From 0edd0d43033fc4b05157d6e2ef2a567cc03b6e84 Mon Sep 17 00:00:00 2001 From: Stephen Parente Date: Thu, 11 Aug 2016 11:10:43 -0700 Subject: [PATCH] Initial commit of node-rdkafka --- .editorconfig | 5 + .gitignore | 11 + .gitmodules | 3 + .jshintignore | 1 + .jshintrc | 23 + .npmignore | 6 + .travis.yml | 20 + CHANGES.md | 5 + CONTRIBUTING.md | 158 + LICENSE.txt | 20 + Makefile | 82 + README.md | 384 ++ bench/consumer-raw-rdkafka.js | 79 + bench/consumer-rdkafka.js | 106 + bench/producer-raw-rdkafka.js | 132 + bench/producer-rdkafka.js | 120 + bench/seed.sh | 38 + binding.gyp | 61 + cpplint.py | 6323 +++++++++++++++++++++++++++++++ deps/librdkafka | 1 + deps/librdkafka.gyp | 188 + docker-compose.yml | 14 + e2e/both.spec.js | 189 + e2e/consumer.spec.js | 60 + e2e/producer.spec.js | 156 + e2e/test-case.js | 115 + examples/consumer.md | 56 + examples/metadata.md | 27 + examples/producer-cluster.md | 101 + examples/producer-multiple.md | 37 + examples/producer.md | 85 + index.js | 12 + kafka-native.js | 12 + lib/client.js | 362 ++ lib/error.js | 233 ++ lib/index.js | 26 + lib/kafka-consumer.js | 455 +++ lib/producer.js | 320 ++ lib/util/topicReadable.js | 206 + lib/util/topicWritable.js | 210 + make_docs.sh | 47 + package.json | 44 + src/binding.cc | 41 + src/binding.h | 23 + src/callbacks.cc | 510 +++ src/callbacks.h | 191 + src/common.cc | 117 + src/common.h | 50 + src/config.cc | 77 + src/config.h | 31 + src/connection.cc | 197 + src/connection.h | 88 + src/consumer.cc | 801 ++++ src/consumer.h | 115 + src/errors.cc | 46 + src/errors.h | 45 + src/message.cc | 135 + src/message.h | 64 + src/producer.cc | 499 +++ src/producer.h | 85 + src/topic.cc | 163 + src/topic.h | 52 + src/workers.cc | 743 ++++ src/workers.h | 318 ++ test/binding.spec.js | 36 + test/consumer.spec.js | 86 + test/error.spec.js | 8 + test/index.spec.js | 8 + test/mocha.opts | 1 + test/mock.js | 55 + test/producer.spec.js | 65 + test/util/topicReadable.spec.js | 123 + test/util/topicWritable.spec.js | 135 + 73 files changed, 15441 insertions(+) create mode 100644 .editorconfig create mode 100644 .gitignore create mode 100644 .gitmodules create mode 100644 .jshintignore create mode 100644 .jshintrc create mode 100644 .npmignore create mode 100644 .travis.yml create mode 100644 CHANGES.md create mode 100644 CONTRIBUTING.md create mode 100644 LICENSE.txt create mode 100644 Makefile create mode 100644 README.md create mode 100644 bench/consumer-raw-rdkafka.js create mode 100644 bench/consumer-rdkafka.js create mode 100644 bench/producer-raw-rdkafka.js create mode 100644 bench/producer-rdkafka.js create mode 100755 bench/seed.sh create mode 100644 binding.gyp create mode 100644 cpplint.py create mode 160000 deps/librdkafka create mode 100644 deps/librdkafka.gyp create mode 100644 docker-compose.yml create mode 100644 e2e/both.spec.js create mode 100644 e2e/consumer.spec.js create mode 100644 e2e/producer.spec.js create mode 100644 e2e/test-case.js create mode 100644 examples/consumer.md create mode 100644 examples/metadata.md create mode 100644 examples/producer-cluster.md create mode 100644 examples/producer-multiple.md create mode 100644 examples/producer.md create mode 100644 index.js create mode 100644 kafka-native.js create mode 100644 lib/client.js create mode 100644 lib/error.js create mode 100644 lib/index.js create mode 100644 lib/kafka-consumer.js create mode 100644 lib/producer.js create mode 100644 lib/util/topicReadable.js create mode 100644 lib/util/topicWritable.js create mode 100755 make_docs.sh create mode 100644 package.json create mode 100644 src/binding.cc create mode 100644 src/binding.h create mode 100644 src/callbacks.cc create mode 100644 src/callbacks.h create mode 100644 src/common.cc create mode 100644 src/common.h create mode 100644 src/config.cc create mode 100644 src/config.h create mode 100644 src/connection.cc create mode 100644 src/connection.h create mode 100644 src/consumer.cc create mode 100644 src/consumer.h create mode 100644 src/errors.cc create mode 100644 src/errors.h create mode 100644 src/message.cc create mode 100644 src/message.h create mode 100644 src/producer.cc create mode 100644 src/producer.h create mode 100644 src/topic.cc create mode 100644 src/topic.h create mode 100644 src/workers.cc create mode 100644 src/workers.h create mode 100644 test/binding.spec.js create mode 100644 test/consumer.spec.js create mode 100644 test/error.spec.js create mode 100644 test/index.spec.js create mode 100644 test/mocha.opts create mode 100644 test/mock.js create mode 100644 test/producer.spec.js create mode 100644 test/util/topicReadable.spec.js create mode 100644 test/util/topicWritable.spec.js diff --git a/.editorconfig b/.editorconfig new file mode 100644 index 00000000..00098894 --- /dev/null +++ b/.editorconfig @@ -0,0 +1,5 @@ +[*] +indent_style = space +indent_size = 2 +trim_trailing_whitespace = true +insert_final_newline = true diff --git a/.gitignore b/.gitignore new file mode 100644 index 00000000..1073c710 --- /dev/null +++ b/.gitignore @@ -0,0 +1,11 @@ +build/ +node_modules/ +deps/librdkafka +npm-debug.log + +docs + +deps/* +!deps/*.gyp + +.DS_Store diff --git a/.gitmodules b/.gitmodules new file mode 100644 index 00000000..4d4e7fb9 --- /dev/null +++ b/.gitmodules @@ -0,0 +1,3 @@ +[submodule "deps/librdkafka"] + path = deps/librdkafka + url = https://github.com/edenhill/librdkafka.git diff --git a/.jshintignore b/.jshintignore new file mode 100644 index 00000000..b43bf86b --- /dev/null +++ b/.jshintignore @@ -0,0 +1 @@ +README.md diff --git a/.jshintrc b/.jshintrc new file mode 100644 index 00000000..09968b8b --- /dev/null +++ b/.jshintrc @@ -0,0 +1,23 @@ +{ + "node": true, + "mocha": true, + "browser": false, + "boss": true, + "curly": true, + "debug": false, + "devel": false, + "eqeqeq": true, + "evil": true, + "forin": false, + "latedef": false, + "noarg": true, + "nonew": true, + "nomen": false, + "onevar": false, + "plusplus": false, + "regexp": false, + "undef": true, + "strict": false, + "white": false, + "eqnull": true +} diff --git a/.npmignore b/.npmignore new file mode 100644 index 00000000..8d384fcd --- /dev/null +++ b/.npmignore @@ -0,0 +1,6 @@ +deps/ +!deps/librdkafka.gyp +!deps/librdkafka +.gitmodules +Dockerfile +deps/librdkafka/config.h diff --git a/.travis.yml b/.travis.yml new file mode 100644 index 00000000..ea2629cf --- /dev/null +++ b/.travis.yml @@ -0,0 +1,20 @@ +dist: trusty +language: node_js +sudo: required +node_js: +- '4' +- '6' +cache: + directories: + - node_modules +services: +- docker +before_install: +- docker-compose up -d +script: +- make lint +- make test +- make e2e +after_success: +- make docs +- echo "Disallow: /" >> docs/robots.txt diff --git a/CHANGES.md b/CHANGES.md new file mode 100644 index 00000000..33532a15 --- /dev/null +++ b/CHANGES.md @@ -0,0 +1,5 @@ +## Changelog + +### 0.3.0 + +* Initial release! diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md new file mode 100644 index 00000000..b3d7fc8e --- /dev/null +++ b/CONTRIBUTING.md @@ -0,0 +1,158 @@ +# Contributing to `node-rdkafka` + +:+1::tada: First off, thanks for taking the time to contribute! :tada::+1: + +The following is a set of guidelines for contributing to `node-rdkafka` +which is hosted in the [Blizzard Organization](https://github.com/blizzard) +on GitHub. This document lists rules, guidelines, and help getting started, +so if you feel something is missing feel free to send a pull request. + +#### Table Of Contents + +[What should I know before I get started?](#what-should-i-know-before-i-get-started) + * [Contributor Agreement](#contributor-agreement) + +[How Can I Contribute?](#how-can-i-contribute) + * [Reporting Bugs](#reporting-bugs) + * [Suggesting Enhancements](#suggesting-enhancements) + * [Pull Requests](#pull-requests) + +[Styleguides](#styleguides) + * [Git Commit Messages](#git-commit-messages) + * [JavaScript Styleguide](#javascript-styleguide) + * [C++ Styleguide](#c++-styleguide) + * [Specs Styleguide](#specs-styleguide) + * [Documentation Styleguide](#documentation-styleguide) + +[Debugging](#debugging) + * [Debugging C++](#debugging-c) + +## What should I know before I get started? + +### Contributor Agreement + +Not currently required. + +## How can I contribute? + +### Reporting Bugs + +Please use __Github Issues__ to report bugs. When filling out an issue report, +make sure to copy any related code and stack traces so we can properly debug. +We need to be able to reproduce a failing test to be able to fix your issue +most of the time, so a custom written failing test is very helpful. + +Please also note the Kafka broker version that you are using and how many +replicas, partitions, and brokers you are connecting to, because some issues +might be related to Kafka. A list of `librdkafka` configuration key-value pairs +also helps. + +### Suggesting Enhancements + +Please use __Github Issues__ to suggest enhancements. We are happy to consider +any extra functionality or features to the library, as long as they add real +and related value to users. Describing your use case and why such an addition +helps the user base can help guide the decision to implement it into the +library's core. + +### Pull Requests + +* Include new test cases (either end-to-end or unit tests) with your change. +* Follow our style guides. +* Make sure all tests are still passing and the `linter` does not report any issues. +* End files with a new line. +* Document the new code in the comments (if it is JavaScript) so the + documentation generator can update the reference documentation. +* Avoid platform-dependent code. +
**Note:** If making modifications to the underlying C++, please use built-in + precompiler directives to detect such platform specificities. Use `Nan` + whenever possible to abstract node/v8 version incompatibility. +* Make sure your branch is up to date and rebased. +* Squash extraneous commits unless their history truly adds value to the library. + +## Styleguides + +### General style guidelines + +Download the [EditorConfig](http://editorconfig.org) plugin for your preferred +text editor to automate the application of the following guidelines: + +* Use 2-space indent (no tabs). +* Do not leave trailing whitespace on lines. +* Files should end with a final newline. + +Also, adhere to the following not enforced by EditorConfig: + +* Limit lines to 80 characters in length. A few extra (<= 5) is fine if it helps + readability, use good judgement. +* Use `lf` line endings. (git's `core.autocrlf` setting can help) + +### Git Commit Messages + +Commit messages should adhere to the guidelines in tpope's +[A Note About Git Commit Messages](http://tbaggery.com/2008/04/19/a-note-about-git-commit-messages.html) + +In short: + +* Use the imperative mood. ("Fix bug", not "Fixed bug" or "Fixes bug") +* Limit the first line to 50 characters or less, followed by a blank line + and detail paragraphs (limit detail lines to about 72 characters). +* Reference issue numbers or pull requests whenever possible. + +### JavaScript Styleguide + +* Place `module.exports` at or near the top of the file. + * Defined functions are hoisted, so it is appropriate to define the + function after you export it. + * When exporting an object, define it first, then export it, and then add + methods or properties. +* Do not use ES2015 specific features (for example, do not use `let`, `const`, + or `class`). +* All callbacks should follow the standard Node.js callback signature. +* Your JavaScript should properly pass the linter (`make jslint`). + +### C++ Styleguide + +* Class member variables should be prefixed with `m_`. +* Use a comment when pointer ownership has changed hands. +* Your C++ should properly pass the `cpplint.py` in the `make lint` test. + +### Specs Styleguide + +* Write all JavaScript tests by using the `mocha` testing framework. +* All `mocha` tests should use exports syntax. +* All `mocha` test files should be suffixed with `.spec.js` instead of `.js`. +* Unit tests should mirror the JavaScript files they test (for example, + `lib/client.js` is tested in `test/client.spec.js`). +* Unit tests should have no outside service dependencies. Any time a dependency, + like Kafka, exists, you should create an end-to-end test. +* You may mock a connection in a unit test if it is reliably similar to its real + variant. + +### Documentation Styleguide + +* Write all JavaScript documentation in jsdoc-compatible inline comments. +* Each docblock should have references to return types and parameters. If an + object is a parameter, you should also document any required subproperties. +* Use `@see` to reference similar pieces of code. +* Use comments to document your code when its intent may be difficult to understand. +* All documentation outside of the code should be in Github-compatible markdown. +* Make good use of font variations like __bold__ and *italics*. +* Use headers and tables of contents when they make sense. + +## Debugging + +### Debugging C++ + +Use `gdb` for debugging (as shown in the following example). + +``` +node-gyp rebuild --debug + +gdb node +(gdb) set args "path/to/file.js" +(gdb) run +[output here] +``` + +You can add breakpoints and so on after that. diff --git a/LICENSE.txt b/LICENSE.txt new file mode 100644 index 00000000..d1a4fa95 --- /dev/null +++ b/LICENSE.txt @@ -0,0 +1,20 @@ +The MIT License (MIT) +Copyright (c) + +Permission is hereby granted, free of charge, to any person obtaining a copy of +this software and associated documentation files (the "Software"), to deal in +the Software without restriction, including without limitation the rights to +use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies +of the Software, and to permit persons to whom the Software is furnished to do +so, subject to the following conditions: + +The above copyright notice and this permission notice shall be included in all +copies or substantial portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS +IN THE SOFTWARE. diff --git a/Makefile b/Makefile new file mode 100644 index 00000000..b8b9c7b8 --- /dev/null +++ b/Makefile @@ -0,0 +1,82 @@ +NODE-GYP ?= node_modules/.bin/node-gyp +PYTHON ?= python +NODE ?= node +CPPLINT ?= cpplint.py +BUILDTYPE ?= Release +TESTS = "test/**/*.js" +E2E_TESTS = $(wildcard e2e/*.spec.js) +TEST_REPORTER = +TEST_OUTPUT = +CONFIG_OUTPUTS = \ + build/bindings.target.mk \ + build/Makefile \ + build/binding.Makefile build/config.gypi + +CPPLINT_FILES = $(wildcard src/*.cc src/*.h) +CPPLINT_FILTER = -legal/copyright +JSLINT_FILES = $(wildcard lib/**/*.js test/**/*.js) + +PACKAGE = $(shell node -pe 'require("./package.json").name.split("/")[1]') +VERSION = $(shell node -pe 'require("./package.json").version') + +GYPBUILDARGS= +ifeq ($(BUILDTYPE),Debug) +GYPBUILDARGS=--debug +endif + +.PHONY: all clean lint test lib docs e2e ghpages + +all: lint lib test e2e + +lint: cpplint jslint + +cpplint: + @$(PYTHON) $(CPPLINT) --filter=$(CPPLINT_FILTER) $(CPPLINT_FILES) + +jslint: node_modules/.dirstamp + @./node_modules/.bin/jshint $(JSLINT_FILES) + +lib: node_modules/.dirstamp $(CONFIG_OUTPUTS) + @$(NODE-GYP) build $(GYPBUILDARGS) + +node_modules/.dirstamp: package.json + @npm update --loglevel warn + @touch $@ + +$(CONFIG_OUTPUTS): node_modules/.dirstamp binding.gyp + @$(NODE-GYP) configure + +test: node_modules/.dirstamp + @./node_modules/.bin/mocha $(TEST_REPORTER) $(TESTS) $(TEST_OUTPUT) + +e2e: $(E2E_TESTS) + @$(NODE) e2e/consumer.spec.js && $(NODE) e2e/producer.spec.js && $(NODE) e2e/both.spec.js + + +define release + NEXT_VERSION=$(shell node -pe 'require("semver").inc("$(VERSION)", "$(1)")') + node -e "\ + var j = require('./package.json');\ + j.version = \"$$NEXT_VERSION\";\ + var s = JSON.stringify(j, null, 2);\ + require('fs').writeFileSync('./package.json', s);" && \ + git commit -m "release $$NEXT_VERSION" -- package.json && \ + git tag "$$NEXT_VERSION" -m "release $$NEXT_VERSION" +endef + +docs: node_modules/.dirstamp + @rm -rf docs + @./node_modules/jsdoc/jsdoc.js --destination docs \ + --recurse -R ./README.md \ + -t "./node_modules/toolkit-jsdoc/" \ + --tutorials examples ./lib + +gh-pages: node_modules/.dirstamp + @./make_docs.sh + +release-patch: + @$(call release,patch) + +clean: node_modules/.dirstamp + @rm -f deps/librdkafka/config.h + @$(NODE-GYP) clean diff --git a/README.md b/README.md new file mode 100644 index 00000000..6b382bfc --- /dev/null +++ b/README.md @@ -0,0 +1,384 @@ +node-rdkafka - Node.js wrapper for Kafka C/C++ library +============================================== + +Copyright (c) 2016 Blizzard Entertainment. + +[https://github.com/blizzard/node-rdkafka](https://github.com/blizzard/node-rdkafka) + +[![Build Status](https://travis-ci.org/Blizzard/node-rdkafka.svg?branch=master)](https://travis-ci.org/Blizzard/node-rdkafka) +[![npm version](https://badge.fury.io/js/node-rdkafka.svg)](https://badge.fury.io/js/node-rdkafka) + +# Overview + +The `node-rdkafka` library is a high-performance NodeJS client for [Apache Kafka](http://kafka.apache.org/) that wraps the native [librdkafka](https://github.com/edenhill/librdkafka) library. All the complexity of balancing writes across partitions and managing (possibly ever-changing) brokers should be encapsulated in the library. + +## Reference Docs + +To view the reference docs for the current version, go [here](https://blizzard.github.io/node-rdkafka/current/) + +## Tests + +This project includes two types of unit tests in this project: +* end-to-end integration tests +* unit tests + +You can run both types of tests by using `Makefile`. Doing so calls `mocha` in your locally installed `node_modules` directory. + +* To run the unit tests, you can run `make lint` or `make test`. +* To run the integration tests, you must have a running Kafka installation available. By default, the test tries to connect to `localhost:9092`; however, you can supply the `KAFKA_HOST` environment variable to override this default behavior. + +## Contributing + +For guidelines on contributing please see [CONTRIBUTING.md](https://github.com/blizzard/node-rdkafka/blob/master/CONTRIBUTING.md) + +# Usage + +You can install the `node-rdkafka` module like any other module: + +``` +npm install node-rdkafka +``` + +To use the module, you must `require` it. + +```js +var Kafka = require('node-rdkafka'); +``` + +## Configuration + +You can pass many configuration options to `librdkafka`. A full list can be found at https://github.com/edenhill/librdkafka/blob/master/CONFIGURATION.md. + +Configuration keys that have the suffix `_cb` are designated as callbacks. Some +of these keys are informational and you can choose to opt-in (for example, `dr_cb`). Others are callbacks designed to +return a value, such as `partitioner_cb`. + +Not all of these options are supported. +The library will throw an error if the value you send in is invalid. + +The library currently supports the following callbacks: +* `partitioner_cb` +* `dr_cb` +* `event_cb` + +## Sending Messages + +A `Producer` sends messages to Kafka. The `Producer` constructor takes a configuration object, as shown in the following example: + +```js +var producer = new Kafka.Producer({ + 'metadata.broker.list': 'kafka-host1:9092,kafka-host2:9092' +}); +``` + +A `Producer` requires only `metadata.broker.list` (the Kafka brokers) to be created. The values in this list are separated by commas. For other configuration options, see the [Configuration.md](https://github.com/edenhill/librdkafka/blob/master/CONFIGURATION.md) file described previously. + +The following example illustrates a list with several `librdkafka` options set. + +```js +var producer = new Kafka.Producer({ + 'client.id': 'kafka', + 'metadata.broker.list': 'localhost:9092', + 'compression.codec': 'gzip', + 'retry.backoff.ms': 200, + 'message.send.max.retries': 10, + 'socket.keepalive.enable': true, + 'queue.buffering.max.messages': 100000, + 'queue.buffering.max.ms': 1000, + 'batch.num.messages': 1000000, + 'dr_cb': true +}); +``` + +#### Stream API + +You can easily use the `Producer` as a writable stream immediately after creation (as shown in the following example): + +```js +// Our producer with its Kafka brokers +var producer = new Kafka.Producer({ + 'metadata.broker.list': 'kafka-host1:9092,kafka-host2:9092' +}); + +// This call returns a new writable stream to our topic 'topic-name' +var stream = producer.getWriteStream('topic-name'); + +// Writes a message to the stream +var queuedSuccess = stream.write(new Buffer('Awesome message')); + +if (queuedSuccess) { + console.log('We queued our message!'); +} else { + // Note that this only tells us if the stream's queue is full, + // it does NOT tell us if the message got to Kafka! See below... + console.log('Too many messages in our queue already'); +} + +stream.on('error', function (err) { + // Here's where we'll know if something went wrong sending to Kafka + console.error('Error in our kafka stream'); + console.error(err); +}) +``` + +Note that `getWriteStream` will create a new stream on every call. You should try to cache the returned stream for a topic after the first call. + +#### Standard API + +The Standard API is more performant, particularly when handling high volumes of messages. +However, it requires more manual setup to use. The following example illustrates its use: + +```js +var producer = new Kafka.Producer({ + 'metadata.broker.list': 'localhost:9092', + 'dr_cb': true +}); + +// Connect to the broker manually +producer.connect(); + +// Wait for the ready event before proceeding +producer.on('ready', function() { + // Create a Topic object with any options our Producer + // should use when writing to that topic. + var topic = producer.Topic('topic', { + // Make the Kafka broker acknowledge our message (optional) + 'request.required.acks': 1 + }); + + producer.produce({ + // Message to send. If a string is supplied, it will be + // converted to a Buffer automatically, but we're being + // explicit here for the sake of example. + message: new Buffer('Awesome message'), + + // The topic object we created above + topic: topic + }, function(err) { + // Called after the message is queued + if (err) { + console.error('A problem occurred when sending our message'); + console.error(err); + } else { + console.log('Message produced successfully!'); + } + }); +}); + +// Any errors we encounter, including connection errors +producer.on('error', function(err) { + console.error('Error from producer'); + console.error(err); +}) +``` + +To see the configuration options available to you, see the [Configuration](#configuration) section. + +##### Methods + +|Method|Description| +|-------|----------| +|`producer.connect()`| Connects to the broker.

The `connect()` method emits the `ready` event when it connects successfully or an `error` when it does not.| +|`producer.disconnect()`| Disconnects from the broker.

The `disconnect()` method emits the `disconnected` event when it has disconnected or `error` if something went wrong. | +|`producer.poll()` | Polls the producer for delivery reports or other events to be transmitted via the emitter.

This happens automatically on transactions such as `produce`. | +|`producer.produce(msg)`| Sends a message.

The `produce()` method takes a JSON object in the format showed above. | + +##### Events + +Some configuration properties that end in `_cb` indicate that an event should be generated for that option. You can either: + +* provide a value of `true` and react to the event +* provide a callback function directly + +The following example illustrates an event: + +```js +var producer = new Kafka.Producer({ + 'client.id': 'my-client', // Specifies an identifier to use to help trace activity in Kafka + 'metadata.broker.list': 'localhost:9092', // Connect to a Kafka instance on localhost + 'dr_cb': true // Specifies that we want a delivery-report event to be generated +}); + +producer.on('delivery-report', function(report) { + // Report of delivery statistics here: + // + console.log(report); +}); +``` + +The following table describes types of events. + +|Event|Description| +|-------|----------| +| `error` | Error reporting is handled through this pipeline.

Most errors will have a value for `code`, `message`, and `origin`. `origin` will be `local` or `kafka` to determine where the error happened. | +| `disconnected` | The `disconnected` event is emitted when the broker has disconnected.

This event is emitted only when `.disconnect` is called. The wrapper will always try to reconnect otherwise. | +| `ready` | The `ready` event is emitted when the `Producer` is ready to send messages. | +| `event` | The `event` event is emitted when `librdkafka` reports an event (if you opted in via the `event_cb` option). | +| `event.log` | The `event.log` event is emitted when logging events come in (if you opted into logging via the `event_cb` option).

You will need to set a value for `debug` if you want to send information. | +| `event.status` | The `event.status` event is emitted when `librdkafka` reports stats (if you opted in). | +| `event.throttle` | The `event.throttle` event emitted when `librdkafka` reports throttling. | +| `delivery-report` | The `delivery-report` event is emitted when a delivery report has been found via polling.

To use this event, you must set `request.required.acks` to `1` or `-1` in topic configuration and `dr_cb` to `true` in the `Producer` constructor options. | + +## Kafka.KafkaConsumer + +To read messages from Kafka, you use a `KafkaConsumer`. You instantiate a `KafkaConsumer` object as follows: + +```js +var consumer = new Kafka.KafkaConsumer({ + 'group.id': 'kafka', + 'metadata.broker.list': 'localhost:9092', +}, {}); +``` + +The first parameter is the global config, while the second parameter is the topic config that gets applied to all subscribed topics. To view a list of all supported configuration properties, see the [Configuration.md](https://github.com/edenhill/librdkafka/blob/master/CONFIGURATION.md) file described previously. Look for the `C` and `*` keys. + +The `group.id` and `metadata.broker.list` properties are required for a consumer. + +### Message Structure + +Messages that are returned by the `KafkaConsumer` have the following structure. + +```js +{ + message: new Buffer('hi'), // message contents as a Buffer + size: 2, // size of the message, in bytes + topic: 'librdtesting-01', // topic the message comes from + offset: 1337, // offset the message was read from + partition: 1 // partition the message was on +} +``` + +### Stream API + +The stream API is the easiest way to consume messages. The following example illustrates the use of the stream API: + +```js +// Read from the librdtesting-01 topic... note that this creates a new stream on each call! +var stream = consumer.getReadStream('librdtesting-01'); + +stream.on('data', function(data) { + console.log('Got message'); + console.log(data.message.toString()); +}); +``` + +### Standard API + +You can also use the Standard API and manage callbacks and events yourself. You can choose different modes for consuming messages: + +* *Flowing mode*. This mode flows all of the messages it can read by maintaining an infinite loop in the event loop. It only stops when it detects the consumer has issued the `unsubscribe` or `disconnect` method. +* *Non-flowing mode*. This mode reads a single message from Kafka at a time manually. + +The following example illustrates flowing mode: +```js +// Flowing mode +consumer.connect(); + +consumer + .on('ready', function() { + // Consume from the librdtesting-01 topic. This is what determines + // the mode we are running in. By consuming an entire topic, + // we will get messages from that topic as soon as they are available + consumer.consume('librdtesting-01'); + }) + .on('data', function(data) { + // Output the actual message contents + console.log(data.message.toString()); + }); +``` +The following example illustrates non-flowing mode: +```js +// Non-flowing mode +consumer.connect(); + +consumer + .on('ready', function() { + // Subscribe to the librdtesting-01 topic + // This makes subsequent consumes read from that topic. + consumer.subscribe('librdtesting-01'); + + // Read one message every 1000 seconds + setInterval(function() { + consumer.consume(); + }, 1000); + }) + .on('data', function(data) { + console.log('Message found! Contents below.'); + console.log(data.message.toString()); + }); +``` + +The following table lists important methods for this API. + +|Method|Description| +|-------|----------| +|`consumer.connect()` | Connects to the broker.

The `connect()` emits the event `ready` when it has successfully connected, or an `error` when it has not. | +|`consumer.disconnect()` | Disconnects from the broker.

The `disconnect()` method emits `disconnected` when it has disconnected or `error` if something went wrong. +|`consumer.subscribe(topics, callback)` | Subscribes to an array of topics.

`topics` can be either an array or a string for a single topic. | +|`consumer.unsubscribe()` | Unsubscribes from the currently subscribed topics.

You cannot subscribe to different topics without calling the `unsubscribe()` method first. | +|`consumer.consume(cb)` | Gets a message from the existing subscription. | +|`consumer.consume(topics, cb)` | Creates a subscription and get messages as they become available.

The `consume()` method keeps a background thread running to do the work. | + +The following table lists events for this API. + +|Event|Description| +|-------|----------| +|`error` | Error reporting is handled through this pipeline.

Most errors will have a `code`, `message`, and `origin` value. The `origin` value will be **local** or **remote** to determine where the error happened. | +|`disconnected` | The `disconnected` event is emitted when the broker disconnects.

This event is only emitted when `.disconnect` is called. The wrapper will always try to reconnect otherwise. | +|`ready` | The `ready` event is emitted when the `Producer` is ready to send messages. | +|`event` | The `event` event is emitted when `librdkafka` reports an event (if you opted in via the `event_cb` option).| +|`event.log` | The `event.log` event is emitted when logging events occur (if you opted in for logging via the `event_cb` option).

You will need to set a value for `debug` if you want information to send. | +|`event.status` | The `event.status` event is emitted when `librdkafka` reports stats (if you opted in). | +|`event.throttle` | The `event.throttle` event is emitted when `librdkafka` reports throttling.| + +## Metadata + +Both `Kafka.Producer` and `Kafka.KafkaConsumer` include a `getMetadata` method to retrieve metadata from Kafka. + +Getting metadata on any connection returns the following data structure: + +```js +{ + orig_broker_id: 1, + orig_broker_name: "broker_name", + brokers: [ + { + id: 1, + host: 'localhost', + port: 40 + } + ], + topics: [ + { + name: 'awesome-topic', + partitions: [ + { + id: 1, + leader: 20, + replicas: [1, 2], + isrs: [1, 2] + } + ] + } + ] +} +``` + +The following example illustrates how to use the `getMetadata` method. + +```js +var opts = { + topic: 'librdtesting-01', + timeout: 10000 +}; + +producer.getMetadata(opts, function(err, metadata) { + if (err) { + console.error('Error getting metadata'); + console.error(err); + } else { + console.log('Got metadata'); + console.log(metadata); + } +}); +``` diff --git a/bench/consumer-raw-rdkafka.js b/bench/consumer-raw-rdkafka.js new file mode 100644 index 00000000..d1d30576 --- /dev/null +++ b/bench/consumer-raw-rdkafka.js @@ -0,0 +1,79 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +var Kafka = require('../'); +var count = 0; +var total = 0; +var store = []; +var host = process.argv[2] || 'localhost:9092'; +var topic = process.argv[3] || 'test'; + +var consumer = new Kafka.KafkaConsumer({ + 'metadata.broker.list': host, + 'group.id': 'node-rdkafka-bench', + 'fetch.wait.max.ms': 100, + 'fetch.message.max.bytes': 1024 * 1024, + 'enable.auto.commit': false + // paused: true, +}, { + 'auto.offset.reset': 'earliest' +}); + +var interval; + +consumer.connect() + .once('ready', function() { + consumer.consume([topic]); + }) + .on('rebalance', function() { + console.log('rebalance'); + }) + .once('data', function() { + interval = setInterval(function() { + console.log('%d messages per second', count); + if (count > 0) { + store.push(count); + } + count = 0; + }, 1000); + }) + .on('data', function(message) { + count += 1; + total += 1; + }); + +process.once('SIGTERM', shutdown); +process.once('SIGINT', shutdown); +process.once('SIGHUP', shutdown); + +function shutdown() { + clearInterval(interval); + + if (store.length > 0) { + var calc = 0; + for (var x in store) { + calc += store[x]; + } + + var mps = parseFloat(calc * 1.0/store.length); + + console.log('%d messages per second on average', mps); + + } + + var killTimer = setTimeout(function() { + process.exit(); + }, 5000); + + consumer.disconnect(function() { + clearTimeout(killTimer); + process.exit(); + }); + +} diff --git a/bench/consumer-rdkafka.js b/bench/consumer-rdkafka.js new file mode 100644 index 00000000..54588fdc --- /dev/null +++ b/bench/consumer-rdkafka.js @@ -0,0 +1,106 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +var Writable = require('stream').Writable; + +var Kafka = require('../'); +var count = 0; +var total = 0; +var store = []; +var host = process.argv[2] || 'localhost:9092'; +var topic = process.argv[3] || 'telemetry-v2-ingest'; + +var consumer = new Kafka.KafkaConsumer({ + 'metadata.broker.list': host, + 'group.id': 'node-rdkafka-bench', + 'fetch.wait.max.ms': 100, + 'fetch.message.max.bytes': 1024 * 1024, + 'enable.auto.commit': false + // paused: true, +}, { + 'auto.offset.reset': 'earliest' +}); + +// Track how many messages we see per second +var interval; + +consumer.on('rebalance', function() { + console.log('Consumer has been rebalanced'); + interval = setInterval(function() { + if (isShuttingDown) { + clearInterval(interval); + } + console.log('%d messages per second', count); + if (count > 0) { + // Don't store ones when we didn't get data i guess? + store.push(count); + // setTimeout(shutdown, 500); + } + count = 0; + }, 1000).unref(); +}); + +var stream = consumer.getReadStream(topic, { + fetchSize: 16 +}); +var isShuttingDown = false; + +stream + .on('error', function(err) { + console.log('Shutting down due to error'); + console.log(err.stack); + shutdown(); + }) + .on('end', function() { + // Can be called more than once without issue because of guard var + console.log('Shutting down due to stream end'); + shutdown(); + }) + .pipe(new Writable({ + objectMode: true, + write: function(batchMessages, encoding, cb) { + if (Array.isArray(batchMessages)) { + count += batchMessages.length; + total += batchMessages.length; + } else { + count += 1; + total += 1; + } + setImmediate(cb); + } + })); + +process.once('SIGTERM', shutdown); +process.once('SIGINT', shutdown); +process.once('SIGHUP', shutdown); + +function shutdown() { + if (isShuttingDown) return; + clearInterval(interval); + isShuttingDown = true; + if (store.length > 0) { + var calc = 0; + for (var x in store) { + calc += store[x]; + } + + var mps = parseFloat(calc * 1.0/store.length); + + console.log('%d messages per second on average', mps); + } + + var killTimer = setTimeout(function() { + process.exit(); + }, 5000).unref(); + + consumer.disconnect(function() { + console.log('total: %d', total); + }); + +} diff --git a/bench/producer-raw-rdkafka.js b/bench/producer-raw-rdkafka.js new file mode 100644 index 00000000..4babd03e --- /dev/null +++ b/bench/producer-raw-rdkafka.js @@ -0,0 +1,132 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +var Kafka = require('../'); +var crypto = require('crypto'); +var count = 0; +var total = 0; +var totalComplete = 0; +var verifiedComplete = 0; +var errors = 0; +var store = []; +var started; +var done = false; +var host = process.argv[2] || '127.0.0.1:9092'; +var topicName = process.argv[3] || 'test'; +var compression = process.argv[4] || 'gzip'; +var MAX = process.argv[5] || 100000; + +var producer = new Kafka.Producer({ + 'metadata.broker.list': host, + 'group.id': 'node-rdkafka-bench', + 'compression.codec': compression, + 'retry.backoff.ms': 200, + 'message.send.max.retries': 10, + 'socket.keepalive.enable': true, + 'queue.buffering.max.messages': 100000, + 'queue.buffering.max.ms': 1000, + 'batch.num.messages': 1000, + // paused: true, +}); + +// Track how many messages we see per second +var interval; +var ok = true; + +function getTimer() { + if (!interval) + interval = setTimeout(function() { + interval = false; + if (!done) { + console.log('%d messages per sent second', count); + store.push(count); + count = 0; + getTimer(); + + } else { + console.log('%d messages remaining sent in last batch <1000ms', count); + } + }, 1000); + + return interval; +} + +var t; + +crypto.randomBytes(4096, function(ex, buffer) { + + producer.connect() + .on('ready', function() { + var topic = producer.Topic(topicName, {}); + getTimer(); + + started = new Date().getTime(); + + var x = function(e) { + if (e) { + console.log(e); + errors += 1; + } else { + verifiedComplete += 1; + } + count += 1; + totalComplete += 1; + if (totalComplete === MAX) { + shutdown(); + } + }; + + var sendMessage = function() { + producer.produce({ + topic: topic, + message: buffer + }, x); + if (total < MAX) { + total += 1; + setImmediate(sendMessage); + } + }; + + sendMessage(); + + }) + .on('error', function(err) { + console.error(err); + process.exit(1); + }) + .on('disconnected', shutdown); + +}); + +process.once('SIGTERM', shutdown); +process.once('SIGINT', shutdown); +process.once('SIGHUP', shutdown); + +function shutdown(e) { + done = true; + + clearInterval(interval); + + var killTimer = setTimeout(function() { + process.exit(); + }, 5000); + + producer.disconnect(function() { + clearTimeout(killTimer); + var ended = new Date().getTime(); + var elapsed = ended - started; + + // console.log('Ended %s', ended); + console.log('total: %d messages over %d ms', total, elapsed); + + console.log('%d messages / second', parseInt(total / (elapsed / 1000))); + process.exit(); + }); + +} diff --git a/bench/producer-rdkafka.js b/bench/producer-rdkafka.js new file mode 100644 index 00000000..0ff2c5d0 --- /dev/null +++ b/bench/producer-rdkafka.js @@ -0,0 +1,120 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +var Kafka = require('../'); +var crypto = require('crypto'); +var count = 0; +var total = 0; +var totalComplete = 0; +var store = []; +var host = process.argv[2] || '127.0.0.1:9092'; +var topicName = process.argv[3] || 'test'; +var compression = process.argv[4] || 'gzip'; +var MAX = process.argv[5] || 100000; + +var producer = new Kafka.Producer({ + 'metadata.broker.list': host, + 'group.id': 'node-rdkafka-bench', + 'compression.codec': compression, + 'retry.backoff.ms': 200, + 'message.send.max.retries': 10, + 'socket.keepalive.enable': true, + 'queue.buffering.max.messages': 100000, + 'queue.buffering.max.ms': 1000, + 'batch.num.messages': 1000, +}); + +producer.on('error', function(e) { + console.log(e); + process.exit(1); +}); + +// Track how many messages we see per second +var interval; +var done = false; + +function log() { + console.log('%d messages per sent second', count); + store.push(count); + count = 0; +} + +crypto.randomBytes(4096, function(ex, buffer) { + + var stream = producer.getWriteStream(topicName); + + var x = function(e) { + if (e) { + console.error(e); + } + count += 1; + totalComplete += 1; + if (totalComplete >= MAX && !done) { + done = true; + clearInterval(interval); + setTimeout(shutdown, 5000); + } + }; + + function write() { + if (!stream.write(buffer, 'base64', x)) { + return stream.once('drain', write); + } else { + total++; + } + + if (total < MAX) { + // we are not done + setImmediate(write); + } + + } + + write(); + interval = setInterval(log, 1000); + stream.on('error', function(err) { + console.log(err); + }); + // stream.on('end', shutdown); + +}); + + +process.once('SIGTERM', shutdown); +process.once('SIGINT', shutdown); +process.once('SIGHUP', shutdown); + +function shutdown() { + + if (store.length > 0) { + var calc = 0; + for (var x in store) { + calc += store[x]; + } + + var mps = parseFloat(calc * 1.0/store.length); + + console.log('%d messages per second on average', mps); + console.log('%d messages total', total); + + } + + clearInterval(interval); + + var killTimer = setTimeout(function() { + process.exit(); + }, 5000); + + producer.disconnect(function() { + clearTimeout(killTimer); + console.log('total: %d', total); + process.exit(); + }); + +} diff --git a/bench/seed.sh b/bench/seed.sh new file mode 100755 index 00000000..5395aaa4 --- /dev/null +++ b/bench/seed.sh @@ -0,0 +1,38 @@ +#!/bin/bash +kafka_root=${KAFKA_ROOT:-/opt/kafka} +# Generate and insert some messages + +OS=$(uname -s) + +function initializeTopic { + topic=$1 + host=$2 + msg_size=$3 + batch_size=$4 + batch_count=$5 + + if [ $host == "localhost:9092" ]; then + ${kafka_root}/bin/kafka-topics.sh --create --zookeeper localhost:2181 \ + --replication-factor 1 --partitions 1 --topic ${topic} + fi + + echo "Generating messages (size: ${msg_size})" + : > /tmp/msgs # Truncate /tmp/msgs + for i in $(seq 1 ${batch_size}); do + if [ $OS == 'Darwin' ]; then + printf %s\\n "$(head -c${msg_size} /dev/urandom | base64)" >> /tmp/msgs + else + printf %s\\n "$(head --bytes=${msg_size} /dev/urandom | base64 --wrap=0)" >> /tmp/msgs + fi + done + + echo "Done generating messages" + + for i in $(seq 1 ${batch_count}); do + echo "Adding $(wc -l /tmp/msgs) messages to topic ${topic}" + "${kafka_root}/bin/kafka-console-producer.sh" \ + --broker-list ${host} --topic ${topic} < /tmp/msgs + done +} + +initializeTopic "librdtesting-01" "localhost:9092" "4096" "5000" "2000" diff --git a/binding.gyp b/binding.gyp new file mode 100644 index 00000000..34082be9 --- /dev/null +++ b/binding.gyp @@ -0,0 +1,61 @@ +{ + "targets": [ + { + "target_name": "node-librdkafka", + "sources": [ + "src/common.cc", + "src/errors.cc", + "src/config.cc", + "src/topic.cc", + "src/callbacks.cc", + "src/connection.cc", + "src/message.cc", + "src/consumer.cc", + "src/producer.cc", + "src/workers.cc", + "src/binding.cc" + ], + "dependencies": [ + "<(module_root_dir)/deps/librdkafka.gyp:librdkafka_cpp" + ], + "include_dirs": [ + "<(module_root_dir)/deps/librdkafka/src-cpp", + " [file] ... + + The style guidelines this tries to follow are those in + http://google-styleguide.googlecode.com/svn/trunk/cppguide.xml + + Every problem is given a confidence score from 1-5, with 5 meaning we are + certain of the problem, and 1 meaning it could be a legitimate construct. + This will miss some errors, and is not a substitute for a code review. + + To suppress false-positive errors of a certain category, add a + 'NOLINT(category)' comment to the line. NOLINT or NOLINT(*) + suppresses errors of all categories on that line. + + The files passed in will be linted; at least one file must be provided. + Default linted extensions are .cc, .cpp, .cu, .cuh and .h. Change the + extensions with the --extensions flag. + + Flags: + + output=vs7 + By default, the output is formatted to ease emacs parsing. Visual Studio + compatible output (vs7) may also be used. Other formats are unsupported. + + verbose=# + Specify a number 0-5 to restrict errors to certain verbosity levels. + + filter=-x,+y,... + Specify a comma-separated list of category-filters to apply: only + error messages whose category names pass the filters will be printed. + (Category names are printed with the message and look like + "[whitespace/indent]".) Filters are evaluated left to right. + "-FOO" and "FOO" means "do not print categories that start with FOO". + "+FOO" means "do print categories that start with FOO". + + Examples: --filter=-whitespace,+whitespace/braces + --filter=whitespace,runtime/printf,+runtime/printf_format + --filter=-,+build/include_what_you_use + + To see a list of all the categories used in cpplint, pass no arg: + --filter= + + counting=total|toplevel|detailed + The total number of errors found is always printed. If + 'toplevel' is provided, then the count of errors in each of + the top-level categories like 'build' and 'whitespace' will + also be printed. If 'detailed' is provided, then a count + is provided for each category like 'build/class'. + + root=subdir + The root directory used for deriving header guard CPP variable. + By default, the header guard CPP variable is calculated as the relative + path to the directory that contains .git, .hg, or .svn. When this flag + is specified, the relative path is calculated from the specified + directory. If the specified directory does not exist, this flag is + ignored. + + Examples: + Assuming that src/.git exists, the header guard CPP variables for + src/chrome/browser/ui/browser.h are: + + No flag => CHROME_BROWSER_UI_BROWSER_H_ + --root=chrome => BROWSER_UI_BROWSER_H_ + --root=chrome/browser => UI_BROWSER_H_ + + linelength=digits + This is the allowed line length for the project. The default value is + 80 characters. + + Examples: + --linelength=120 + + extensions=extension,extension,... + The allowed file extensions that cpplint will check + + Examples: + --extensions=hpp,cpp + + cpplint.py supports per-directory configurations specified in CPPLINT.cfg + files. CPPLINT.cfg file can contain a number of key=value pairs. + Currently the following options are supported: + + set noparent + filter=+filter1,-filter2,... + exclude_files=regex + linelength=80 + + "set noparent" option prevents cpplint from traversing directory tree + upwards looking for more .cfg files in parent directories. This option + is usually placed in the top-level project directory. + + The "filter" option is similar in function to --filter flag. It specifies + message filters in addition to the |_DEFAULT_FILTERS| and those specified + through --filter command-line flag. + + "exclude_files" allows to specify a regular expression to be matched against + a file name. If the expression matches, the file is skipped and not run + through liner. + + "linelength" allows to specify the allowed line length for the project. + + CPPLINT.cfg has an effect on files in the same directory and all + sub-directories, unless overridden by a nested configuration file. + + Example file: + filter=-build/include_order,+build/include_alpha + exclude_files=.*\.cc + + The above example disables build/include_order warning and enables + build/include_alpha as well as excludes all .cc from being + processed by linter, in the current directory (where the .cfg + file is located) and all sub-directories. +""" + +# We categorize each error message we print. Here are the categories. +# We want an explicit list so we can list them all in cpplint --filter=. +# If you add a new error message with a new category, add it to the list +# here! cpplint_unittest.py should tell you if you forget to do this. +_ERROR_CATEGORIES = [ + 'build/class', + 'build/c++11', + 'build/deprecated', + 'build/endif_comment', + 'build/explicit_make_pair', + 'build/forward_decl', + 'build/header_guard', + 'build/include', + 'build/include_alpha', + 'build/include_order', + 'build/include_what_you_use', + 'build/namespaces', + 'build/printf_format', + 'build/storage_class', + 'legal/copyright', + 'readability/alt_tokens', + 'readability/braces', + 'readability/casting', + 'readability/check', + 'readability/constructors', + 'readability/fn_size', + 'readability/function', + 'readability/inheritance', + 'readability/multiline_comment', + 'readability/multiline_string', + 'readability/namespace', + 'readability/nolint', + 'readability/nul', + 'readability/strings', + 'readability/todo', + 'readability/utf8', + 'runtime/arrays', + 'runtime/casting', + 'runtime/explicit', + 'runtime/int', + 'runtime/init', + 'runtime/invalid_increment', + 'runtime/member_string_references', + 'runtime/memset', + 'runtime/indentation_namespace', + 'runtime/operator', + 'runtime/printf', + 'runtime/printf_format', + 'runtime/references', + 'runtime/string', + 'runtime/threadsafe_fn', + 'runtime/vlog', + 'whitespace/blank_line', + 'whitespace/braces', + 'whitespace/comma', + 'whitespace/comments', + 'whitespace/empty_conditional_body', + 'whitespace/empty_loop_body', + 'whitespace/end_of_line', + 'whitespace/ending_newline', + 'whitespace/forcolon', + 'whitespace/indent', + 'whitespace/line_length', + 'whitespace/newline', + 'whitespace/operators', + 'whitespace/parens', + 'whitespace/semicolon', + 'whitespace/tab', + 'whitespace/todo', + ] + +# These error categories are no longer enforced by cpplint, but for backwards- +# compatibility they may still appear in NOLINT comments. +_LEGACY_ERROR_CATEGORIES = [ + 'readability/streams', + ] + +# The default state of the category filter. This is overridden by the --filter= +# flag. By default all errors are on, so only add here categories that should be +# off by default (i.e., categories that must be enabled by the --filter= flags). +# All entries here should start with a '-' or '+', as in the --filter= flag. +_DEFAULT_FILTERS = ['-build/include_alpha'] + +# We used to check for high-bit characters, but after much discussion we +# decided those were OK, as long as they were in UTF-8 and didn't represent +# hard-coded international strings, which belong in a separate i18n file. + +# C++ headers +_CPP_HEADERS = frozenset([ + # Legacy + 'algobase.h', + 'algo.h', + 'alloc.h', + 'builtinbuf.h', + 'bvector.h', + 'complex.h', + 'defalloc.h', + 'deque.h', + 'editbuf.h', + 'fstream.h', + 'function.h', + 'hash_map', + 'hash_map.h', + 'hash_set', + 'hash_set.h', + 'hashtable.h', + 'heap.h', + 'indstream.h', + 'iomanip.h', + 'iostream.h', + 'istream.h', + 'iterator.h', + 'list.h', + 'map.h', + 'multimap.h', + 'multiset.h', + 'ostream.h', + 'pair.h', + 'parsestream.h', + 'pfstream.h', + 'procbuf.h', + 'pthread_alloc', + 'pthread_alloc.h', + 'rope', + 'rope.h', + 'ropeimpl.h', + 'set.h', + 'slist', + 'slist.h', + 'stack.h', + 'stdiostream.h', + 'stl_alloc.h', + 'stl_relops.h', + 'streambuf.h', + 'stream.h', + 'strfile.h', + 'strstream.h', + 'tempbuf.h', + 'tree.h', + 'type_traits.h', + 'vector.h', + # 17.6.1.2 C++ library headers + 'algorithm', + 'array', + 'atomic', + 'bitset', + 'chrono', + 'codecvt', + 'complex', + 'condition_variable', + 'deque', + 'exception', + 'forward_list', + 'fstream', + 'functional', + 'future', + 'initializer_list', + 'iomanip', + 'ios', + 'iosfwd', + 'iostream', + 'istream', + 'iterator', + 'limits', + 'list', + 'locale', + 'map', + 'memory', + 'mutex', + 'new', + 'numeric', + 'ostream', + 'queue', + 'random', + 'ratio', + 'regex', + 'set', + 'sstream', + 'stack', + 'stdexcept', + 'streambuf', + 'string', + 'strstream', + 'system_error', + 'thread', + 'tuple', + 'typeindex', + 'typeinfo', + 'type_traits', + 'unordered_map', + 'unordered_set', + 'utility', + 'valarray', + 'vector', + # 17.6.1.2 C++ headers for C library facilities + 'cassert', + 'ccomplex', + 'cctype', + 'cerrno', + 'cfenv', + 'cfloat', + 'cinttypes', + 'ciso646', + 'climits', + 'clocale', + 'cmath', + 'csetjmp', + 'csignal', + 'cstdalign', + 'cstdarg', + 'cstdbool', + 'cstddef', + 'cstdint', + 'cstdio', + 'cstdlib', + 'cstring', + 'ctgmath', + 'ctime', + 'cuchar', + 'cwchar', + 'cwctype', + ]) + + +# These headers are excluded from [build/include] and [build/include_order] +# checks: +# - Anything not following google file name conventions (containing an +# uppercase character, such as Python.h or nsStringAPI.h, for example). +# - Lua headers. +_THIRD_PARTY_HEADERS_PATTERN = re.compile( + r'^(?:[^/]*[A-Z][^/]*\.h|lua\.h|lauxlib\.h|lualib\.h)$') + + +# Assertion macros. These are defined in base/logging.h and +# testing/base/gunit.h. Note that the _M versions need to come first +# for substring matching to work. +_CHECK_MACROS = [ + 'DCHECK', 'CHECK', + 'EXPECT_TRUE_M', 'EXPECT_TRUE', + 'ASSERT_TRUE_M', 'ASSERT_TRUE', + 'EXPECT_FALSE_M', 'EXPECT_FALSE', + 'ASSERT_FALSE_M', 'ASSERT_FALSE', + ] + +# Replacement macros for CHECK/DCHECK/EXPECT_TRUE/EXPECT_FALSE +_CHECK_REPLACEMENT = dict([(m, {}) for m in _CHECK_MACROS]) + +for op, replacement in [('==', 'EQ'), ('!=', 'NE'), + ('>=', 'GE'), ('>', 'GT'), + ('<=', 'LE'), ('<', 'LT')]: + _CHECK_REPLACEMENT['DCHECK'][op] = 'DCHECK_%s' % replacement + _CHECK_REPLACEMENT['CHECK'][op] = 'CHECK_%s' % replacement + _CHECK_REPLACEMENT['EXPECT_TRUE'][op] = 'EXPECT_%s' % replacement + _CHECK_REPLACEMENT['ASSERT_TRUE'][op] = 'ASSERT_%s' % replacement + _CHECK_REPLACEMENT['EXPECT_TRUE_M'][op] = 'EXPECT_%s_M' % replacement + _CHECK_REPLACEMENT['ASSERT_TRUE_M'][op] = 'ASSERT_%s_M' % replacement + +for op, inv_replacement in [('==', 'NE'), ('!=', 'EQ'), + ('>=', 'LT'), ('>', 'LE'), + ('<=', 'GT'), ('<', 'GE')]: + _CHECK_REPLACEMENT['EXPECT_FALSE'][op] = 'EXPECT_%s' % inv_replacement + _CHECK_REPLACEMENT['ASSERT_FALSE'][op] = 'ASSERT_%s' % inv_replacement + _CHECK_REPLACEMENT['EXPECT_FALSE_M'][op] = 'EXPECT_%s_M' % inv_replacement + _CHECK_REPLACEMENT['ASSERT_FALSE_M'][op] = 'ASSERT_%s_M' % inv_replacement + +# Alternative tokens and their replacements. For full list, see section 2.5 +# Alternative tokens [lex.digraph] in the C++ standard. +# +# Digraphs (such as '%:') are not included here since it's a mess to +# match those on a word boundary. +_ALT_TOKEN_REPLACEMENT = { + 'and': '&&', + 'bitor': '|', + 'or': '||', + 'xor': '^', + 'compl': '~', + 'bitand': '&', + 'and_eq': '&=', + 'or_eq': '|=', + 'xor_eq': '^=', + 'not': '!', + 'not_eq': '!=' + } + +# Compile regular expression that matches all the above keywords. The "[ =()]" +# bit is meant to avoid matching these keywords outside of boolean expressions. +# +# False positives include C-style multi-line comments and multi-line strings +# but those have always been troublesome for cpplint. +_ALT_TOKEN_REPLACEMENT_PATTERN = re.compile( + r'[ =()](' + ('|'.join(_ALT_TOKEN_REPLACEMENT.keys())) + r')(?=[ (]|$)') + + +# These constants define types of headers for use with +# _IncludeState.CheckNextIncludeOrder(). +_C_SYS_HEADER = 1 +_CPP_SYS_HEADER = 2 +_LIKELY_MY_HEADER = 3 +_POSSIBLE_MY_HEADER = 4 +_OTHER_HEADER = 5 + +# These constants define the current inline assembly state +_NO_ASM = 0 # Outside of inline assembly block +_INSIDE_ASM = 1 # Inside inline assembly block +_END_ASM = 2 # Last line of inline assembly block +_BLOCK_ASM = 3 # The whole block is an inline assembly block + +# Match start of assembly blocks +_MATCH_ASM = re.compile(r'^\s*(?:asm|_asm|__asm|__asm__)' + r'(?:\s+(volatile|__volatile__))?' + r'\s*[{(]') + + +_regexp_compile_cache = {} + +# {str, set(int)}: a map from error categories to sets of linenumbers +# on which those errors are expected and should be suppressed. +_error_suppressions = {} + +# The root directory used for deriving header guard CPP variable. +# This is set by --root flag. +_root = None + +# The allowed line length of files. +# This is set by --linelength flag. +_line_length = 80 + +# The allowed extensions for file names +# This is set by --extensions flag. +_valid_extensions = set(['cc', 'h', 'cpp', 'cu', 'cuh']) + +def ParseNolintSuppressions(filename, raw_line, linenum, error): + """Updates the global list of error-suppressions. + + Parses any NOLINT comments on the current line, updating the global + error_suppressions store. Reports an error if the NOLINT comment + was malformed. + + Args: + filename: str, the name of the input file. + raw_line: str, the line of input text, with comments. + linenum: int, the number of the current line. + error: function, an error handler. + """ + matched = Search(r'\bNOLINT(NEXTLINE)?\b(\([^)]+\))?', raw_line) + if matched: + if matched.group(1): + suppressed_line = linenum + 1 + else: + suppressed_line = linenum + category = matched.group(2) + if category in (None, '(*)'): # => "suppress all" + _error_suppressions.setdefault(None, set()).add(suppressed_line) + else: + if category.startswith('(') and category.endswith(')'): + category = category[1:-1] + if category in _ERROR_CATEGORIES: + _error_suppressions.setdefault(category, set()).add(suppressed_line) + elif category not in _LEGACY_ERROR_CATEGORIES: + error(filename, linenum, 'readability/nolint', 5, + 'Unknown NOLINT error category: %s' % category) + + +def ResetNolintSuppressions(): + """Resets the set of NOLINT suppressions to empty.""" + _error_suppressions.clear() + + +def IsErrorSuppressedByNolint(category, linenum): + """Returns true if the specified error category is suppressed on this line. + + Consults the global error_suppressions map populated by + ParseNolintSuppressions/ResetNolintSuppressions. + + Args: + category: str, the category of the error. + linenum: int, the current line number. + Returns: + bool, True iff the error should be suppressed due to a NOLINT comment. + """ + return (linenum in _error_suppressions.get(category, set()) or + linenum in _error_suppressions.get(None, set())) + + +def Match(pattern, s): + """Matches the string with the pattern, caching the compiled regexp.""" + # The regexp compilation caching is inlined in both Match and Search for + # performance reasons; factoring it out into a separate function turns out + # to be noticeably expensive. + if pattern not in _regexp_compile_cache: + _regexp_compile_cache[pattern] = sre_compile.compile(pattern) + return _regexp_compile_cache[pattern].match(s) + + +def ReplaceAll(pattern, rep, s): + """Replaces instances of pattern in a string with a replacement. + + The compiled regex is kept in a cache shared by Match and Search. + + Args: + pattern: regex pattern + rep: replacement text + s: search string + + Returns: + string with replacements made (or original string if no replacements) + """ + if pattern not in _regexp_compile_cache: + _regexp_compile_cache[pattern] = sre_compile.compile(pattern) + return _regexp_compile_cache[pattern].sub(rep, s) + + +def Search(pattern, s): + """Searches the string for the pattern, caching the compiled regexp.""" + if pattern not in _regexp_compile_cache: + _regexp_compile_cache[pattern] = sre_compile.compile(pattern) + return _regexp_compile_cache[pattern].search(s) + + +class _IncludeState(object): + """Tracks line numbers for includes, and the order in which includes appear. + + include_list contains list of lists of (header, line number) pairs. + It's a lists of lists rather than just one flat list to make it + easier to update across preprocessor boundaries. + + Call CheckNextIncludeOrder() once for each header in the file, passing + in the type constants defined above. Calls in an illegal order will + raise an _IncludeError with an appropriate error message. + + """ + # self._section will move monotonically through this set. If it ever + # needs to move backwards, CheckNextIncludeOrder will raise an error. + _INITIAL_SECTION = 0 + _MY_H_SECTION = 1 + _C_SECTION = 2 + _CPP_SECTION = 3 + _OTHER_H_SECTION = 4 + + _TYPE_NAMES = { + _C_SYS_HEADER: 'C system header', + _CPP_SYS_HEADER: 'C++ system header', + _LIKELY_MY_HEADER: 'header this file implements', + _POSSIBLE_MY_HEADER: 'header this file may implement', + _OTHER_HEADER: 'other header', + } + _SECTION_NAMES = { + _INITIAL_SECTION: "... nothing. (This can't be an error.)", + _MY_H_SECTION: 'a header this file implements', + _C_SECTION: 'C system header', + _CPP_SECTION: 'C++ system header', + _OTHER_H_SECTION: 'other header', + } + + def __init__(self): + self.include_list = [[]] + self.ResetSection('') + + def FindHeader(self, header): + """Check if a header has already been included. + + Args: + header: header to check. + Returns: + Line number of previous occurrence, or -1 if the header has not + been seen before. + """ + for section_list in self.include_list: + for f in section_list: + if f[0] == header: + return f[1] + return -1 + + def ResetSection(self, directive): + """Reset section checking for preprocessor directive. + + Args: + directive: preprocessor directive (e.g. "if", "else"). + """ + # The name of the current section. + self._section = self._INITIAL_SECTION + # The path of last found header. + self._last_header = '' + + # Update list of includes. Note that we never pop from the + # include list. + if directive in ('if', 'ifdef', 'ifndef'): + self.include_list.append([]) + elif directive in ('else', 'elif'): + self.include_list[-1] = [] + + def SetLastHeader(self, header_path): + self._last_header = header_path + + def CanonicalizeAlphabeticalOrder(self, header_path): + """Returns a path canonicalized for alphabetical comparison. + + - replaces "-" with "_" so they both cmp the same. + - removes '-inl' since we don't require them to be after the main header. + - lowercase everything, just in case. + + Args: + header_path: Path to be canonicalized. + + Returns: + Canonicalized path. + """ + return header_path.replace('-inl.h', '.h').replace('-', '_').lower() + + def IsInAlphabeticalOrder(self, clean_lines, linenum, header_path): + """Check if a header is in alphabetical order with the previous header. + + Args: + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + header_path: Canonicalized header to be checked. + + Returns: + Returns true if the header is in alphabetical order. + """ + # If previous section is different from current section, _last_header will + # be reset to empty string, so it's always less than current header. + # + # If previous line was a blank line, assume that the headers are + # intentionally sorted the way they are. + if (self._last_header > header_path and + Match(r'^\s*#\s*include\b', clean_lines.elided[linenum - 1])): + return False + return True + + def CheckNextIncludeOrder(self, header_type): + """Returns a non-empty error message if the next header is out of order. + + This function also updates the internal state to be ready to check + the next include. + + Args: + header_type: One of the _XXX_HEADER constants defined above. + + Returns: + The empty string if the header is in the right order, or an + error message describing what's wrong. + + """ + error_message = ('Found %s after %s' % + (self._TYPE_NAMES[header_type], + self._SECTION_NAMES[self._section])) + + last_section = self._section + + if header_type == _C_SYS_HEADER: + if self._section <= self._C_SECTION: + self._section = self._C_SECTION + else: + self._last_header = '' + return error_message + elif header_type == _CPP_SYS_HEADER: + if self._section <= self._CPP_SECTION: + self._section = self._CPP_SECTION + else: + self._last_header = '' + return error_message + elif header_type == _LIKELY_MY_HEADER: + if self._section <= self._MY_H_SECTION: + self._section = self._MY_H_SECTION + else: + self._section = self._OTHER_H_SECTION + elif header_type == _POSSIBLE_MY_HEADER: + if self._section <= self._MY_H_SECTION: + self._section = self._MY_H_SECTION + else: + # This will always be the fallback because we're not sure + # enough that the header is associated with this file. + self._section = self._OTHER_H_SECTION + else: + assert header_type == _OTHER_HEADER + self._section = self._OTHER_H_SECTION + + if last_section != self._section: + self._last_header = '' + + return '' + + +class _CppLintState(object): + """Maintains module-wide state..""" + + def __init__(self): + self.verbose_level = 1 # global setting. + self.error_count = 0 # global count of reported errors + # filters to apply when emitting error messages + self.filters = _DEFAULT_FILTERS[:] + # backup of filter list. Used to restore the state after each file. + self._filters_backup = self.filters[:] + self.counting = 'total' # In what way are we counting errors? + self.errors_by_category = {} # string to int dict storing error counts + + # output format: + # "emacs" - format that emacs can parse (default) + # "vs7" - format that Microsoft Visual Studio 7 can parse + self.output_format = 'emacs' + + def SetOutputFormat(self, output_format): + """Sets the output format for errors.""" + self.output_format = output_format + + def SetVerboseLevel(self, level): + """Sets the module's verbosity, and returns the previous setting.""" + last_verbose_level = self.verbose_level + self.verbose_level = level + return last_verbose_level + + def SetCountingStyle(self, counting_style): + """Sets the module's counting options.""" + self.counting = counting_style + + def SetFilters(self, filters): + """Sets the error-message filters. + + These filters are applied when deciding whether to emit a given + error message. + + Args: + filters: A string of comma-separated filters (eg "+whitespace/indent"). + Each filter should start with + or -; else we die. + + Raises: + ValueError: The comma-separated filters did not all start with '+' or '-'. + E.g. "-,+whitespace,-whitespace/indent,whitespace/badfilter" + """ + # Default filters always have less priority than the flag ones. + self.filters = _DEFAULT_FILTERS[:] + self.AddFilters(filters) + + def AddFilters(self, filters): + """ Adds more filters to the existing list of error-message filters. """ + for filt in filters.split(','): + clean_filt = filt.strip() + if clean_filt: + self.filters.append(clean_filt) + for filt in self.filters: + if not (filt.startswith('+') or filt.startswith('-')): + raise ValueError('Every filter in --filters must start with + or -' + ' (%s does not)' % filt) + + def BackupFilters(self): + """ Saves the current filter list to backup storage.""" + self._filters_backup = self.filters[:] + + def RestoreFilters(self): + """ Restores filters previously backed up.""" + self.filters = self._filters_backup[:] + + def ResetErrorCounts(self): + """Sets the module's error statistic back to zero.""" + self.error_count = 0 + self.errors_by_category = {} + + def IncrementErrorCount(self, category): + """Bumps the module's error statistic.""" + self.error_count += 1 + if self.counting in ('toplevel', 'detailed'): + if self.counting != 'detailed': + category = category.split('/')[0] + if category not in self.errors_by_category: + self.errors_by_category[category] = 0 + self.errors_by_category[category] += 1 + + def PrintErrorCounts(self): + """Print a summary of errors by category, and the total.""" + for category, count in self.errors_by_category.iteritems(): + sys.stderr.write('Category \'%s\' errors found: %d\n' % + (category, count)) + sys.stderr.write('Total errors found: %d\n' % self.error_count) + +_cpplint_state = _CppLintState() + + +def _OutputFormat(): + """Gets the module's output format.""" + return _cpplint_state.output_format + + +def _SetOutputFormat(output_format): + """Sets the module's output format.""" + _cpplint_state.SetOutputFormat(output_format) + + +def _VerboseLevel(): + """Returns the module's verbosity setting.""" + return _cpplint_state.verbose_level + + +def _SetVerboseLevel(level): + """Sets the module's verbosity, and returns the previous setting.""" + return _cpplint_state.SetVerboseLevel(level) + + +def _SetCountingStyle(level): + """Sets the module's counting options.""" + _cpplint_state.SetCountingStyle(level) + + +def _Filters(): + """Returns the module's list of output filters, as a list.""" + return _cpplint_state.filters + + +def _SetFilters(filters): + """Sets the module's error-message filters. + + These filters are applied when deciding whether to emit a given + error message. + + Args: + filters: A string of comma-separated filters (eg "whitespace/indent"). + Each filter should start with + or -; else we die. + """ + _cpplint_state.SetFilters(filters) + +def _AddFilters(filters): + """Adds more filter overrides. + + Unlike _SetFilters, this function does not reset the current list of filters + available. + + Args: + filters: A string of comma-separated filters (eg "whitespace/indent"). + Each filter should start with + or -; else we die. + """ + _cpplint_state.AddFilters(filters) + +def _BackupFilters(): + """ Saves the current filter list to backup storage.""" + _cpplint_state.BackupFilters() + +def _RestoreFilters(): + """ Restores filters previously backed up.""" + _cpplint_state.RestoreFilters() + +class _FunctionState(object): + """Tracks current function name and the number of lines in its body.""" + + _NORMAL_TRIGGER = 250 # for --v=0, 500 for --v=1, etc. + _TEST_TRIGGER = 400 # about 50% more than _NORMAL_TRIGGER. + + def __init__(self): + self.in_a_function = False + self.lines_in_function = 0 + self.current_function = '' + + def Begin(self, function_name): + """Start analyzing function body. + + Args: + function_name: The name of the function being tracked. + """ + self.in_a_function = True + self.lines_in_function = 0 + self.current_function = function_name + + def Count(self): + """Count line in current function body.""" + if self.in_a_function: + self.lines_in_function += 1 + + def Check(self, error, filename, linenum): + """Report if too many lines in function body. + + Args: + error: The function to call with any errors found. + filename: The name of the current file. + linenum: The number of the line to check. + """ + if Match(r'T(EST|est)', self.current_function): + base_trigger = self._TEST_TRIGGER + else: + base_trigger = self._NORMAL_TRIGGER + trigger = base_trigger * 2**_VerboseLevel() + + if self.lines_in_function > trigger: + error_level = int(math.log(self.lines_in_function / base_trigger, 2)) + # 50 => 0, 100 => 1, 200 => 2, 400 => 3, 800 => 4, 1600 => 5, ... + if error_level > 5: + error_level = 5 + error(filename, linenum, 'readability/fn_size', error_level, + 'Small and focused functions are preferred:' + ' %s has %d non-comment lines' + ' (error triggered by exceeding %d lines).' % ( + self.current_function, self.lines_in_function, trigger)) + + def End(self): + """Stop analyzing function body.""" + self.in_a_function = False + + +class _IncludeError(Exception): + """Indicates a problem with the include order in a file.""" + pass + + +class FileInfo(object): + """Provides utility functions for filenames. + + FileInfo provides easy access to the components of a file's path + relative to the project root. + """ + + def __init__(self, filename): + self._filename = filename + + def FullName(self): + """Make Windows paths like Unix.""" + return os.path.abspath(self._filename).replace('\\', '/') + + def RepositoryName(self): + """FullName after removing the local path to the repository. + + If we have a real absolute path name here we can try to do something smart: + detecting the root of the checkout and truncating /path/to/checkout from + the name so that we get header guards that don't include things like + "C:\Documents and Settings\..." or "/home/username/..." in them and thus + people on different computers who have checked the source out to different + locations won't see bogus errors. + """ + fullname = self.FullName() + + if os.path.exists(fullname): + project_dir = os.path.dirname(fullname) + + if os.path.exists(os.path.join(project_dir, ".svn")): + # If there's a .svn file in the current directory, we recursively look + # up the directory tree for the top of the SVN checkout + root_dir = project_dir + one_up_dir = os.path.dirname(root_dir) + while os.path.exists(os.path.join(one_up_dir, ".svn")): + root_dir = os.path.dirname(root_dir) + one_up_dir = os.path.dirname(one_up_dir) + + prefix = os.path.commonprefix([root_dir, project_dir]) + return fullname[len(prefix) + 1:] + + # Not SVN <= 1.6? Try to find a git, hg, or svn top level directory by + # searching up from the current path. + root_dir = os.path.dirname(fullname) + while (root_dir != os.path.dirname(root_dir) and + not os.path.exists(os.path.join(root_dir, ".git")) and + not os.path.exists(os.path.join(root_dir, ".hg")) and + not os.path.exists(os.path.join(root_dir, ".svn"))): + root_dir = os.path.dirname(root_dir) + + if (os.path.exists(os.path.join(root_dir, ".git")) or + os.path.exists(os.path.join(root_dir, ".hg")) or + os.path.exists(os.path.join(root_dir, ".svn"))): + prefix = os.path.commonprefix([root_dir, project_dir]) + return fullname[len(prefix) + 1:] + + # Don't know what to do; header guard warnings may be wrong... + return fullname + + def Split(self): + """Splits the file into the directory, basename, and extension. + + For 'chrome/browser/browser.cc', Split() would + return ('chrome/browser', 'browser', '.cc') + + Returns: + A tuple of (directory, basename, extension). + """ + + googlename = self.RepositoryName() + project, rest = os.path.split(googlename) + return (project,) + os.path.splitext(rest) + + def BaseName(self): + """File base name - text after the final slash, before the final period.""" + return self.Split()[1] + + def Extension(self): + """File extension - text following the final period.""" + return self.Split()[2] + + def NoExtension(self): + """File has no source file extension.""" + return '/'.join(self.Split()[0:2]) + + def IsSource(self): + """File has a source file extension.""" + return self.Extension()[1:] in ('c', 'cc', 'cpp', 'cxx') + + +def _ShouldPrintError(category, confidence, linenum): + """If confidence >= verbose, category passes filter and is not suppressed.""" + + # There are three ways we might decide not to print an error message: + # a "NOLINT(category)" comment appears in the source, + # the verbosity level isn't high enough, or the filters filter it out. + if IsErrorSuppressedByNolint(category, linenum): + return False + + if confidence < _cpplint_state.verbose_level: + return False + + is_filtered = False + for one_filter in _Filters(): + if one_filter.startswith('-'): + if category.startswith(one_filter[1:]): + is_filtered = True + elif one_filter.startswith('+'): + if category.startswith(one_filter[1:]): + is_filtered = False + else: + assert False # should have been checked for in SetFilter. + if is_filtered: + return False + + return True + + +def Error(filename, linenum, category, confidence, message): + """Logs the fact we've found a lint error. + + We log where the error was found, and also our confidence in the error, + that is, how certain we are this is a legitimate style regression, and + not a misidentification or a use that's sometimes justified. + + False positives can be suppressed by the use of + "cpplint(category)" comments on the offending line. These are + parsed into _error_suppressions. + + Args: + filename: The name of the file containing the error. + linenum: The number of the line containing the error. + category: A string used to describe the "category" this bug + falls under: "whitespace", say, or "runtime". Categories + may have a hierarchy separated by slashes: "whitespace/indent". + confidence: A number from 1-5 representing a confidence score for + the error, with 5 meaning that we are certain of the problem, + and 1 meaning that it could be a legitimate construct. + message: The error message. + """ + if _ShouldPrintError(category, confidence, linenum): + _cpplint_state.IncrementErrorCount(category) + if _cpplint_state.output_format == 'vs7': + sys.stderr.write('%s(%s): %s [%s] [%d]\n' % ( + filename, linenum, message, category, confidence)) + elif _cpplint_state.output_format == 'eclipse': + sys.stderr.write('%s:%s: warning: %s [%s] [%d]\n' % ( + filename, linenum, message, category, confidence)) + else: + sys.stderr.write('%s:%s: %s [%s] [%d]\n' % ( + filename, linenum, message, category, confidence)) + + +# Matches standard C++ escape sequences per 2.13.2.3 of the C++ standard. +_RE_PATTERN_CLEANSE_LINE_ESCAPES = re.compile( + r'\\([abfnrtv?"\\\']|\d+|x[0-9a-fA-F]+)') +# Match a single C style comment on the same line. +_RE_PATTERN_C_COMMENTS = r'/\*(?:[^*]|\*(?!/))*\*/' +# Matches multi-line C style comments. +# This RE is a little bit more complicated than one might expect, because we +# have to take care of space removals tools so we can handle comments inside +# statements better. +# The current rule is: We only clear spaces from both sides when we're at the +# end of the line. Otherwise, we try to remove spaces from the right side, +# if this doesn't work we try on left side but only if there's a non-character +# on the right. +_RE_PATTERN_CLEANSE_LINE_C_COMMENTS = re.compile( + r'(\s*' + _RE_PATTERN_C_COMMENTS + r'\s*$|' + + _RE_PATTERN_C_COMMENTS + r'\s+|' + + r'\s+' + _RE_PATTERN_C_COMMENTS + r'(?=\W)|' + + _RE_PATTERN_C_COMMENTS + r')') + + +def IsCppString(line): + """Does line terminate so, that the next symbol is in string constant. + + This function does not consider single-line nor multi-line comments. + + Args: + line: is a partial line of code starting from the 0..n. + + Returns: + True, if next character appended to 'line' is inside a + string constant. + """ + + line = line.replace(r'\\', 'XX') # after this, \\" does not match to \" + return ((line.count('"') - line.count(r'\"') - line.count("'\"'")) & 1) == 1 + + +def CleanseRawStrings(raw_lines): + """Removes C++11 raw strings from lines. + + Before: + static const char kData[] = R"( + multi-line string + )"; + + After: + static const char kData[] = "" + (replaced by blank line) + ""; + + Args: + raw_lines: list of raw lines. + + Returns: + list of lines with C++11 raw strings replaced by empty strings. + """ + + delimiter = None + lines_without_raw_strings = [] + for line in raw_lines: + if delimiter: + # Inside a raw string, look for the end + end = line.find(delimiter) + if end >= 0: + # Found the end of the string, match leading space for this + # line and resume copying the original lines, and also insert + # a "" on the last line. + leading_space = Match(r'^(\s*)\S', line) + line = leading_space.group(1) + '""' + line[end + len(delimiter):] + delimiter = None + else: + # Haven't found the end yet, append a blank line. + line = '""' + + # Look for beginning of a raw string, and replace them with + # empty strings. This is done in a loop to handle multiple raw + # strings on the same line. + while delimiter is None: + # Look for beginning of a raw string. + # See 2.14.15 [lex.string] for syntax. + matched = Match(r'^(.*)\b(?:R|u8R|uR|UR|LR)"([^\s\\()]*)\((.*)$', line) + if matched: + delimiter = ')' + matched.group(2) + '"' + + end = matched.group(3).find(delimiter) + if end >= 0: + # Raw string ended on same line + line = (matched.group(1) + '""' + + matched.group(3)[end + len(delimiter):]) + delimiter = None + else: + # Start of a multi-line raw string + line = matched.group(1) + '""' + else: + break + + lines_without_raw_strings.append(line) + + # TODO(unknown): if delimiter is not None here, we might want to + # emit a warning for unterminated string. + return lines_without_raw_strings + + +def FindNextMultiLineCommentStart(lines, lineix): + """Find the beginning marker for a multiline comment.""" + while lineix < len(lines): + if lines[lineix].strip().startswith('/*'): + # Only return this marker if the comment goes beyond this line + if lines[lineix].strip().find('*/', 2) < 0: + return lineix + lineix += 1 + return len(lines) + + +def FindNextMultiLineCommentEnd(lines, lineix): + """We are inside a comment, find the end marker.""" + while lineix < len(lines): + if lines[lineix].strip().endswith('*/'): + return lineix + lineix += 1 + return len(lines) + + +def RemoveMultiLineCommentsFromRange(lines, begin, end): + """Clears a range of lines for multi-line comments.""" + # Having // dummy comments makes the lines non-empty, so we will not get + # unnecessary blank line warnings later in the code. + for i in range(begin, end): + lines[i] = '/**/' + + +def RemoveMultiLineComments(filename, lines, error): + """Removes multiline (c-style) comments from lines.""" + lineix = 0 + while lineix < len(lines): + lineix_begin = FindNextMultiLineCommentStart(lines, lineix) + if lineix_begin >= len(lines): + return + lineix_end = FindNextMultiLineCommentEnd(lines, lineix_begin) + if lineix_end >= len(lines): + error(filename, lineix_begin + 1, 'readability/multiline_comment', 5, + 'Could not find end of multi-line comment') + return + RemoveMultiLineCommentsFromRange(lines, lineix_begin, lineix_end + 1) + lineix = lineix_end + 1 + + +def CleanseComments(line): + """Removes //-comments and single-line C-style /* */ comments. + + Args: + line: A line of C++ source. + + Returns: + The line with single-line comments removed. + """ + commentpos = line.find('//') + if commentpos != -1 and not IsCppString(line[:commentpos]): + line = line[:commentpos].rstrip() + # get rid of /* ... */ + return _RE_PATTERN_CLEANSE_LINE_C_COMMENTS.sub('', line) + + +class CleansedLines(object): + """Holds 4 copies of all lines with different preprocessing applied to them. + + 1) elided member contains lines without strings and comments. + 2) lines member contains lines without comments. + 3) raw_lines member contains all the lines without processing. + 4) lines_without_raw_strings member is same as raw_lines, but with C++11 raw + strings removed. + All these members are of , and of the same length. + """ + + def __init__(self, lines): + self.elided = [] + self.lines = [] + self.raw_lines = lines + self.num_lines = len(lines) + self.lines_without_raw_strings = CleanseRawStrings(lines) + for linenum in range(len(self.lines_without_raw_strings)): + self.lines.append(CleanseComments( + self.lines_without_raw_strings[linenum])) + elided = self._CollapseStrings(self.lines_without_raw_strings[linenum]) + self.elided.append(CleanseComments(elided)) + + def NumLines(self): + """Returns the number of lines represented.""" + return self.num_lines + + @staticmethod + def _CollapseStrings(elided): + """Collapses strings and chars on a line to simple "" or '' blocks. + + We nix strings first so we're not fooled by text like '"http://"' + + Args: + elided: The line being processed. + + Returns: + The line with collapsed strings. + """ + if _RE_PATTERN_INCLUDE.match(elided): + return elided + + # Remove escaped characters first to make quote/single quote collapsing + # basic. Things that look like escaped characters shouldn't occur + # outside of strings and chars. + elided = _RE_PATTERN_CLEANSE_LINE_ESCAPES.sub('', elided) + + # Replace quoted strings and digit separators. Both single quotes + # and double quotes are processed in the same loop, otherwise + # nested quotes wouldn't work. + collapsed = '' + while True: + # Find the first quote character + match = Match(r'^([^\'"]*)([\'"])(.*)$', elided) + if not match: + collapsed += elided + break + head, quote, tail = match.groups() + + if quote == '"': + # Collapse double quoted strings + second_quote = tail.find('"') + if second_quote >= 0: + collapsed += head + '""' + elided = tail[second_quote + 1:] + else: + # Unmatched double quote, don't bother processing the rest + # of the line since this is probably a multiline string. + collapsed += elided + break + else: + # Found single quote, check nearby text to eliminate digit separators. + # + # There is no special handling for floating point here, because + # the integer/fractional/exponent parts would all be parsed + # correctly as long as there are digits on both sides of the + # separator. So we are fine as long as we don't see something + # like "0.'3" (gcc 4.9.0 will not allow this literal). + if Search(r'\b(?:0[bBxX]?|[1-9])[0-9a-fA-F]*$', head): + match_literal = Match(r'^((?:\'?[0-9a-zA-Z_])*)(.*)$', "'" + tail) + collapsed += head + match_literal.group(1).replace("'", '') + elided = match_literal.group(2) + else: + second_quote = tail.find('\'') + if second_quote >= 0: + collapsed += head + "''" + elided = tail[second_quote + 1:] + else: + # Unmatched single quote + collapsed += elided + break + + return collapsed + + +def FindEndOfExpressionInLine(line, startpos, stack): + """Find the position just after the end of current parenthesized expression. + + Args: + line: a CleansedLines line. + startpos: start searching at this position. + stack: nesting stack at startpos. + + Returns: + On finding matching end: (index just after matching end, None) + On finding an unclosed expression: (-1, None) + Otherwise: (-1, new stack at end of this line) + """ + for i in xrange(startpos, len(line)): + char = line[i] + if char in '([{': + # Found start of parenthesized expression, push to expression stack + stack.append(char) + elif char == '<': + # Found potential start of template argument list + if i > 0 and line[i - 1] == '<': + # Left shift operator + if stack and stack[-1] == '<': + stack.pop() + if not stack: + return (-1, None) + elif i > 0 and Search(r'\boperator\s*$', line[0:i]): + # operator<, don't add to stack + continue + else: + # Tentative start of template argument list + stack.append('<') + elif char in ')]}': + # Found end of parenthesized expression. + # + # If we are currently expecting a matching '>', the pending '<' + # must have been an operator. Remove them from expression stack. + while stack and stack[-1] == '<': + stack.pop() + if not stack: + return (-1, None) + if ((stack[-1] == '(' and char == ')') or + (stack[-1] == '[' and char == ']') or + (stack[-1] == '{' and char == '}')): + stack.pop() + if not stack: + return (i + 1, None) + else: + # Mismatched parentheses + return (-1, None) + elif char == '>': + # Found potential end of template argument list. + + # Ignore "->" and operator functions + if (i > 0 and + (line[i - 1] == '-' or Search(r'\boperator\s*$', line[0:i - 1]))): + continue + + # Pop the stack if there is a matching '<'. Otherwise, ignore + # this '>' since it must be an operator. + if stack: + if stack[-1] == '<': + stack.pop() + if not stack: + return (i + 1, None) + elif char == ';': + # Found something that look like end of statements. If we are currently + # expecting a '>', the matching '<' must have been an operator, since + # template argument list should not contain statements. + while stack and stack[-1] == '<': + stack.pop() + if not stack: + return (-1, None) + + # Did not find end of expression or unbalanced parentheses on this line + return (-1, stack) + + +def CloseExpression(clean_lines, linenum, pos): + """If input points to ( or { or [ or <, finds the position that closes it. + + If lines[linenum][pos] points to a '(' or '{' or '[' or '<', finds the + linenum/pos that correspond to the closing of the expression. + + TODO(unknown): cpplint spends a fair bit of time matching parentheses. + Ideally we would want to index all opening and closing parentheses once + and have CloseExpression be just a simple lookup, but due to preprocessor + tricks, this is not so easy. + + Args: + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + pos: A position on the line. + + Returns: + A tuple (line, linenum, pos) pointer *past* the closing brace, or + (line, len(lines), -1) if we never find a close. Note we ignore + strings and comments when matching; and the line we return is the + 'cleansed' line at linenum. + """ + + line = clean_lines.elided[linenum] + if (line[pos] not in '({[<') or Match(r'<[<=]', line[pos:]): + return (line, clean_lines.NumLines(), -1) + + # Check first line + (end_pos, stack) = FindEndOfExpressionInLine(line, pos, []) + if end_pos > -1: + return (line, linenum, end_pos) + + # Continue scanning forward + while stack and linenum < clean_lines.NumLines() - 1: + linenum += 1 + line = clean_lines.elided[linenum] + (end_pos, stack) = FindEndOfExpressionInLine(line, 0, stack) + if end_pos > -1: + return (line, linenum, end_pos) + + # Did not find end of expression before end of file, give up + return (line, clean_lines.NumLines(), -1) + + +def FindStartOfExpressionInLine(line, endpos, stack): + """Find position at the matching start of current expression. + + This is almost the reverse of FindEndOfExpressionInLine, but note + that the input position and returned position differs by 1. + + Args: + line: a CleansedLines line. + endpos: start searching at this position. + stack: nesting stack at endpos. + + Returns: + On finding matching start: (index at matching start, None) + On finding an unclosed expression: (-1, None) + Otherwise: (-1, new stack at beginning of this line) + """ + i = endpos + while i >= 0: + char = line[i] + if char in ')]}': + # Found end of expression, push to expression stack + stack.append(char) + elif char == '>': + # Found potential end of template argument list. + # + # Ignore it if it's a "->" or ">=" or "operator>" + if (i > 0 and + (line[i - 1] == '-' or + Match(r'\s>=\s', line[i - 1:]) or + Search(r'\boperator\s*$', line[0:i]))): + i -= 1 + else: + stack.append('>') + elif char == '<': + # Found potential start of template argument list + if i > 0 and line[i - 1] == '<': + # Left shift operator + i -= 1 + else: + # If there is a matching '>', we can pop the expression stack. + # Otherwise, ignore this '<' since it must be an operator. + if stack and stack[-1] == '>': + stack.pop() + if not stack: + return (i, None) + elif char in '([{': + # Found start of expression. + # + # If there are any unmatched '>' on the stack, they must be + # operators. Remove those. + while stack and stack[-1] == '>': + stack.pop() + if not stack: + return (-1, None) + if ((char == '(' and stack[-1] == ')') or + (char == '[' and stack[-1] == ']') or + (char == '{' and stack[-1] == '}')): + stack.pop() + if not stack: + return (i, None) + else: + # Mismatched parentheses + return (-1, None) + elif char == ';': + # Found something that look like end of statements. If we are currently + # expecting a '<', the matching '>' must have been an operator, since + # template argument list should not contain statements. + while stack and stack[-1] == '>': + stack.pop() + if not stack: + return (-1, None) + + i -= 1 + + return (-1, stack) + + +def ReverseCloseExpression(clean_lines, linenum, pos): + """If input points to ) or } or ] or >, finds the position that opens it. + + If lines[linenum][pos] points to a ')' or '}' or ']' or '>', finds the + linenum/pos that correspond to the opening of the expression. + + Args: + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + pos: A position on the line. + + Returns: + A tuple (line, linenum, pos) pointer *at* the opening brace, or + (line, 0, -1) if we never find the matching opening brace. Note + we ignore strings and comments when matching; and the line we + return is the 'cleansed' line at linenum. + """ + line = clean_lines.elided[linenum] + if line[pos] not in ')}]>': + return (line, 0, -1) + + # Check last line + (start_pos, stack) = FindStartOfExpressionInLine(line, pos, []) + if start_pos > -1: + return (line, linenum, start_pos) + + # Continue scanning backward + while stack and linenum > 0: + linenum -= 1 + line = clean_lines.elided[linenum] + (start_pos, stack) = FindStartOfExpressionInLine(line, len(line) - 1, stack) + if start_pos > -1: + return (line, linenum, start_pos) + + # Did not find start of expression before beginning of file, give up + return (line, 0, -1) + + +def CheckForCopyright(filename, lines, error): + """Logs an error if no Copyright message appears at the top of the file.""" + + # We'll say it should occur by line 10. Don't forget there's a + # dummy line at the front. + for line in xrange(1, min(len(lines), 11)): + if re.search(r'Copyright', lines[line], re.I): break + else: # means no copyright line was found + error(filename, 0, 'legal/copyright', 5, + 'No copyright message found. ' + 'You should have a line: "Copyright [year] "') + + +def GetIndentLevel(line): + """Return the number of leading spaces in line. + + Args: + line: A string to check. + + Returns: + An integer count of leading spaces, possibly zero. + """ + indent = Match(r'^( *)\S', line) + if indent: + return len(indent.group(1)) + else: + return 0 + + +def GetHeaderGuardCPPVariable(filename): + """Returns the CPP variable that should be used as a header guard. + + Args: + filename: The name of a C++ header file. + + Returns: + The CPP variable that should be used as a header guard in the + named file. + + """ + + # Restores original filename in case that cpplint is invoked from Emacs's + # flymake. + filename = re.sub(r'_flymake\.h$', '.h', filename) + filename = re.sub(r'/\.flymake/([^/]*)$', r'/\1', filename) + # Replace 'c++' with 'cpp'. + filename = filename.replace('C++', 'cpp').replace('c++', 'cpp') + + fileinfo = FileInfo(filename) + file_path_from_root = fileinfo.RepositoryName() + if _root: + file_path_from_root = re.sub('^' + _root + os.sep, '', file_path_from_root) + return re.sub(r'[^a-zA-Z0-9]', '_', file_path_from_root).upper() + '_' + + +def CheckForHeaderGuard(filename, clean_lines, error): + """Checks that the file contains a header guard. + + Logs an error if no #ifndef header guard is present. For other + headers, checks that the full pathname is used. + + Args: + filename: The name of the C++ header file. + clean_lines: A CleansedLines instance containing the file. + error: The function to call with any errors found. + """ + + # Don't check for header guards if there are error suppression + # comments somewhere in this file. + # + # Because this is silencing a warning for a nonexistent line, we + # only support the very specific NOLINT(build/header_guard) syntax, + # and not the general NOLINT or NOLINT(*) syntax. + raw_lines = clean_lines.lines_without_raw_strings + for i in raw_lines: + if Search(r'//\s*NOLINT\(build/header_guard\)', i): + return + + cppvar = GetHeaderGuardCPPVariable(filename) + + ifndef = '' + ifndef_linenum = 0 + define = '' + endif = '' + endif_linenum = 0 + for linenum, line in enumerate(raw_lines): + linesplit = line.split() + if len(linesplit) >= 2: + # find the first occurrence of #ifndef and #define, save arg + if not ifndef and linesplit[0] == '#ifndef': + # set ifndef to the header guard presented on the #ifndef line. + ifndef = linesplit[1] + ifndef_linenum = linenum + if not define and linesplit[0] == '#define': + define = linesplit[1] + # find the last occurrence of #endif, save entire line + if line.startswith('#endif'): + endif = line + endif_linenum = linenum + + if not ifndef or not define or ifndef != define: + error(filename, 0, 'build/header_guard', 5, + 'No #ifndef header guard found, suggested CPP variable is: %s' % + cppvar) + return + + # The guard should be PATH_FILE_H_, but we also allow PATH_FILE_H__ + # for backward compatibility. + if ifndef != cppvar: + error_level = 0 + if ifndef != cppvar + '_': + error_level = 5 + + ParseNolintSuppressions(filename, raw_lines[ifndef_linenum], ifndef_linenum, + error) + error(filename, ifndef_linenum, 'build/header_guard', error_level, + '#ifndef header guard has wrong style, please use: %s' % cppvar) + + # Check for "//" comments on endif line. + ParseNolintSuppressions(filename, raw_lines[endif_linenum], endif_linenum, + error) + match = Match(r'#endif\s*//\s*' + cppvar + r'(_)?\b', endif) + if match: + if match.group(1) == '_': + # Issue low severity warning for deprecated double trailing underscore + error(filename, endif_linenum, 'build/header_guard', 0, + '#endif line should be "#endif // %s"' % cppvar) + return + + # Didn't find the corresponding "//" comment. If this file does not + # contain any "//" comments at all, it could be that the compiler + # only wants "/**/" comments, look for those instead. + no_single_line_comments = True + for i in xrange(1, len(raw_lines) - 1): + line = raw_lines[i] + if Match(r'^(?:(?:\'(?:\.|[^\'])*\')|(?:"(?:\.|[^"])*")|[^\'"])*//', line): + no_single_line_comments = False + break + + if no_single_line_comments: + match = Match(r'#endif\s*/\*\s*' + cppvar + r'(_)?\s*\*/', endif) + if match: + if match.group(1) == '_': + # Low severity warning for double trailing underscore + error(filename, endif_linenum, 'build/header_guard', 0, + '#endif line should be "#endif /* %s */"' % cppvar) + return + + # Didn't find anything + error(filename, endif_linenum, 'build/header_guard', 5, + '#endif line should be "#endif // %s"' % cppvar) + + +def CheckHeaderFileIncluded(filename, include_state, error): + """Logs an error if a .cc file does not include its header.""" + + # Do not check test files + if filename.endswith('_test.cc') or filename.endswith('_unittest.cc'): + return + + fileinfo = FileInfo(filename) + headerfile = filename[0:len(filename) - 2] + 'h' + if not os.path.exists(headerfile): + return + headername = FileInfo(headerfile).RepositoryName() + first_include = 0 + for section_list in include_state.include_list: + for f in section_list: + if headername in f[0] or f[0] in headername: + return + if not first_include: + first_include = f[1] + + error(filename, first_include, 'build/include', 5, + '%s should include its header file %s' % (fileinfo.RepositoryName(), + headername)) + + +def CheckForBadCharacters(filename, lines, error): + """Logs an error for each line containing bad characters. + + Two kinds of bad characters: + + 1. Unicode replacement characters: These indicate that either the file + contained invalid UTF-8 (likely) or Unicode replacement characters (which + it shouldn't). Note that it's possible for this to throw off line + numbering if the invalid UTF-8 occurred adjacent to a newline. + + 2. NUL bytes. These are problematic for some tools. + + Args: + filename: The name of the current file. + lines: An array of strings, each representing a line of the file. + error: The function to call with any errors found. + """ + for linenum, line in enumerate(lines): + if u'\ufffd' in line: + error(filename, linenum, 'readability/utf8', 5, + 'Line contains invalid UTF-8 (or Unicode replacement character).') + if '\0' in line: + error(filename, linenum, 'readability/nul', 5, 'Line contains NUL byte.') + + +def CheckForNewlineAtEOF(filename, lines, error): + """Logs an error if there is no newline char at the end of the file. + + Args: + filename: The name of the current file. + lines: An array of strings, each representing a line of the file. + error: The function to call with any errors found. + """ + + # The array lines() was created by adding two newlines to the + # original file (go figure), then splitting on \n. + # To verify that the file ends in \n, we just have to make sure the + # last-but-two element of lines() exists and is empty. + if len(lines) < 3 or lines[-2]: + error(filename, len(lines) - 2, 'whitespace/ending_newline', 5, + 'Could not find a newline character at the end of the file.') + + +def CheckForMultilineCommentsAndStrings(filename, clean_lines, linenum, error): + """Logs an error if we see /* ... */ or "..." that extend past one line. + + /* ... */ comments are legit inside macros, for one line. + Otherwise, we prefer // comments, so it's ok to warn about the + other. Likewise, it's ok for strings to extend across multiple + lines, as long as a line continuation character (backslash) + terminates each line. Although not currently prohibited by the C++ + style guide, it's ugly and unnecessary. We don't do well with either + in this lint program, so we warn about both. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + line = clean_lines.elided[linenum] + + # Remove all \\ (escaped backslashes) from the line. They are OK, and the + # second (escaped) slash may trigger later \" detection erroneously. + line = line.replace('\\\\', '') + + if line.count('/*') > line.count('*/'): + error(filename, linenum, 'readability/multiline_comment', 5, + 'Complex multi-line /*...*/-style comment found. ' + 'Lint may give bogus warnings. ' + 'Consider replacing these with //-style comments, ' + 'with #if 0...#endif, ' + 'or with more clearly structured multi-line comments.') + + if (line.count('"') - line.count('\\"')) % 2: + error(filename, linenum, 'readability/multiline_string', 5, + 'Multi-line string ("...") found. This lint script doesn\'t ' + 'do well with such strings, and may give bogus warnings. ' + 'Use C++11 raw strings or concatenation instead.') + + +# (non-threadsafe name, thread-safe alternative, validation pattern) +# +# The validation pattern is used to eliminate false positives such as: +# _rand(); // false positive due to substring match. +# ->rand(); // some member function rand(). +# ACMRandom rand(seed); // some variable named rand. +# ISAACRandom rand(); // another variable named rand. +# +# Basically we require the return value of these functions to be used +# in some expression context on the same line by matching on some +# operator before the function name. This eliminates constructors and +# member function calls. +_UNSAFE_FUNC_PREFIX = r'(?:[-+*/=%^&|(<]\s*|>\s+)' +_THREADING_LIST = ( + ('asctime(', 'asctime_r(', _UNSAFE_FUNC_PREFIX + r'asctime\([^)]+\)'), + ('ctime(', 'ctime_r(', _UNSAFE_FUNC_PREFIX + r'ctime\([^)]+\)'), + ('getgrgid(', 'getgrgid_r(', _UNSAFE_FUNC_PREFIX + r'getgrgid\([^)]+\)'), + ('getgrnam(', 'getgrnam_r(', _UNSAFE_FUNC_PREFIX + r'getgrnam\([^)]+\)'), + ('getlogin(', 'getlogin_r(', _UNSAFE_FUNC_PREFIX + r'getlogin\(\)'), + ('getpwnam(', 'getpwnam_r(', _UNSAFE_FUNC_PREFIX + r'getpwnam\([^)]+\)'), + ('getpwuid(', 'getpwuid_r(', _UNSAFE_FUNC_PREFIX + r'getpwuid\([^)]+\)'), + ('gmtime(', 'gmtime_r(', _UNSAFE_FUNC_PREFIX + r'gmtime\([^)]+\)'), + ('localtime(', 'localtime_r(', _UNSAFE_FUNC_PREFIX + r'localtime\([^)]+\)'), + ('rand(', 'rand_r(', _UNSAFE_FUNC_PREFIX + r'rand\(\)'), + ('strtok(', 'strtok_r(', + _UNSAFE_FUNC_PREFIX + r'strtok\([^)]+\)'), + ('ttyname(', 'ttyname_r(', _UNSAFE_FUNC_PREFIX + r'ttyname\([^)]+\)'), + ) + + +def CheckPosixThreading(filename, clean_lines, linenum, error): + """Checks for calls to thread-unsafe functions. + + Much code has been originally written without consideration of + multi-threading. Also, engineers are relying on their old experience; + they have learned posix before threading extensions were added. These + tests guide the engineers to use thread-safe functions (when using + posix directly). + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + line = clean_lines.elided[linenum] + for single_thread_func, multithread_safe_func, pattern in _THREADING_LIST: + # Additional pattern matching check to confirm that this is the + # function we are looking for + if Search(pattern, line): + error(filename, linenum, 'runtime/threadsafe_fn', 2, + 'Consider using ' + multithread_safe_func + + '...) instead of ' + single_thread_func + + '...) for improved thread safety.') + + +def CheckVlogArguments(filename, clean_lines, linenum, error): + """Checks that VLOG() is only used for defining a logging level. + + For example, VLOG(2) is correct. VLOG(INFO), VLOG(WARNING), VLOG(ERROR), and + VLOG(FATAL) are not. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + line = clean_lines.elided[linenum] + if Search(r'\bVLOG\((INFO|ERROR|WARNING|DFATAL|FATAL)\)', line): + error(filename, linenum, 'runtime/vlog', 5, + 'VLOG() should be used with numeric verbosity level. ' + 'Use LOG() if you want symbolic severity levels.') + +# Matches invalid increment: *count++, which moves pointer instead of +# incrementing a value. +_RE_PATTERN_INVALID_INCREMENT = re.compile( + r'^\s*\*\w+(\+\+|--);') + + +def CheckInvalidIncrement(filename, clean_lines, linenum, error): + """Checks for invalid increment *count++. + + For example following function: + void increment_counter(int* count) { + *count++; + } + is invalid, because it effectively does count++, moving pointer, and should + be replaced with ++*count, (*count)++ or *count += 1. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + line = clean_lines.elided[linenum] + if _RE_PATTERN_INVALID_INCREMENT.match(line): + error(filename, linenum, 'runtime/invalid_increment', 5, + 'Changing pointer instead of value (or unused value of operator*).') + + +def IsMacroDefinition(clean_lines, linenum): + if Search(r'^#define', clean_lines[linenum]): + return True + + if linenum > 0 and Search(r'\\$', clean_lines[linenum - 1]): + return True + + return False + + +def IsForwardClassDeclaration(clean_lines, linenum): + return Match(r'^\s*(\btemplate\b)*.*class\s+\w+;\s*$', clean_lines[linenum]) + + +class _BlockInfo(object): + """Stores information about a generic block of code.""" + + def __init__(self, seen_open_brace): + self.seen_open_brace = seen_open_brace + self.open_parentheses = 0 + self.inline_asm = _NO_ASM + self.check_namespace_indentation = False + + def CheckBegin(self, filename, clean_lines, linenum, error): + """Run checks that applies to text up to the opening brace. + + This is mostly for checking the text after the class identifier + and the "{", usually where the base class is specified. For other + blocks, there isn't much to check, so we always pass. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + pass + + def CheckEnd(self, filename, clean_lines, linenum, error): + """Run checks that applies to text after the closing brace. + + This is mostly used for checking end of namespace comments. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + pass + + def IsBlockInfo(self): + """Returns true if this block is a _BlockInfo. + + This is convenient for verifying that an object is an instance of + a _BlockInfo, but not an instance of any of the derived classes. + + Returns: + True for this class, False for derived classes. + """ + return self.__class__ == _BlockInfo + + +class _ExternCInfo(_BlockInfo): + """Stores information about an 'extern "C"' block.""" + + def __init__(self): + _BlockInfo.__init__(self, True) + + +class _ClassInfo(_BlockInfo): + """Stores information about a class.""" + + def __init__(self, name, class_or_struct, clean_lines, linenum): + _BlockInfo.__init__(self, False) + self.name = name + self.starting_linenum = linenum + self.is_derived = False + self.check_namespace_indentation = True + if class_or_struct == 'struct': + self.access = 'public' + self.is_struct = True + else: + self.access = 'private' + self.is_struct = False + + # Remember initial indentation level for this class. Using raw_lines here + # instead of elided to account for leading comments. + self.class_indent = GetIndentLevel(clean_lines.raw_lines[linenum]) + + # Try to find the end of the class. This will be confused by things like: + # class A { + # } *x = { ... + # + # But it's still good enough for CheckSectionSpacing. + self.last_line = 0 + depth = 0 + for i in range(linenum, clean_lines.NumLines()): + line = clean_lines.elided[i] + depth += line.count('{') - line.count('}') + if not depth: + self.last_line = i + break + + def CheckBegin(self, filename, clean_lines, linenum, error): + # Look for a bare ':' + if Search('(^|[^:]):($|[^:])', clean_lines.elided[linenum]): + self.is_derived = True + + def CheckEnd(self, filename, clean_lines, linenum, error): + # If there is a DISALLOW macro, it should appear near the end of + # the class. + seen_last_thing_in_class = False + for i in xrange(linenum - 1, self.starting_linenum, -1): + match = Search( + r'\b(DISALLOW_COPY_AND_ASSIGN|DISALLOW_IMPLICIT_CONSTRUCTORS)\(' + + self.name + r'\)', + clean_lines.elided[i]) + if match: + if seen_last_thing_in_class: + error(filename, i, 'readability/constructors', 3, + match.group(1) + ' should be the last thing in the class') + break + + if not Match(r'^\s*$', clean_lines.elided[i]): + seen_last_thing_in_class = True + + # Check that closing brace is aligned with beginning of the class. + # Only do this if the closing brace is indented by only whitespaces. + # This means we will not check single-line class definitions. + indent = Match(r'^( *)\}', clean_lines.elided[linenum]) + if indent and len(indent.group(1)) != self.class_indent: + if self.is_struct: + parent = 'struct ' + self.name + else: + parent = 'class ' + self.name + error(filename, linenum, 'whitespace/indent', 3, + 'Closing brace should be aligned with beginning of %s' % parent) + + +class _NamespaceInfo(_BlockInfo): + """Stores information about a namespace.""" + + def __init__(self, name, linenum): + _BlockInfo.__init__(self, False) + self.name = name or '' + self.starting_linenum = linenum + self.check_namespace_indentation = True + + def CheckEnd(self, filename, clean_lines, linenum, error): + """Check end of namespace comments.""" + line = clean_lines.raw_lines[linenum] + + # Check how many lines is enclosed in this namespace. Don't issue + # warning for missing namespace comments if there aren't enough + # lines. However, do apply checks if there is already an end of + # namespace comment and it's incorrect. + # + # TODO(unknown): We always want to check end of namespace comments + # if a namespace is large, but sometimes we also want to apply the + # check if a short namespace contained nontrivial things (something + # other than forward declarations). There is currently no logic on + # deciding what these nontrivial things are, so this check is + # triggered by namespace size only, which works most of the time. + if (linenum - self.starting_linenum < 10 + and not Match(r'};*\s*(//|/\*).*\bnamespace\b', line)): + return + + # Look for matching comment at end of namespace. + # + # Note that we accept C style "/* */" comments for terminating + # namespaces, so that code that terminate namespaces inside + # preprocessor macros can be cpplint clean. + # + # We also accept stuff like "// end of namespace ." with the + # period at the end. + # + # Besides these, we don't accept anything else, otherwise we might + # get false negatives when existing comment is a substring of the + # expected namespace. + if self.name: + # Named namespace + if not Match((r'};*\s*(//|/\*).*\bnamespace\s+' + re.escape(self.name) + + r'[\*/\.\\\s]*$'), + line): + error(filename, linenum, 'readability/namespace', 5, + 'Namespace should be terminated with "// namespace %s"' % + self.name) + else: + # Anonymous namespace + if not Match(r'};*\s*(//|/\*).*\bnamespace[\*/\.\\\s]*$', line): + # If "// namespace anonymous" or "// anonymous namespace (more text)", + # mention "// anonymous namespace" as an acceptable form + if Match(r'}.*\b(namespace anonymous|anonymous namespace)\b', line): + error(filename, linenum, 'readability/namespace', 5, + 'Anonymous namespace should be terminated with "// namespace"' + ' or "// anonymous namespace"') + else: + error(filename, linenum, 'readability/namespace', 5, + 'Anonymous namespace should be terminated with "// namespace"') + + +class _PreprocessorInfo(object): + """Stores checkpoints of nesting stacks when #if/#else is seen.""" + + def __init__(self, stack_before_if): + # The entire nesting stack before #if + self.stack_before_if = stack_before_if + + # The entire nesting stack up to #else + self.stack_before_else = [] + + # Whether we have already seen #else or #elif + self.seen_else = False + + +class NestingState(object): + """Holds states related to parsing braces.""" + + def __init__(self): + # Stack for tracking all braces. An object is pushed whenever we + # see a "{", and popped when we see a "}". Only 3 types of + # objects are possible: + # - _ClassInfo: a class or struct. + # - _NamespaceInfo: a namespace. + # - _BlockInfo: some other type of block. + self.stack = [] + + # Top of the previous stack before each Update(). + # + # Because the nesting_stack is updated at the end of each line, we + # had to do some convoluted checks to find out what is the current + # scope at the beginning of the line. This check is simplified by + # saving the previous top of nesting stack. + # + # We could save the full stack, but we only need the top. Copying + # the full nesting stack would slow down cpplint by ~10%. + self.previous_stack_top = [] + + # Stack of _PreprocessorInfo objects. + self.pp_stack = [] + + def SeenOpenBrace(self): + """Check if we have seen the opening brace for the innermost block. + + Returns: + True if we have seen the opening brace, False if the innermost + block is still expecting an opening brace. + """ + return (not self.stack) or self.stack[-1].seen_open_brace + + def InNamespaceBody(self): + """Check if we are currently one level inside a namespace body. + + Returns: + True if top of the stack is a namespace block, False otherwise. + """ + return self.stack and isinstance(self.stack[-1], _NamespaceInfo) + + def InExternC(self): + """Check if we are currently one level inside an 'extern "C"' block. + + Returns: + True if top of the stack is an extern block, False otherwise. + """ + return self.stack and isinstance(self.stack[-1], _ExternCInfo) + + def InClassDeclaration(self): + """Check if we are currently one level inside a class or struct declaration. + + Returns: + True if top of the stack is a class/struct, False otherwise. + """ + return self.stack and isinstance(self.stack[-1], _ClassInfo) + + def InAsmBlock(self): + """Check if we are currently one level inside an inline ASM block. + + Returns: + True if the top of the stack is a block containing inline ASM. + """ + return self.stack and self.stack[-1].inline_asm != _NO_ASM + + def InTemplateArgumentList(self, clean_lines, linenum, pos): + """Check if current position is inside template argument list. + + Args: + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + pos: position just after the suspected template argument. + Returns: + True if (linenum, pos) is inside template arguments. + """ + while linenum < clean_lines.NumLines(): + # Find the earliest character that might indicate a template argument + line = clean_lines.elided[linenum] + match = Match(r'^[^{};=\[\]\.<>]*(.)', line[pos:]) + if not match: + linenum += 1 + pos = 0 + continue + token = match.group(1) + pos += len(match.group(0)) + + # These things do not look like template argument list: + # class Suspect { + # class Suspect x; } + if token in ('{', '}', ';'): return False + + # These things look like template argument list: + # template + # template + # template + # template + if token in ('>', '=', '[', ']', '.'): return True + + # Check if token is an unmatched '<'. + # If not, move on to the next character. + if token != '<': + pos += 1 + if pos >= len(line): + linenum += 1 + pos = 0 + continue + + # We can't be sure if we just find a single '<', and need to + # find the matching '>'. + (_, end_line, end_pos) = CloseExpression(clean_lines, linenum, pos - 1) + if end_pos < 0: + # Not sure if template argument list or syntax error in file + return False + linenum = end_line + pos = end_pos + return False + + def UpdatePreprocessor(self, line): + """Update preprocessor stack. + + We need to handle preprocessors due to classes like this: + #ifdef SWIG + struct ResultDetailsPageElementExtensionPoint { + #else + struct ResultDetailsPageElementExtensionPoint : public Extension { + #endif + + We make the following assumptions (good enough for most files): + - Preprocessor condition evaluates to true from #if up to first + #else/#elif/#endif. + + - Preprocessor condition evaluates to false from #else/#elif up + to #endif. We still perform lint checks on these lines, but + these do not affect nesting stack. + + Args: + line: current line to check. + """ + if Match(r'^\s*#\s*(if|ifdef|ifndef)\b', line): + # Beginning of #if block, save the nesting stack here. The saved + # stack will allow us to restore the parsing state in the #else case. + self.pp_stack.append(_PreprocessorInfo(copy.deepcopy(self.stack))) + elif Match(r'^\s*#\s*(else|elif)\b', line): + # Beginning of #else block + if self.pp_stack: + if not self.pp_stack[-1].seen_else: + # This is the first #else or #elif block. Remember the + # whole nesting stack up to this point. This is what we + # keep after the #endif. + self.pp_stack[-1].seen_else = True + self.pp_stack[-1].stack_before_else = copy.deepcopy(self.stack) + + # Restore the stack to how it was before the #if + self.stack = copy.deepcopy(self.pp_stack[-1].stack_before_if) + else: + # TODO(unknown): unexpected #else, issue warning? + pass + elif Match(r'^\s*#\s*endif\b', line): + # End of #if or #else blocks. + if self.pp_stack: + # If we saw an #else, we will need to restore the nesting + # stack to its former state before the #else, otherwise we + # will just continue from where we left off. + if self.pp_stack[-1].seen_else: + # Here we can just use a shallow copy since we are the last + # reference to it. + self.stack = self.pp_stack[-1].stack_before_else + # Drop the corresponding #if + self.pp_stack.pop() + else: + # TODO(unknown): unexpected #endif, issue warning? + pass + + # TODO(unknown): Update() is too long, but we will refactor later. + def Update(self, filename, clean_lines, linenum, error): + """Update nesting state with current line. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + line = clean_lines.elided[linenum] + + # Remember top of the previous nesting stack. + # + # The stack is always pushed/popped and not modified in place, so + # we can just do a shallow copy instead of copy.deepcopy. Using + # deepcopy would slow down cpplint by ~28%. + if self.stack: + self.previous_stack_top = self.stack[-1] + else: + self.previous_stack_top = None + + # Update pp_stack + self.UpdatePreprocessor(line) + + # Count parentheses. This is to avoid adding struct arguments to + # the nesting stack. + if self.stack: + inner_block = self.stack[-1] + depth_change = line.count('(') - line.count(')') + inner_block.open_parentheses += depth_change + + # Also check if we are starting or ending an inline assembly block. + if inner_block.inline_asm in (_NO_ASM, _END_ASM): + if (depth_change != 0 and + inner_block.open_parentheses == 1 and + _MATCH_ASM.match(line)): + # Enter assembly block + inner_block.inline_asm = _INSIDE_ASM + else: + # Not entering assembly block. If previous line was _END_ASM, + # we will now shift to _NO_ASM state. + inner_block.inline_asm = _NO_ASM + elif (inner_block.inline_asm == _INSIDE_ASM and + inner_block.open_parentheses == 0): + # Exit assembly block + inner_block.inline_asm = _END_ASM + + # Consume namespace declaration at the beginning of the line. Do + # this in a loop so that we catch same line declarations like this: + # namespace proto2 { namespace bridge { class MessageSet; } } + while True: + # Match start of namespace. The "\b\s*" below catches namespace + # declarations even if it weren't followed by a whitespace, this + # is so that we don't confuse our namespace checker. The + # missing spaces will be flagged by CheckSpacing. + namespace_decl_match = Match(r'^\s*namespace\b\s*([:\w]+)?(.*)$', line) + if not namespace_decl_match: + break + + new_namespace = _NamespaceInfo(namespace_decl_match.group(1), linenum) + self.stack.append(new_namespace) + + line = namespace_decl_match.group(2) + if line.find('{') != -1: + new_namespace.seen_open_brace = True + line = line[line.find('{') + 1:] + + # Look for a class declaration in whatever is left of the line + # after parsing namespaces. The regexp accounts for decorated classes + # such as in: + # class LOCKABLE API Object { + # }; + class_decl_match = Match( + r'^(\s*(?:template\s*<[\w\s<>,:]*>\s*)?' + r'(class|struct)\s+(?:[A-Z_]+\s+)*(\w+(?:::\w+)*))' + r'(.*)$', line) + if (class_decl_match and + (not self.stack or self.stack[-1].open_parentheses == 0)): + # We do not want to accept classes that are actually template arguments: + # template , + # template class Ignore3> + # void Function() {}; + # + # To avoid template argument cases, we scan forward and look for + # an unmatched '>'. If we see one, assume we are inside a + # template argument list. + end_declaration = len(class_decl_match.group(1)) + if not self.InTemplateArgumentList(clean_lines, linenum, end_declaration): + self.stack.append(_ClassInfo( + class_decl_match.group(3), class_decl_match.group(2), + clean_lines, linenum)) + line = class_decl_match.group(4) + + # If we have not yet seen the opening brace for the innermost block, + # run checks here. + if not self.SeenOpenBrace(): + self.stack[-1].CheckBegin(filename, clean_lines, linenum, error) + + # Update access control if we are inside a class/struct + if self.stack and isinstance(self.stack[-1], _ClassInfo): + classinfo = self.stack[-1] + access_match = Match( + r'^(.*)\b(public|private|protected|signals)(\s+(?:slots\s*)?)?' + r':(?:[^:]|$)', + line) + if access_match: + classinfo.access = access_match.group(2) + + # Check that access keywords are indented +1 space. Skip this + # check if the keywords are not preceded by whitespaces. + indent = access_match.group(1) + if (len(indent) != classinfo.class_indent + 1 and + Match(r'^\s*$', indent)): + if classinfo.is_struct: + parent = 'struct ' + classinfo.name + else: + parent = 'class ' + classinfo.name + slots = '' + if access_match.group(3): + slots = access_match.group(3) + error(filename, linenum, 'whitespace/indent', 3, + '%s%s: should be indented +1 space inside %s' % ( + access_match.group(2), slots, parent)) + + # Consume braces or semicolons from what's left of the line + while True: + # Match first brace, semicolon, or closed parenthesis. + matched = Match(r'^[^{;)}]*([{;)}])(.*)$', line) + if not matched: + break + + token = matched.group(1) + if token == '{': + # If namespace or class hasn't seen a opening brace yet, mark + # namespace/class head as complete. Push a new block onto the + # stack otherwise. + if not self.SeenOpenBrace(): + self.stack[-1].seen_open_brace = True + elif Match(r'^extern\s*"[^"]*"\s*\{', line): + self.stack.append(_ExternCInfo()) + else: + self.stack.append(_BlockInfo(True)) + if _MATCH_ASM.match(line): + self.stack[-1].inline_asm = _BLOCK_ASM + + elif token == ';' or token == ')': + # If we haven't seen an opening brace yet, but we already saw + # a semicolon, this is probably a forward declaration. Pop + # the stack for these. + # + # Similarly, if we haven't seen an opening brace yet, but we + # already saw a closing parenthesis, then these are probably + # function arguments with extra "class" or "struct" keywords. + # Also pop these stack for these. + if not self.SeenOpenBrace(): + self.stack.pop() + else: # token == '}' + # Perform end of block checks and pop the stack. + if self.stack: + self.stack[-1].CheckEnd(filename, clean_lines, linenum, error) + self.stack.pop() + line = matched.group(2) + + def InnermostClass(self): + """Get class info on the top of the stack. + + Returns: + A _ClassInfo object if we are inside a class, or None otherwise. + """ + for i in range(len(self.stack), 0, -1): + classinfo = self.stack[i - 1] + if isinstance(classinfo, _ClassInfo): + return classinfo + return None + + def CheckCompletedBlocks(self, filename, error): + """Checks that all classes and namespaces have been completely parsed. + + Call this when all lines in a file have been processed. + Args: + filename: The name of the current file. + error: The function to call with any errors found. + """ + # Note: This test can result in false positives if #ifdef constructs + # get in the way of brace matching. See the testBuildClass test in + # cpplint_unittest.py for an example of this. + for obj in self.stack: + if isinstance(obj, _ClassInfo): + error(filename, obj.starting_linenum, 'build/class', 5, + 'Failed to find complete declaration of class %s' % + obj.name) + elif isinstance(obj, _NamespaceInfo): + error(filename, obj.starting_linenum, 'build/namespaces', 5, + 'Failed to find complete declaration of namespace %s' % + obj.name) + + +def CheckForNonStandardConstructs(filename, clean_lines, linenum, + nesting_state, error): + r"""Logs an error if we see certain non-ANSI constructs ignored by gcc-2. + + Complain about several constructs which gcc-2 accepts, but which are + not standard C++. Warning about these in lint is one way to ease the + transition to new compilers. + - put storage class first (e.g. "static const" instead of "const static"). + - "%lld" instead of %qd" in printf-type functions. + - "%1$d" is non-standard in printf-type functions. + - "\%" is an undefined character escape sequence. + - text after #endif is not allowed. + - invalid inner-style forward declaration. + - >? and ?= and )\?=?\s*(\w+|[+-]?\d+)(\.\d*)?', + line): + error(filename, linenum, 'build/deprecated', 3, + '>? and ))?' + # r'\s*const\s*' + type_name + '\s*&\s*\w+\s*;' + error(filename, linenum, 'runtime/member_string_references', 2, + 'const string& members are dangerous. It is much better to use ' + 'alternatives, such as pointers or simple constants.') + + # Everything else in this function operates on class declarations. + # Return early if the top of the nesting stack is not a class, or if + # the class head is not completed yet. + classinfo = nesting_state.InnermostClass() + if not classinfo or not classinfo.seen_open_brace: + return + + # The class may have been declared with namespace or classname qualifiers. + # The constructor and destructor will not have those qualifiers. + base_classname = classinfo.name.split('::')[-1] + + # Look for single-argument constructors that aren't marked explicit. + # Technically a valid construct, but against style. Also look for + # non-single-argument constructors which are also technically valid, but + # strongly suggest something is wrong. + explicit_constructor_match = Match( + r'\s+(?:inline\s+)?(explicit\s+)?(?:inline\s+)?%s\s*' + r'\(((?:[^()]|\([^()]*\))*)\)' + % re.escape(base_classname), + line) + + if explicit_constructor_match: + is_marked_explicit = explicit_constructor_match.group(1) + + if not explicit_constructor_match.group(2): + constructor_args = [] + else: + constructor_args = explicit_constructor_match.group(2).split(',') + + # collapse arguments so that commas in template parameter lists and function + # argument parameter lists don't split arguments in two + i = 0 + while i < len(constructor_args): + constructor_arg = constructor_args[i] + while (constructor_arg.count('<') > constructor_arg.count('>') or + constructor_arg.count('(') > constructor_arg.count(')')): + constructor_arg += ',' + constructor_args[i + 1] + del constructor_args[i + 1] + constructor_args[i] = constructor_arg + i += 1 + + defaulted_args = [arg for arg in constructor_args if '=' in arg] + noarg_constructor = (not constructor_args or # empty arg list + # 'void' arg specifier + (len(constructor_args) == 1 and + constructor_args[0].strip() == 'void')) + onearg_constructor = ((len(constructor_args) == 1 and # exactly one arg + not noarg_constructor) or + # all but at most one arg defaulted + (len(constructor_args) >= 1 and + not noarg_constructor and + len(defaulted_args) >= len(constructor_args) - 1)) + initializer_list_constructor = bool( + onearg_constructor and + Search(r'\bstd\s*::\s*initializer_list\b', constructor_args[0])) + copy_constructor = bool( + onearg_constructor and + Match(r'(const\s+)?%s(\s*<[^>]*>)?(\s+const)?\s*(?:<\w+>\s*)?&' + % re.escape(base_classname), constructor_args[0].strip())) + + if (not is_marked_explicit and + onearg_constructor and + not initializer_list_constructor and + not copy_constructor): + if defaulted_args: + error(filename, linenum, 'runtime/explicit', 5, + 'Constructors callable with one argument ' + 'should be marked explicit.') + else: + error(filename, linenum, 'runtime/explicit', 5, + 'Single-parameter constructors should be marked explicit.') + elif is_marked_explicit and not onearg_constructor: + if noarg_constructor: + error(filename, linenum, 'runtime/explicit', 5, + 'Zero-parameter constructors should not be marked explicit.') + else: + error(filename, linenum, 'runtime/explicit', 0, + 'Constructors that require multiple arguments ' + 'should not be marked explicit.') + + +def CheckSpacingForFunctionCall(filename, clean_lines, linenum, error): + """Checks for the correctness of various spacing around function calls. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + line = clean_lines.elided[linenum] + + # Since function calls often occur inside if/for/while/switch + # expressions - which have their own, more liberal conventions - we + # first see if we should be looking inside such an expression for a + # function call, to which we can apply more strict standards. + fncall = line # if there's no control flow construct, look at whole line + for pattern in (r'\bif\s*\((.*)\)\s*{', + r'\bfor\s*\((.*)\)\s*{', + r'\bwhile\s*\((.*)\)\s*[{;]', + r'\bswitch\s*\((.*)\)\s*{'): + match = Search(pattern, line) + if match: + fncall = match.group(1) # look inside the parens for function calls + break + + # Except in if/for/while/switch, there should never be space + # immediately inside parens (eg "f( 3, 4 )"). We make an exception + # for nested parens ( (a+b) + c ). Likewise, there should never be + # a space before a ( when it's a function argument. I assume it's a + # function argument when the char before the whitespace is legal in + # a function name (alnum + _) and we're not starting a macro. Also ignore + # pointers and references to arrays and functions coz they're too tricky: + # we use a very simple way to recognize these: + # " (something)(maybe-something)" or + # " (something)(maybe-something," or + # " (something)[something]" + # Note that we assume the contents of [] to be short enough that + # they'll never need to wrap. + if ( # Ignore control structures. + not Search(r'\b(if|for|while|switch|return|new|delete|catch|sizeof)\b', + fncall) and + # Ignore pointers/references to functions. + not Search(r' \([^)]+\)\([^)]*(\)|,$)', fncall) and + # Ignore pointers/references to arrays. + not Search(r' \([^)]+\)\[[^\]]+\]', fncall)): + if Search(r'\w\s*\(\s(?!\s*\\$)', fncall): # a ( used for a fn call + error(filename, linenum, 'whitespace/parens', 4, + 'Extra space after ( in function call') + elif Search(r'\(\s+(?!(\s*\\)|\()', fncall): + error(filename, linenum, 'whitespace/parens', 2, + 'Extra space after (') + if (Search(r'\w\s+\(', fncall) and + not Search(r'#\s*define|typedef|using\s+\w+\s*=', fncall) and + not Search(r'\w\s+\((\w+::)*\*\w+\)\(', fncall) and + not Search(r'\bcase\s+\(', fncall)): + # TODO(unknown): Space after an operator function seem to be a common + # error, silence those for now by restricting them to highest verbosity. + if Search(r'\boperator_*\b', line): + error(filename, linenum, 'whitespace/parens', 0, + 'Extra space before ( in function call') + else: + error(filename, linenum, 'whitespace/parens', 4, + 'Extra space before ( in function call') + # If the ) is followed only by a newline or a { + newline, assume it's + # part of a control statement (if/while/etc), and don't complain + if Search(r'[^)]\s+\)\s*[^{\s]', fncall): + # If the closing parenthesis is preceded by only whitespaces, + # try to give a more descriptive error message. + if Search(r'^\s+\)', fncall): + error(filename, linenum, 'whitespace/parens', 2, + 'Closing ) should be moved to the previous line') + else: + error(filename, linenum, 'whitespace/parens', 2, + 'Extra space before )') + + +def IsBlankLine(line): + """Returns true if the given line is blank. + + We consider a line to be blank if the line is empty or consists of + only white spaces. + + Args: + line: A line of a string. + + Returns: + True, if the given line is blank. + """ + return not line or line.isspace() + + +def CheckForNamespaceIndentation(filename, nesting_state, clean_lines, line, + error): + is_namespace_indent_item = ( + len(nesting_state.stack) > 1 and + nesting_state.stack[-1].check_namespace_indentation and + isinstance(nesting_state.previous_stack_top, _NamespaceInfo) and + nesting_state.previous_stack_top == nesting_state.stack[-2]) + + if ShouldCheckNamespaceIndentation(nesting_state, is_namespace_indent_item, + clean_lines.elided, line): + CheckItemIndentationInNamespace(filename, clean_lines.elided, + line, error) + + +def CheckForFunctionLengths(filename, clean_lines, linenum, + function_state, error): + """Reports for long function bodies. + + For an overview why this is done, see: + http://google-styleguide.googlecode.com/svn/trunk/cppguide.xml#Write_Short_Functions + + Uses a simplistic algorithm assuming other style guidelines + (especially spacing) are followed. + Only checks unindented functions, so class members are unchecked. + Trivial bodies are unchecked, so constructors with huge initializer lists + may be missed. + Blank/comment lines are not counted so as to avoid encouraging the removal + of vertical space and comments just to get through a lint check. + NOLINT *on the last line of a function* disables this check. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + function_state: Current function name and lines in body so far. + error: The function to call with any errors found. + """ + lines = clean_lines.lines + line = lines[linenum] + joined_line = '' + + starting_func = False + regexp = r'(\w(\w|::|\*|\&|\s)*)\(' # decls * & space::name( ... + match_result = Match(regexp, line) + if match_result: + # If the name is all caps and underscores, figure it's a macro and + # ignore it, unless it's TEST or TEST_F. + function_name = match_result.group(1).split()[-1] + if function_name == 'TEST' or function_name == 'TEST_F' or ( + not Match(r'[A-Z_]+$', function_name)): + starting_func = True + + if starting_func: + body_found = False + for start_linenum in xrange(linenum, clean_lines.NumLines()): + start_line = lines[start_linenum] + joined_line += ' ' + start_line.lstrip() + if Search(r'(;|})', start_line): # Declarations and trivial functions + body_found = True + break # ... ignore + elif Search(r'{', start_line): + body_found = True + function = Search(r'((\w|:)*)\(', line).group(1) + if Match(r'TEST', function): # Handle TEST... macros + parameter_regexp = Search(r'(\(.*\))', joined_line) + if parameter_regexp: # Ignore bad syntax + function += parameter_regexp.group(1) + else: + function += '()' + function_state.Begin(function) + break + if not body_found: + # No body for the function (or evidence of a non-function) was found. + error(filename, linenum, 'readability/fn_size', 5, + 'Lint failed to find start of function body.') + elif Match(r'^\}\s*$', line): # function end + function_state.Check(error, filename, linenum) + function_state.End() + elif not Match(r'^\s*$', line): + function_state.Count() # Count non-blank/non-comment lines. + + +_RE_PATTERN_TODO = re.compile(r'^//(\s*)TODO(\(.+?\))?:?(\s|$)?') + + +def CheckComment(line, filename, linenum, next_line_start, error): + """Checks for common mistakes in comments. + + Args: + line: The line in question. + filename: The name of the current file. + linenum: The number of the line to check. + next_line_start: The first non-whitespace column of the next line. + error: The function to call with any errors found. + """ + commentpos = line.find('//') + if commentpos != -1: + # Check if the // may be in quotes. If so, ignore it + # Comparisons made explicit for clarity -- pylint: disable=g-explicit-bool-comparison + if (line.count('"', 0, commentpos) - + line.count('\\"', 0, commentpos)) % 2 == 0: # not in quotes + # Allow one space for new scopes, two spaces otherwise: + if (not (Match(r'^.*{ *//', line) and next_line_start == commentpos) and + ((commentpos >= 1 and + line[commentpos-1] not in string.whitespace) or + (commentpos >= 2 and + line[commentpos-2] not in string.whitespace))): + error(filename, linenum, 'whitespace/comments', 2, + 'At least two spaces is best between code and comments') + + # Checks for common mistakes in TODO comments. + comment = line[commentpos:] + match = _RE_PATTERN_TODO.match(comment) + if match: + # One whitespace is correct; zero whitespace is handled elsewhere. + leading_whitespace = match.group(1) + if len(leading_whitespace) > 1: + error(filename, linenum, 'whitespace/todo', 2, + 'Too many spaces before TODO') + + username = match.group(2) + if not username: + error(filename, linenum, 'readability/todo', 2, + 'Missing username in TODO; it should look like ' + '"// TODO(my_username): Stuff."') + + middle_whitespace = match.group(3) + # Comparisons made explicit for correctness -- pylint: disable=g-explicit-bool-comparison + if middle_whitespace != ' ' and middle_whitespace != '': + error(filename, linenum, 'whitespace/todo', 2, + 'TODO(my_username) should be followed by a space') + + # If the comment contains an alphanumeric character, there + # should be a space somewhere between it and the // unless + # it's a /// or //! Doxygen comment. + if (Match(r'//[^ ]*\w', comment) and + not Match(r'(///|//\!)(\s+|$)', comment)): + error(filename, linenum, 'whitespace/comments', 4, + 'Should have a space between // and comment') + + +def CheckAccess(filename, clean_lines, linenum, nesting_state, error): + """Checks for improper use of DISALLOW* macros. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + nesting_state: A NestingState instance which maintains information about + the current stack of nested blocks being parsed. + error: The function to call with any errors found. + """ + line = clean_lines.elided[linenum] # get rid of comments and strings + + matched = Match((r'\s*(DISALLOW_COPY_AND_ASSIGN|' + r'DISALLOW_IMPLICIT_CONSTRUCTORS)'), line) + if not matched: + return + if nesting_state.stack and isinstance(nesting_state.stack[-1], _ClassInfo): + if nesting_state.stack[-1].access != 'private': + error(filename, linenum, 'readability/constructors', 3, + '%s must be in the private: section' % matched.group(1)) + + else: + # Found DISALLOW* macro outside a class declaration, or perhaps it + # was used inside a function when it should have been part of the + # class declaration. We could issue a warning here, but it + # probably resulted in a compiler error already. + pass + + +def CheckSpacing(filename, clean_lines, linenum, nesting_state, error): + """Checks for the correctness of various spacing issues in the code. + + Things we check for: spaces around operators, spaces after + if/for/while/switch, no spaces around parens in function calls, two + spaces between code and comment, don't start a block with a blank + line, don't end a function with a blank line, don't add a blank line + after public/protected/private, don't have too many blank lines in a row. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + nesting_state: A NestingState instance which maintains information about + the current stack of nested blocks being parsed. + error: The function to call with any errors found. + """ + + # Don't use "elided" lines here, otherwise we can't check commented lines. + # Don't want to use "raw" either, because we don't want to check inside C++11 + # raw strings, + raw = clean_lines.lines_without_raw_strings + line = raw[linenum] + + # Before nixing comments, check if the line is blank for no good + # reason. This includes the first line after a block is opened, and + # blank lines at the end of a function (ie, right before a line like '}' + # + # Skip all the blank line checks if we are immediately inside a + # namespace body. In other words, don't issue blank line warnings + # for this block: + # namespace { + # + # } + # + # A warning about missing end of namespace comments will be issued instead. + # + # Also skip blank line checks for 'extern "C"' blocks, which are formatted + # like namespaces. + if (IsBlankLine(line) and + not nesting_state.InNamespaceBody() and + not nesting_state.InExternC()): + elided = clean_lines.elided + prev_line = elided[linenum - 1] + prevbrace = prev_line.rfind('{') + # TODO(unknown): Don't complain if line before blank line, and line after, + # both start with alnums and are indented the same amount. + # This ignores whitespace at the start of a namespace block + # because those are not usually indented. + if prevbrace != -1 and prev_line[prevbrace:].find('}') == -1: + # OK, we have a blank line at the start of a code block. Before we + # complain, we check if it is an exception to the rule: The previous + # non-empty line has the parameters of a function header that are indented + # 4 spaces (because they did not fit in a 80 column line when placed on + # the same line as the function name). We also check for the case where + # the previous line is indented 6 spaces, which may happen when the + # initializers of a constructor do not fit into a 80 column line. + exception = False + if Match(r' {6}\w', prev_line): # Initializer list? + # We are looking for the opening column of initializer list, which + # should be indented 4 spaces to cause 6 space indentation afterwards. + search_position = linenum-2 + while (search_position >= 0 + and Match(r' {6}\w', elided[search_position])): + search_position -= 1 + exception = (search_position >= 0 + and elided[search_position][:5] == ' :') + else: + # Search for the function arguments or an initializer list. We use a + # simple heuristic here: If the line is indented 4 spaces; and we have a + # closing paren, without the opening paren, followed by an opening brace + # or colon (for initializer lists) we assume that it is the last line of + # a function header. If we have a colon indented 4 spaces, it is an + # initializer list. + exception = (Match(r' {4}\w[^\(]*\)\s*(const\s*)?(\{\s*$|:)', + prev_line) + or Match(r' {4}:', prev_line)) + + if not exception: + error(filename, linenum, 'whitespace/blank_line', 2, + 'Redundant blank line at the start of a code block ' + 'should be deleted.') + # Ignore blank lines at the end of a block in a long if-else + # chain, like this: + # if (condition1) { + # // Something followed by a blank line + # + # } else if (condition2) { + # // Something else + # } + if linenum + 1 < clean_lines.NumLines(): + next_line = raw[linenum + 1] + if (next_line + and Match(r'\s*}', next_line) + and next_line.find('} else ') == -1): + error(filename, linenum, 'whitespace/blank_line', 3, + 'Redundant blank line at the end of a code block ' + 'should be deleted.') + + matched = Match(r'\s*(public|protected|private):', prev_line) + if matched: + error(filename, linenum, 'whitespace/blank_line', 3, + 'Do not leave a blank line after "%s:"' % matched.group(1)) + + # Next, check comments + next_line_start = 0 + if linenum + 1 < clean_lines.NumLines(): + next_line = raw[linenum + 1] + next_line_start = len(next_line) - len(next_line.lstrip()) + CheckComment(line, filename, linenum, next_line_start, error) + + # get rid of comments and strings + line = clean_lines.elided[linenum] + + # You shouldn't have spaces before your brackets, except maybe after + # 'delete []' or 'return []() {};' + if Search(r'\w\s+\[', line) and not Search(r'(?:delete|return)\s+\[', line): + error(filename, linenum, 'whitespace/braces', 5, + 'Extra space before [') + + # In range-based for, we wanted spaces before and after the colon, but + # not around "::" tokens that might appear. + if (Search(r'for *\(.*[^:]:[^: ]', line) or + Search(r'for *\(.*[^: ]:[^:]', line)): + error(filename, linenum, 'whitespace/forcolon', 2, + 'Missing space around colon in range-based for loop') + + +def CheckOperatorSpacing(filename, clean_lines, linenum, error): + """Checks for horizontal spacing around operators. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + line = clean_lines.elided[linenum] + + # Don't try to do spacing checks for operator methods. Do this by + # replacing the troublesome characters with something else, + # preserving column position for all other characters. + # + # The replacement is done repeatedly to avoid false positives from + # operators that call operators. + while True: + match = Match(r'^(.*\boperator\b)(\S+)(\s*\(.*)$', line) + if match: + line = match.group(1) + ('_' * len(match.group(2))) + match.group(3) + else: + break + + # We allow no-spaces around = within an if: "if ( (a=Foo()) == 0 )". + # Otherwise not. Note we only check for non-spaces on *both* sides; + # sometimes people put non-spaces on one side when aligning ='s among + # many lines (not that this is behavior that I approve of...) + if ((Search(r'[\w.]=', line) or + Search(r'=[\w.]', line)) + and not Search(r'\b(if|while|for) ', line) + # Operators taken from [lex.operators] in C++11 standard. + and not Search(r'(>=|<=|==|!=|&=|\^=|\|=|\+=|\*=|\/=|\%=)', line) + and not Search(r'operator=', line)): + error(filename, linenum, 'whitespace/operators', 4, + 'Missing spaces around =') + + # It's ok not to have spaces around binary operators like + - * /, but if + # there's too little whitespace, we get concerned. It's hard to tell, + # though, so we punt on this one for now. TODO. + + # You should always have whitespace around binary operators. + # + # Check <= and >= first to avoid false positives with < and >, then + # check non-include lines for spacing around < and >. + # + # If the operator is followed by a comma, assume it's be used in a + # macro context and don't do any checks. This avoids false + # positives. + # + # Note that && is not included here. Those are checked separately + # in CheckRValueReference + match = Search(r'[^<>=!\s](==|!=|<=|>=|\|\|)[^<>=!\s,;\)]', line) + if match: + error(filename, linenum, 'whitespace/operators', 3, + 'Missing spaces around %s' % match.group(1)) + elif not Match(r'#.*include', line): + # Look for < that is not surrounded by spaces. This is only + # triggered if both sides are missing spaces, even though + # technically should should flag if at least one side is missing a + # space. This is done to avoid some false positives with shifts. + match = Match(r'^(.*[^\s<])<[^\s=<,]', line) + if match: + (_, _, end_pos) = CloseExpression( + clean_lines, linenum, len(match.group(1))) + if end_pos <= -1: + error(filename, linenum, 'whitespace/operators', 3, + 'Missing spaces around <') + + # Look for > that is not surrounded by spaces. Similar to the + # above, we only trigger if both sides are missing spaces to avoid + # false positives with shifts. + match = Match(r'^(.*[^-\s>])>[^\s=>,]', line) + if match: + (_, _, start_pos) = ReverseCloseExpression( + clean_lines, linenum, len(match.group(1))) + if start_pos <= -1: + error(filename, linenum, 'whitespace/operators', 3, + 'Missing spaces around >') + + # We allow no-spaces around << when used like this: 10<<20, but + # not otherwise (particularly, not when used as streams) + # + # We also allow operators following an opening parenthesis, since + # those tend to be macros that deal with operators. + match = Search(r'(operator|[^\s(<])(?:L|UL|ULL|l|ul|ull)?<<([^\s,=<])', line) + if (match and not (match.group(1).isdigit() and match.group(2).isdigit()) and + not (match.group(1) == 'operator' and match.group(2) == ';')): + error(filename, linenum, 'whitespace/operators', 3, + 'Missing spaces around <<') + + # We allow no-spaces around >> for almost anything. This is because + # C++11 allows ">>" to close nested templates, which accounts for + # most cases when ">>" is not followed by a space. + # + # We still warn on ">>" followed by alpha character, because that is + # likely due to ">>" being used for right shifts, e.g.: + # value >> alpha + # + # When ">>" is used to close templates, the alphanumeric letter that + # follows would be part of an identifier, and there should still be + # a space separating the template type and the identifier. + # type> alpha + match = Search(r'>>[a-zA-Z_]', line) + if match: + error(filename, linenum, 'whitespace/operators', 3, + 'Missing spaces around >>') + + # There shouldn't be space around unary operators + match = Search(r'(!\s|~\s|[\s]--[\s;]|[\s]\+\+[\s;])', line) + if match: + error(filename, linenum, 'whitespace/operators', 4, + 'Extra space for operator %s' % match.group(1)) + + +def CheckParenthesisSpacing(filename, clean_lines, linenum, error): + """Checks for horizontal spacing around parentheses. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + line = clean_lines.elided[linenum] + + # No spaces after an if, while, switch, or for + match = Search(r' (if\(|for\(|while\(|switch\()', line) + if match: + error(filename, linenum, 'whitespace/parens', 5, + 'Missing space before ( in %s' % match.group(1)) + + # For if/for/while/switch, the left and right parens should be + # consistent about how many spaces are inside the parens, and + # there should either be zero or one spaces inside the parens. + # We don't want: "if ( foo)" or "if ( foo )". + # Exception: "for ( ; foo; bar)" and "for (foo; bar; )" are allowed. + match = Search(r'\b(if|for|while|switch)\s*' + r'\(([ ]*)(.).*[^ ]+([ ]*)\)\s*{\s*$', + line) + if match: + if len(match.group(2)) != len(match.group(4)): + if not (match.group(3) == ';' and + len(match.group(2)) == 1 + len(match.group(4)) or + not match.group(2) and Search(r'\bfor\s*\(.*; \)', line)): + error(filename, linenum, 'whitespace/parens', 5, + 'Mismatching spaces inside () in %s' % match.group(1)) + if len(match.group(2)) not in [0, 1]: + error(filename, linenum, 'whitespace/parens', 5, + 'Should have zero or one spaces inside ( and ) in %s' % + match.group(1)) + + +def CheckCommaSpacing(filename, clean_lines, linenum, error): + """Checks for horizontal spacing near commas and semicolons. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + raw = clean_lines.lines_without_raw_strings + line = clean_lines.elided[linenum] + + # You should always have a space after a comma (either as fn arg or operator) + # + # This does not apply when the non-space character following the + # comma is another comma, since the only time when that happens is + # for empty macro arguments. + # + # We run this check in two passes: first pass on elided lines to + # verify that lines contain missing whitespaces, second pass on raw + # lines to confirm that those missing whitespaces are not due to + # elided comments. + if (Search(r',[^,\s]', ReplaceAll(r'\boperator\s*,\s*\(', 'F(', line)) and + Search(r',[^,\s]', raw[linenum])): + error(filename, linenum, 'whitespace/comma', 3, + 'Missing space after ,') + + # You should always have a space after a semicolon + # except for few corner cases + # TODO(unknown): clarify if 'if (1) { return 1;}' is requires one more + # space after ; + if Search(r';[^\s};\\)/]', line): + error(filename, linenum, 'whitespace/semicolon', 3, + 'Missing space after ;') + + +def CheckBracesSpacing(filename, clean_lines, linenum, error): + """Checks for horizontal spacing near commas. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + line = clean_lines.elided[linenum] + + # Except after an opening paren, or after another opening brace (in case of + # an initializer list, for instance), you should have spaces before your + # braces. And since you should never have braces at the beginning of a line, + # this is an easy test. + match = Match(r'^(.*[^ ({>]){', line) + if match: + # Try a bit harder to check for brace initialization. This + # happens in one of the following forms: + # Constructor() : initializer_list_{} { ... } + # Constructor{}.MemberFunction() + # Type variable{}; + # FunctionCall(type{}, ...); + # LastArgument(..., type{}); + # LOG(INFO) << type{} << " ..."; + # map_of_type[{...}] = ...; + # ternary = expr ? new type{} : nullptr; + # OuterTemplate{}> + # + # We check for the character following the closing brace, and + # silence the warning if it's one of those listed above, i.e. + # "{.;,)<>]:". + # + # To account for nested initializer list, we allow any number of + # closing braces up to "{;,)<". We can't simply silence the + # warning on first sight of closing brace, because that would + # cause false negatives for things that are not initializer lists. + # Silence this: But not this: + # Outer{ if (...) { + # Inner{...} if (...){ // Missing space before { + # }; } + # + # There is a false negative with this approach if people inserted + # spurious semicolons, e.g. "if (cond){};", but we will catch the + # spurious semicolon with a separate check. + (endline, endlinenum, endpos) = CloseExpression( + clean_lines, linenum, len(match.group(1))) + trailing_text = '' + if endpos > -1: + trailing_text = endline[endpos:] + for offset in xrange(endlinenum + 1, + min(endlinenum + 3, clean_lines.NumLines() - 1)): + trailing_text += clean_lines.elided[offset] + if not Match(r'^[\s}]*[{.;,)<>\]:]', trailing_text): + error(filename, linenum, 'whitespace/braces', 5, + 'Missing space before {') + + # Make sure '} else {' has spaces. + if Search(r'}else', line): + error(filename, linenum, 'whitespace/braces', 5, + 'Missing space before else') + + # You shouldn't have a space before a semicolon at the end of the line. + # There's a special case for "for" since the style guide allows space before + # the semicolon there. + if Search(r':\s*;\s*$', line): + error(filename, linenum, 'whitespace/semicolon', 5, + 'Semicolon defining empty statement. Use {} instead.') + elif Search(r'^\s*;\s*$', line): + error(filename, linenum, 'whitespace/semicolon', 5, + 'Line contains only semicolon. If this should be an empty statement, ' + 'use {} instead.') + elif (Search(r'\s+;\s*$', line) and + not Search(r'\bfor\b', line)): + error(filename, linenum, 'whitespace/semicolon', 5, + 'Extra space before last semicolon. If this should be an empty ' + 'statement, use {} instead.') + + +def IsDecltype(clean_lines, linenum, column): + """Check if the token ending on (linenum, column) is decltype(). + + Args: + clean_lines: A CleansedLines instance containing the file. + linenum: the number of the line to check. + column: end column of the token to check. + Returns: + True if this token is decltype() expression, False otherwise. + """ + (text, _, start_col) = ReverseCloseExpression(clean_lines, linenum, column) + if start_col < 0: + return False + if Search(r'\bdecltype\s*$', text[0:start_col]): + return True + return False + + +def IsTemplateParameterList(clean_lines, linenum, column): + """Check if the token ending on (linenum, column) is the end of template<>. + + Args: + clean_lines: A CleansedLines instance containing the file. + linenum: the number of the line to check. + column: end column of the token to check. + Returns: + True if this token is end of a template parameter list, False otherwise. + """ + (_, startline, startpos) = ReverseCloseExpression( + clean_lines, linenum, column) + if (startpos > -1 and + Search(r'\btemplate\s*$', clean_lines.elided[startline][0:startpos])): + return True + return False + + +def IsRValueType(typenames, clean_lines, nesting_state, linenum, column): + """Check if the token ending on (linenum, column) is a type. + + Assumes that text to the right of the column is "&&" or a function + name. + + Args: + typenames: set of type names from template-argument-list. + clean_lines: A CleansedLines instance containing the file. + nesting_state: A NestingState instance which maintains information about + the current stack of nested blocks being parsed. + linenum: the number of the line to check. + column: end column of the token to check. + Returns: + True if this token is a type, False if we are not sure. + """ + prefix = clean_lines.elided[linenum][0:column] + + # Get one word to the left. If we failed to do so, this is most + # likely not a type, since it's unlikely that the type name and "&&" + # would be split across multiple lines. + match = Match(r'^(.*)(\b\w+|[>*)&])\s*$', prefix) + if not match: + return False + + # Check text following the token. If it's "&&>" or "&&," or "&&...", it's + # most likely a rvalue reference used inside a template. + suffix = clean_lines.elided[linenum][column:] + if Match(r'&&\s*(?:[>,]|\.\.\.)', suffix): + return True + + # Check for known types and end of templates: + # int&& variable + # vector&& variable + # + # Because this function is called recursively, we also need to + # recognize pointer and reference types: + # int* Function() + # int& Function() + if (match.group(2) in typenames or + match.group(2) in ['char', 'char16_t', 'char32_t', 'wchar_t', 'bool', + 'short', 'int', 'long', 'signed', 'unsigned', + 'float', 'double', 'void', 'auto', '>', '*', '&']): + return True + + # If we see a close parenthesis, look for decltype on the other side. + # decltype would unambiguously identify a type, anything else is + # probably a parenthesized expression and not a type. + if match.group(2) == ')': + return IsDecltype( + clean_lines, linenum, len(match.group(1)) + len(match.group(2)) - 1) + + # Check for casts and cv-qualifiers. + # match.group(1) remainder + # -------------- --------- + # const_cast< type&& + # const type&& + # type const&& + if Search(r'\b(?:const_cast\s*<|static_cast\s*<|dynamic_cast\s*<|' + r'reinterpret_cast\s*<|\w+\s)\s*$', + match.group(1)): + return True + + # Look for a preceding symbol that might help differentiate the context. + # These are the cases that would be ambiguous: + # match.group(1) remainder + # -------------- --------- + # Call ( expression && + # Declaration ( type&& + # sizeof ( type&& + # if ( expression && + # while ( expression && + # for ( type&& + # for( ; expression && + # statement ; type&& + # block { type&& + # constructor { expression && + start = linenum + line = match.group(1) + match_symbol = None + while start >= 0: + # We want to skip over identifiers and commas to get to a symbol. + # Commas are skipped so that we can find the opening parenthesis + # for function parameter lists. + match_symbol = Match(r'^(.*)([^\w\s,])[\w\s,]*$', line) + if match_symbol: + break + start -= 1 + line = clean_lines.elided[start] + + if not match_symbol: + # Probably the first statement in the file is an rvalue reference + return True + + if match_symbol.group(2) == '}': + # Found closing brace, probably an indicate of this: + # block{} type&& + return True + + if match_symbol.group(2) == ';': + # Found semicolon, probably one of these: + # for(; expression && + # statement; type&& + + # Look for the previous 'for(' in the previous lines. + before_text = match_symbol.group(1) + for i in xrange(start - 1, max(start - 6, 0), -1): + before_text = clean_lines.elided[i] + before_text + if Search(r'for\s*\([^{};]*$', before_text): + # This is the condition inside a for-loop + return False + + # Did not find a for-init-statement before this semicolon, so this + # is probably a new statement and not a condition. + return True + + if match_symbol.group(2) == '{': + # Found opening brace, probably one of these: + # block{ type&& = ... ; } + # constructor{ expression && expression } + + # Look for a closing brace or a semicolon. If we see a semicolon + # first, this is probably a rvalue reference. + line = clean_lines.elided[start][0:len(match_symbol.group(1)) + 1] + end = start + depth = 1 + while True: + for ch in line: + if ch == ';': + return True + elif ch == '{': + depth += 1 + elif ch == '}': + depth -= 1 + if depth == 0: + return False + end += 1 + if end >= clean_lines.NumLines(): + break + line = clean_lines.elided[end] + # Incomplete program? + return False + + if match_symbol.group(2) == '(': + # Opening parenthesis. Need to check what's to the left of the + # parenthesis. Look back one extra line for additional context. + before_text = match_symbol.group(1) + if linenum > 1: + before_text = clean_lines.elided[linenum - 1] + before_text + before_text = match_symbol.group(1) + + # Patterns that are likely to be types: + # [](type&& + # for (type&& + # sizeof(type&& + # operator=(type&& + # + if Search(r'(?:\]|\bfor|\bsizeof|\boperator\s*\S+\s*)\s*$', before_text): + return True + + # Patterns that are likely to be expressions: + # if (expression && + # while (expression && + # : initializer(expression && + # , initializer(expression && + # ( FunctionCall(expression && + # + FunctionCall(expression && + # + (expression && + # + # The last '+' represents operators such as '+' and '-'. + if Search(r'(?:\bif|\bwhile|[-+=%^(]*>)?\s*$', + match_symbol.group(1)) + if match_func: + # Check for constructors, which don't have return types. + if Search(r'\b(?:explicit|inline)$', match_func.group(1)): + return True + implicit_constructor = Match(r'\s*(\w+)\((?:const\s+)?(\w+)', prefix) + if (implicit_constructor and + implicit_constructor.group(1) == implicit_constructor.group(2)): + return True + return IsRValueType(typenames, clean_lines, nesting_state, linenum, + len(match_func.group(1))) + + # Nothing before the function name. If this is inside a block scope, + # this is probably a function call. + return not (nesting_state.previous_stack_top and + nesting_state.previous_stack_top.IsBlockInfo()) + + if match_symbol.group(2) == '>': + # Possibly a closing bracket, check that what's on the other side + # looks like the start of a template. + return IsTemplateParameterList( + clean_lines, start, len(match_symbol.group(1))) + + # Some other symbol, usually something like "a=b&&c". This is most + # likely not a type. + return False + + +def IsDeletedOrDefault(clean_lines, linenum): + """Check if current constructor or operator is deleted or default. + + Args: + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + Returns: + True if this is a deleted or default constructor. + """ + open_paren = clean_lines.elided[linenum].find('(') + if open_paren < 0: + return False + (close_line, _, close_paren) = CloseExpression( + clean_lines, linenum, open_paren) + if close_paren < 0: + return False + return Match(r'\s*=\s*(?:delete|default)\b', close_line[close_paren:]) + + +def IsRValueAllowed(clean_lines, linenum, typenames): + """Check if RValue reference is allowed on a particular line. + + Args: + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + typenames: set of type names from template-argument-list. + Returns: + True if line is within the region where RValue references are allowed. + """ + # Allow region marked by PUSH/POP macros + for i in xrange(linenum, 0, -1): + line = clean_lines.elided[i] + if Match(r'GOOGLE_ALLOW_RVALUE_REFERENCES_(?:PUSH|POP)', line): + if not line.endswith('PUSH'): + return False + for j in xrange(linenum, clean_lines.NumLines(), 1): + line = clean_lines.elided[j] + if Match(r'GOOGLE_ALLOW_RVALUE_REFERENCES_(?:PUSH|POP)', line): + return line.endswith('POP') + + # Allow operator= + line = clean_lines.elided[linenum] + if Search(r'\boperator\s*=\s*\(', line): + return IsDeletedOrDefault(clean_lines, linenum) + + # Allow constructors + match = Match(r'\s*(?:[\w<>]+::)*([\w<>]+)\s*::\s*([\w<>]+)\s*\(', line) + if match and match.group(1) == match.group(2): + return IsDeletedOrDefault(clean_lines, linenum) + if Search(r'\b(?:explicit|inline)\s+[\w<>]+\s*\(', line): + return IsDeletedOrDefault(clean_lines, linenum) + + if Match(r'\s*[\w<>]+\s*\(', line): + previous_line = 'ReturnType' + if linenum > 0: + previous_line = clean_lines.elided[linenum - 1] + if Match(r'^\s*$', previous_line) or Search(r'[{}:;]\s*$', previous_line): + return IsDeletedOrDefault(clean_lines, linenum) + + # Reject types not mentioned in template-argument-list + while line: + match = Match(r'^.*?(\w+)\s*&&(.*)$', line) + if not match: + break + if match.group(1) not in typenames: + return False + line = match.group(2) + + # All RValue types that were in template-argument-list should have + # been removed by now. Those were allowed, assuming that they will + # be forwarded. + # + # If there are no remaining RValue types left (i.e. types that were + # not found in template-argument-list), flag those as not allowed. + return line.find('&&') < 0 + + +def GetTemplateArgs(clean_lines, linenum): + """Find list of template arguments associated with this function declaration. + + Args: + clean_lines: A CleansedLines instance containing the file. + linenum: Line number containing the start of the function declaration, + usually one line after the end of the template-argument-list. + Returns: + Set of type names, or empty set if this does not appear to have + any template parameters. + """ + # Find start of function + func_line = linenum + while func_line > 0: + line = clean_lines.elided[func_line] + if Match(r'^\s*$', line): + return set() + if line.find('(') >= 0: + break + func_line -= 1 + if func_line == 0: + return set() + + # Collapse template-argument-list into a single string + argument_list = '' + match = Match(r'^(\s*template\s*)<', clean_lines.elided[func_line]) + if match: + # template-argument-list on the same line as function name + start_col = len(match.group(1)) + _, end_line, end_col = CloseExpression(clean_lines, func_line, start_col) + if end_col > -1 and end_line == func_line: + start_col += 1 # Skip the opening bracket + argument_list = clean_lines.elided[func_line][start_col:end_col] + + elif func_line > 1: + # template-argument-list one line before function name + match = Match(r'^(.*)>\s*$', clean_lines.elided[func_line - 1]) + if match: + end_col = len(match.group(1)) + _, start_line, start_col = ReverseCloseExpression( + clean_lines, func_line - 1, end_col) + if start_col > -1: + start_col += 1 # Skip the opening bracket + while start_line < func_line - 1: + argument_list += clean_lines.elided[start_line][start_col:] + start_col = 0 + start_line += 1 + argument_list += clean_lines.elided[func_line - 1][start_col:end_col] + + if not argument_list: + return set() + + # Extract type names + typenames = set() + while True: + match = Match(r'^[,\s]*(?:typename|class)(?:\.\.\.)?\s+(\w+)(.*)$', + argument_list) + if not match: + break + typenames.add(match.group(1)) + argument_list = match.group(2) + return typenames + + +def CheckRValueReference(filename, clean_lines, linenum, nesting_state, error): + """Check for rvalue references. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + nesting_state: A NestingState instance which maintains information about + the current stack of nested blocks being parsed. + error: The function to call with any errors found. + """ + # Find lines missing spaces around &&. + # TODO(unknown): currently we don't check for rvalue references + # with spaces surrounding the && to avoid false positives with + # boolean expressions. + line = clean_lines.elided[linenum] + match = Match(r'^(.*\S)&&', line) + if not match: + match = Match(r'(.*)&&\S', line) + if (not match) or '(&&)' in line or Search(r'\boperator\s*$', match.group(1)): + return + + # Either poorly formed && or an rvalue reference, check the context + # to get a more accurate error message. Mostly we want to determine + # if what's to the left of "&&" is a type or not. + typenames = GetTemplateArgs(clean_lines, linenum) + and_pos = len(match.group(1)) + if IsRValueType(typenames, clean_lines, nesting_state, linenum, and_pos): + if not IsRValueAllowed(clean_lines, linenum, typenames): + error(filename, linenum, 'build/c++11', 3, + 'RValue references are an unapproved C++ feature.') + else: + error(filename, linenum, 'whitespace/operators', 3, + 'Missing spaces around &&') + + +def CheckSectionSpacing(filename, clean_lines, class_info, linenum, error): + """Checks for additional blank line issues related to sections. + + Currently the only thing checked here is blank line before protected/private. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + class_info: A _ClassInfo objects. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + # Skip checks if the class is small, where small means 25 lines or less. + # 25 lines seems like a good cutoff since that's the usual height of + # terminals, and any class that can't fit in one screen can't really + # be considered "small". + # + # Also skip checks if we are on the first line. This accounts for + # classes that look like + # class Foo { public: ... }; + # + # If we didn't find the end of the class, last_line would be zero, + # and the check will be skipped by the first condition. + if (class_info.last_line - class_info.starting_linenum <= 24 or + linenum <= class_info.starting_linenum): + return + + matched = Match(r'\s*(public|protected|private):', clean_lines.lines[linenum]) + if matched: + # Issue warning if the line before public/protected/private was + # not a blank line, but don't do this if the previous line contains + # "class" or "struct". This can happen two ways: + # - We are at the beginning of the class. + # - We are forward-declaring an inner class that is semantically + # private, but needed to be public for implementation reasons. + # Also ignores cases where the previous line ends with a backslash as can be + # common when defining classes in C macros. + prev_line = clean_lines.lines[linenum - 1] + if (not IsBlankLine(prev_line) and + not Search(r'\b(class|struct)\b', prev_line) and + not Search(r'\\$', prev_line)): + # Try a bit harder to find the beginning of the class. This is to + # account for multi-line base-specifier lists, e.g.: + # class Derived + # : public Base { + end_class_head = class_info.starting_linenum + for i in range(class_info.starting_linenum, linenum): + if Search(r'\{\s*$', clean_lines.lines[i]): + end_class_head = i + break + if end_class_head < linenum - 1: + error(filename, linenum, 'whitespace/blank_line', 3, + '"%s:" should be preceded by a blank line' % matched.group(1)) + + +def GetPreviousNonBlankLine(clean_lines, linenum): + """Return the most recent non-blank line and its line number. + + Args: + clean_lines: A CleansedLines instance containing the file contents. + linenum: The number of the line to check. + + Returns: + A tuple with two elements. The first element is the contents of the last + non-blank line before the current line, or the empty string if this is the + first non-blank line. The second is the line number of that line, or -1 + if this is the first non-blank line. + """ + + prevlinenum = linenum - 1 + while prevlinenum >= 0: + prevline = clean_lines.elided[prevlinenum] + if not IsBlankLine(prevline): # if not a blank line... + return (prevline, prevlinenum) + prevlinenum -= 1 + return ('', -1) + + +def CheckBraces(filename, clean_lines, linenum, error): + """Looks for misplaced braces (e.g. at the end of line). + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + + line = clean_lines.elided[linenum] # get rid of comments and strings + + if Match(r'\s*{\s*$', line): + # We allow an open brace to start a line in the case where someone is using + # braces in a block to explicitly create a new scope, which is commonly used + # to control the lifetime of stack-allocated variables. Braces are also + # used for brace initializers inside function calls. We don't detect this + # perfectly: we just don't complain if the last non-whitespace character on + # the previous non-blank line is ',', ';', ':', '(', '{', or '}', or if the + # previous line starts a preprocessor block. + prevline = GetPreviousNonBlankLine(clean_lines, linenum)[0] + if (not Search(r'[,;:}{(]\s*$', prevline) and + not Match(r'\s*#', prevline)): + error(filename, linenum, 'whitespace/braces', 4, + '{ should almost always be at the end of the previous line') + + # An else clause should be on the same line as the preceding closing brace. + if Match(r'\s*else\b\s*(?:if\b|\{|$)', line): + prevline = GetPreviousNonBlankLine(clean_lines, linenum)[0] + if Match(r'\s*}\s*$', prevline): + error(filename, linenum, 'whitespace/newline', 4, + 'An else should appear on the same line as the preceding }') + + # If braces come on one side of an else, they should be on both. + # However, we have to worry about "else if" that spans multiple lines! + if Search(r'else if\s*\(', line): # could be multi-line if + brace_on_left = bool(Search(r'}\s*else if\s*\(', line)) + # find the ( after the if + pos = line.find('else if') + pos = line.find('(', pos) + if pos > 0: + (endline, _, endpos) = CloseExpression(clean_lines, linenum, pos) + brace_on_right = endline[endpos:].find('{') != -1 + if brace_on_left != brace_on_right: # must be brace after if + error(filename, linenum, 'readability/braces', 5, + 'If an else has a brace on one side, it should have it on both') + elif Search(r'}\s*else[^{]*$', line) or Match(r'[^}]*else\s*{', line): + error(filename, linenum, 'readability/braces', 5, + 'If an else has a brace on one side, it should have it on both') + + # Likewise, an else should never have the else clause on the same line + if Search(r'\belse [^\s{]', line) and not Search(r'\belse if\b', line): + error(filename, linenum, 'whitespace/newline', 4, + 'Else clause should never be on same line as else (use 2 lines)') + + # In the same way, a do/while should never be on one line + if Match(r'\s*do [^\s{]', line): + error(filename, linenum, 'whitespace/newline', 4, + 'do/while clauses should not be on a single line') + + # Check single-line if/else bodies. The style guide says 'curly braces are not + # required for single-line statements'. We additionally allow multi-line, + # single statements, but we reject anything with more than one semicolon in + # it. This means that the first semicolon after the if should be at the end of + # its line, and the line after that should have an indent level equal to or + # lower than the if. We also check for ambiguous if/else nesting without + # braces. + if_else_match = Search(r'\b(if\s*\(|else\b)', line) + if if_else_match and not Match(r'\s*#', line): + if_indent = GetIndentLevel(line) + endline, endlinenum, endpos = line, linenum, if_else_match.end() + if_match = Search(r'\bif\s*\(', line) + if if_match: + # This could be a multiline if condition, so find the end first. + pos = if_match.end() - 1 + (endline, endlinenum, endpos) = CloseExpression(clean_lines, linenum, pos) + # Check for an opening brace, either directly after the if or on the next + # line. If found, this isn't a single-statement conditional. + if (not Match(r'\s*{', endline[endpos:]) + and not (Match(r'\s*$', endline[endpos:]) + and endlinenum < (len(clean_lines.elided) - 1) + and Match(r'\s*{', clean_lines.elided[endlinenum + 1]))): + while (endlinenum < len(clean_lines.elided) + and ';' not in clean_lines.elided[endlinenum][endpos:]): + endlinenum += 1 + endpos = 0 + if endlinenum < len(clean_lines.elided): + endline = clean_lines.elided[endlinenum] + # We allow a mix of whitespace and closing braces (e.g. for one-liner + # methods) and a single \ after the semicolon (for macros) + endpos = endline.find(';') + if not Match(r';[\s}]*(\\?)$', endline[endpos:]): + # Semicolon isn't the last character, there's something trailing. + # Output a warning if the semicolon is not contained inside + # a lambda expression. + if not Match(r'^[^{};]*\[[^\[\]]*\][^{}]*\{[^{}]*\}\s*\)*[;,]\s*$', + endline): + error(filename, linenum, 'readability/braces', 4, + 'If/else bodies with multiple statements require braces') + elif endlinenum < len(clean_lines.elided) - 1: + # Make sure the next line is dedented + next_line = clean_lines.elided[endlinenum + 1] + next_indent = GetIndentLevel(next_line) + # With ambiguous nested if statements, this will error out on the + # if that *doesn't* match the else, regardless of whether it's the + # inner one or outer one. + if (if_match and Match(r'\s*else\b', next_line) + and next_indent != if_indent): + error(filename, linenum, 'readability/braces', 4, + 'Else clause should be indented at the same level as if. ' + 'Ambiguous nested if/else chains require braces.') + elif next_indent > if_indent: + error(filename, linenum, 'readability/braces', 4, + 'If/else bodies with multiple statements require braces') + + +def CheckTrailingSemicolon(filename, clean_lines, linenum, error): + """Looks for redundant trailing semicolon. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + + line = clean_lines.elided[linenum] + + # Block bodies should not be followed by a semicolon. Due to C++11 + # brace initialization, there are more places where semicolons are + # required than not, so we use a whitelist approach to check these + # rather than a blacklist. These are the places where "};" should + # be replaced by just "}": + # 1. Some flavor of block following closing parenthesis: + # for (;;) {}; + # while (...) {}; + # switch (...) {}; + # Function(...) {}; + # if (...) {}; + # if (...) else if (...) {}; + # + # 2. else block: + # if (...) else {}; + # + # 3. const member function: + # Function(...) const {}; + # + # 4. Block following some statement: + # x = 42; + # {}; + # + # 5. Block at the beginning of a function: + # Function(...) { + # {}; + # } + # + # Note that naively checking for the preceding "{" will also match + # braces inside multi-dimensional arrays, but this is fine since + # that expression will not contain semicolons. + # + # 6. Block following another block: + # while (true) {} + # {}; + # + # 7. End of namespaces: + # namespace {}; + # + # These semicolons seems far more common than other kinds of + # redundant semicolons, possibly due to people converting classes + # to namespaces. For now we do not warn for this case. + # + # Try matching case 1 first. + match = Match(r'^(.*\)\s*)\{', line) + if match: + # Matched closing parenthesis (case 1). Check the token before the + # matching opening parenthesis, and don't warn if it looks like a + # macro. This avoids these false positives: + # - macro that defines a base class + # - multi-line macro that defines a base class + # - macro that defines the whole class-head + # + # But we still issue warnings for macros that we know are safe to + # warn, specifically: + # - TEST, TEST_F, TEST_P, MATCHER, MATCHER_P + # - TYPED_TEST + # - INTERFACE_DEF + # - EXCLUSIVE_LOCKS_REQUIRED, SHARED_LOCKS_REQUIRED, LOCKS_EXCLUDED: + # + # We implement a whitelist of safe macros instead of a blacklist of + # unsafe macros, even though the latter appears less frequently in + # google code and would have been easier to implement. This is because + # the downside for getting the whitelist wrong means some extra + # semicolons, while the downside for getting the blacklist wrong + # would result in compile errors. + # + # In addition to macros, we also don't want to warn on + # - Compound literals + # - Lambdas + # - alignas specifier with anonymous structs: + closing_brace_pos = match.group(1).rfind(')') + opening_parenthesis = ReverseCloseExpression( + clean_lines, linenum, closing_brace_pos) + if opening_parenthesis[2] > -1: + line_prefix = opening_parenthesis[0][0:opening_parenthesis[2]] + macro = Search(r'\b([A-Z_]+)\s*$', line_prefix) + func = Match(r'^(.*\])\s*$', line_prefix) + if ((macro and + macro.group(1) not in ( + 'TEST', 'TEST_F', 'MATCHER', 'MATCHER_P', 'TYPED_TEST', + 'EXCLUSIVE_LOCKS_REQUIRED', 'SHARED_LOCKS_REQUIRED', + 'LOCKS_EXCLUDED', 'INTERFACE_DEF')) or + (func and not Search(r'\boperator\s*\[\s*\]', func.group(1))) or + Search(r'\b(?:struct|union)\s+alignas\s*$', line_prefix) or + Search(r'\s+=\s*$', line_prefix)): + match = None + if (match and + opening_parenthesis[1] > 1 and + Search(r'\]\s*$', clean_lines.elided[opening_parenthesis[1] - 1])): + # Multi-line lambda-expression + match = None + + else: + # Try matching cases 2-3. + match = Match(r'^(.*(?:else|\)\s*const)\s*)\{', line) + if not match: + # Try matching cases 4-6. These are always matched on separate lines. + # + # Note that we can't simply concatenate the previous line to the + # current line and do a single match, otherwise we may output + # duplicate warnings for the blank line case: + # if (cond) { + # // blank line + # } + prevline = GetPreviousNonBlankLine(clean_lines, linenum)[0] + if prevline and Search(r'[;{}]\s*$', prevline): + match = Match(r'^(\s*)\{', line) + + # Check matching closing brace + if match: + (endline, endlinenum, endpos) = CloseExpression( + clean_lines, linenum, len(match.group(1))) + if endpos > -1 and Match(r'^\s*;', endline[endpos:]): + # Current {} pair is eligible for semicolon check, and we have found + # the redundant semicolon, output warning here. + # + # Note: because we are scanning forward for opening braces, and + # outputting warnings for the matching closing brace, if there are + # nested blocks with trailing semicolons, we will get the error + # messages in reversed order. + error(filename, endlinenum, 'readability/braces', 4, + "You don't need a ; after a }") + + +def CheckEmptyBlockBody(filename, clean_lines, linenum, error): + """Look for empty loop/conditional body with only a single semicolon. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + + # Search for loop keywords at the beginning of the line. Because only + # whitespaces are allowed before the keywords, this will also ignore most + # do-while-loops, since those lines should start with closing brace. + # + # We also check "if" blocks here, since an empty conditional block + # is likely an error. + line = clean_lines.elided[linenum] + matched = Match(r'\s*(for|while|if)\s*\(', line) + if matched: + # Find the end of the conditional expression + (end_line, end_linenum, end_pos) = CloseExpression( + clean_lines, linenum, line.find('(')) + + # Output warning if what follows the condition expression is a semicolon. + # No warning for all other cases, including whitespace or newline, since we + # have a separate check for semicolons preceded by whitespace. + if end_pos >= 0 and Match(r';', end_line[end_pos:]): + if matched.group(1) == 'if': + error(filename, end_linenum, 'whitespace/empty_conditional_body', 5, + 'Empty conditional bodies should use {}') + else: + error(filename, end_linenum, 'whitespace/empty_loop_body', 5, + 'Empty loop bodies should use {} or continue') + + +def FindCheckMacro(line): + """Find a replaceable CHECK-like macro. + + Args: + line: line to search on. + Returns: + (macro name, start position), or (None, -1) if no replaceable + macro is found. + """ + for macro in _CHECK_MACROS: + i = line.find(macro) + if i >= 0: + # Find opening parenthesis. Do a regular expression match here + # to make sure that we are matching the expected CHECK macro, as + # opposed to some other macro that happens to contain the CHECK + # substring. + matched = Match(r'^(.*\b' + macro + r'\s*)\(', line) + if not matched: + continue + return (macro, len(matched.group(1))) + return (None, -1) + + +def CheckCheck(filename, clean_lines, linenum, error): + """Checks the use of CHECK and EXPECT macros. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + + # Decide the set of replacement macros that should be suggested + lines = clean_lines.elided + (check_macro, start_pos) = FindCheckMacro(lines[linenum]) + if not check_macro: + return + + # Find end of the boolean expression by matching parentheses + (last_line, end_line, end_pos) = CloseExpression( + clean_lines, linenum, start_pos) + if end_pos < 0: + return + + # If the check macro is followed by something other than a + # semicolon, assume users will log their own custom error messages + # and don't suggest any replacements. + if not Match(r'\s*;', last_line[end_pos:]): + return + + if linenum == end_line: + expression = lines[linenum][start_pos + 1:end_pos - 1] + else: + expression = lines[linenum][start_pos + 1:] + for i in xrange(linenum + 1, end_line): + expression += lines[i] + expression += last_line[0:end_pos - 1] + + # Parse expression so that we can take parentheses into account. + # This avoids false positives for inputs like "CHECK((a < 4) == b)", + # which is not replaceable by CHECK_LE. + lhs = '' + rhs = '' + operator = None + while expression: + matched = Match(r'^\s*(<<|<<=|>>|>>=|->\*|->|&&|\|\||' + r'==|!=|>=|>|<=|<|\()(.*)$', expression) + if matched: + token = matched.group(1) + if token == '(': + # Parenthesized operand + expression = matched.group(2) + (end, _) = FindEndOfExpressionInLine(expression, 0, ['(']) + if end < 0: + return # Unmatched parenthesis + lhs += '(' + expression[0:end] + expression = expression[end:] + elif token in ('&&', '||'): + # Logical and/or operators. This means the expression + # contains more than one term, for example: + # CHECK(42 < a && a < b); + # + # These are not replaceable with CHECK_LE, so bail out early. + return + elif token in ('<<', '<<=', '>>', '>>=', '->*', '->'): + # Non-relational operator + lhs += token + expression = matched.group(2) + else: + # Relational operator + operator = token + rhs = matched.group(2) + break + else: + # Unparenthesized operand. Instead of appending to lhs one character + # at a time, we do another regular expression match to consume several + # characters at once if possible. Trivial benchmark shows that this + # is more efficient when the operands are longer than a single + # character, which is generally the case. + matched = Match(r'^([^-=!<>()&|]+)(.*)$', expression) + if not matched: + matched = Match(r'^(\s*\S)(.*)$', expression) + if not matched: + break + lhs += matched.group(1) + expression = matched.group(2) + + # Only apply checks if we got all parts of the boolean expression + if not (lhs and operator and rhs): + return + + # Check that rhs do not contain logical operators. We already know + # that lhs is fine since the loop above parses out && and ||. + if rhs.find('&&') > -1 or rhs.find('||') > -1: + return + + # At least one of the operands must be a constant literal. This is + # to avoid suggesting replacements for unprintable things like + # CHECK(variable != iterator) + # + # The following pattern matches decimal, hex integers, strings, and + # characters (in that order). + lhs = lhs.strip() + rhs = rhs.strip() + match_constant = r'^([-+]?(\d+|0[xX][0-9a-fA-F]+)[lLuU]{0,3}|".*"|\'.*\')$' + if Match(match_constant, lhs) or Match(match_constant, rhs): + # Note: since we know both lhs and rhs, we can provide a more + # descriptive error message like: + # Consider using CHECK_EQ(x, 42) instead of CHECK(x == 42) + # Instead of: + # Consider using CHECK_EQ instead of CHECK(a == b) + # + # We are still keeping the less descriptive message because if lhs + # or rhs gets long, the error message might become unreadable. + error(filename, linenum, 'readability/check', 2, + 'Consider using %s instead of %s(a %s b)' % ( + _CHECK_REPLACEMENT[check_macro][operator], + check_macro, operator)) + + +def CheckAltTokens(filename, clean_lines, linenum, error): + """Check alternative keywords being used in boolean expressions. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + line = clean_lines.elided[linenum] + + # Avoid preprocessor lines + if Match(r'^\s*#', line): + return + + # Last ditch effort to avoid multi-line comments. This will not help + # if the comment started before the current line or ended after the + # current line, but it catches most of the false positives. At least, + # it provides a way to workaround this warning for people who use + # multi-line comments in preprocessor macros. + # + # TODO(unknown): remove this once cpplint has better support for + # multi-line comments. + if line.find('/*') >= 0 or line.find('*/') >= 0: + return + + for match in _ALT_TOKEN_REPLACEMENT_PATTERN.finditer(line): + error(filename, linenum, 'readability/alt_tokens', 2, + 'Use operator %s instead of %s' % ( + _ALT_TOKEN_REPLACEMENT[match.group(1)], match.group(1))) + + +def GetLineWidth(line): + """Determines the width of the line in column positions. + + Args: + line: A string, which may be a Unicode string. + + Returns: + The width of the line in column positions, accounting for Unicode + combining characters and wide characters. + """ + if isinstance(line, unicode): + width = 0 + for uc in unicodedata.normalize('NFC', line): + if unicodedata.east_asian_width(uc) in ('W', 'F'): + width += 2 + elif not unicodedata.combining(uc): + width += 1 + return width + else: + return len(line) + + +def CheckStyle(filename, clean_lines, linenum, file_extension, nesting_state, + error): + """Checks rules from the 'C++ style rules' section of cppguide.html. + + Most of these rules are hard to test (naming, comment style), but we + do what we can. In particular we check for 2-space indents, line lengths, + tab usage, spaces inside code, etc. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + file_extension: The extension (without the dot) of the filename. + nesting_state: A NestingState instance which maintains information about + the current stack of nested blocks being parsed. + error: The function to call with any errors found. + """ + + # Don't use "elided" lines here, otherwise we can't check commented lines. + # Don't want to use "raw" either, because we don't want to check inside C++11 + # raw strings, + raw_lines = clean_lines.lines_without_raw_strings + line = raw_lines[linenum] + + if line.find('\t') != -1: + error(filename, linenum, 'whitespace/tab', 1, + 'Tab found; better to use spaces') + + # One or three blank spaces at the beginning of the line is weird; it's + # hard to reconcile that with 2-space indents. + # NOTE: here are the conditions rob pike used for his tests. Mine aren't + # as sophisticated, but it may be worth becoming so: RLENGTH==initial_spaces + # if(RLENGTH > 20) complain = 0; + # if(match($0, " +(error|private|public|protected):")) complain = 0; + # if(match(prev, "&& *$")) complain = 0; + # if(match(prev, "\\|\\| *$")) complain = 0; + # if(match(prev, "[\",=><] *$")) complain = 0; + # if(match($0, " <<")) complain = 0; + # if(match(prev, " +for \\(")) complain = 0; + # if(prevodd && match(prevprev, " +for \\(")) complain = 0; + scope_or_label_pattern = r'\s*\w+\s*:\s*\\?$' + classinfo = nesting_state.InnermostClass() + initial_spaces = 0 + cleansed_line = clean_lines.elided[linenum] + while initial_spaces < len(line) and line[initial_spaces] == ' ': + initial_spaces += 1 + if line and line[-1].isspace(): + error(filename, linenum, 'whitespace/end_of_line', 4, + 'Line ends in whitespace. Consider deleting these extra spaces.') + # There are certain situations we allow one space, notably for + # section labels, and also lines containing multi-line raw strings. + elif ((initial_spaces == 1 or initial_spaces == 3) and + not Match(scope_or_label_pattern, cleansed_line) and + not (clean_lines.raw_lines[linenum] != line and + Match(r'^\s*""', line))): + error(filename, linenum, 'whitespace/indent', 3, + 'Weird number of spaces at line-start. ' + 'Are you using a 2-space indent?') + + # Check if the line is a header guard. + is_header_guard = False + if file_extension == 'h': + cppvar = GetHeaderGuardCPPVariable(filename) + if (line.startswith('#ifndef %s' % cppvar) or + line.startswith('#define %s' % cppvar) or + line.startswith('#endif // %s' % cppvar)): + is_header_guard = True + # #include lines and header guards can be long, since there's no clean way to + # split them. + # + # URLs can be long too. It's possible to split these, but it makes them + # harder to cut&paste. + # + # The "$Id:...$" comment may also get very long without it being the + # developers fault. + if (not line.startswith('#include') and not is_header_guard and + not Match(r'^\s*//.*http(s?)://\S*$', line) and + not Match(r'^// \$Id:.*#[0-9]+ \$$', line)): + line_width = GetLineWidth(line) + extended_length = int((_line_length * 1.25)) + if line_width > extended_length: + error(filename, linenum, 'whitespace/line_length', 4, + 'Lines should very rarely be longer than %i characters' % + extended_length) + elif line_width > _line_length: + error(filename, linenum, 'whitespace/line_length', 2, + 'Lines should be <= %i characters long' % _line_length) + + if (cleansed_line.count(';') > 1 and + # for loops are allowed two ;'s (and may run over two lines). + cleansed_line.find('for') == -1 and + (GetPreviousNonBlankLine(clean_lines, linenum)[0].find('for') == -1 or + GetPreviousNonBlankLine(clean_lines, linenum)[0].find(';') != -1) and + # It's ok to have many commands in a switch case that fits in 1 line + not ((cleansed_line.find('case ') != -1 or + cleansed_line.find('default:') != -1) and + cleansed_line.find('break;') != -1)): + error(filename, linenum, 'whitespace/newline', 0, + 'More than one command on the same line') + + # Some more style checks + CheckBraces(filename, clean_lines, linenum, error) + CheckTrailingSemicolon(filename, clean_lines, linenum, error) + CheckEmptyBlockBody(filename, clean_lines, linenum, error) + CheckAccess(filename, clean_lines, linenum, nesting_state, error) + CheckSpacing(filename, clean_lines, linenum, nesting_state, error) + CheckOperatorSpacing(filename, clean_lines, linenum, error) + CheckParenthesisSpacing(filename, clean_lines, linenum, error) + CheckCommaSpacing(filename, clean_lines, linenum, error) + CheckBracesSpacing(filename, clean_lines, linenum, error) + CheckSpacingForFunctionCall(filename, clean_lines, linenum, error) + CheckRValueReference(filename, clean_lines, linenum, nesting_state, error) + CheckCheck(filename, clean_lines, linenum, error) + CheckAltTokens(filename, clean_lines, linenum, error) + classinfo = nesting_state.InnermostClass() + if classinfo: + CheckSectionSpacing(filename, clean_lines, classinfo, linenum, error) + + +_RE_PATTERN_INCLUDE = re.compile(r'^\s*#\s*include\s*([<"])([^>"]*)[>"].*$') +# Matches the first component of a filename delimited by -s and _s. That is: +# _RE_FIRST_COMPONENT.match('foo').group(0) == 'foo' +# _RE_FIRST_COMPONENT.match('foo.cc').group(0) == 'foo' +# _RE_FIRST_COMPONENT.match('foo-bar_baz.cc').group(0) == 'foo' +# _RE_FIRST_COMPONENT.match('foo_bar-baz.cc').group(0) == 'foo' +_RE_FIRST_COMPONENT = re.compile(r'^[^-_.]+') + + +def _DropCommonSuffixes(filename): + """Drops common suffixes like _test.cc or -inl.h from filename. + + For example: + >>> _DropCommonSuffixes('foo/foo-inl.h') + 'foo/foo' + >>> _DropCommonSuffixes('foo/bar/foo.cc') + 'foo/bar/foo' + >>> _DropCommonSuffixes('foo/foo_internal.h') + 'foo/foo' + >>> _DropCommonSuffixes('foo/foo_unusualinternal.h') + 'foo/foo_unusualinternal' + + Args: + filename: The input filename. + + Returns: + The filename with the common suffix removed. + """ + for suffix in ('test.cc', 'regtest.cc', 'unittest.cc', + 'inl.h', 'impl.h', 'internal.h'): + if (filename.endswith(suffix) and len(filename) > len(suffix) and + filename[-len(suffix) - 1] in ('-', '_')): + return filename[:-len(suffix) - 1] + return os.path.splitext(filename)[0] + + +def _IsTestFilename(filename): + """Determines if the given filename has a suffix that identifies it as a test. + + Args: + filename: The input filename. + + Returns: + True if 'filename' looks like a test, False otherwise. + """ + if (filename.endswith('_test.cc') or + filename.endswith('_unittest.cc') or + filename.endswith('_regtest.cc')): + return True + else: + return False + + +def _ClassifyInclude(fileinfo, include, is_system): + """Figures out what kind of header 'include' is. + + Args: + fileinfo: The current file cpplint is running over. A FileInfo instance. + include: The path to a #included file. + is_system: True if the #include used <> rather than "". + + Returns: + One of the _XXX_HEADER constants. + + For example: + >>> _ClassifyInclude(FileInfo('foo/foo.cc'), 'stdio.h', True) + _C_SYS_HEADER + >>> _ClassifyInclude(FileInfo('foo/foo.cc'), 'string', True) + _CPP_SYS_HEADER + >>> _ClassifyInclude(FileInfo('foo/foo.cc'), 'foo/foo.h', False) + _LIKELY_MY_HEADER + >>> _ClassifyInclude(FileInfo('foo/foo_unknown_extension.cc'), + ... 'bar/foo_other_ext.h', False) + _POSSIBLE_MY_HEADER + >>> _ClassifyInclude(FileInfo('foo/foo.cc'), 'foo/bar.h', False) + _OTHER_HEADER + """ + # This is a list of all standard c++ header files, except + # those already checked for above. + is_cpp_h = include in _CPP_HEADERS + + if is_system: + if is_cpp_h: + return _CPP_SYS_HEADER + else: + return _C_SYS_HEADER + + # If the target file and the include we're checking share a + # basename when we drop common extensions, and the include + # lives in . , then it's likely to be owned by the target file. + target_dir, target_base = ( + os.path.split(_DropCommonSuffixes(fileinfo.RepositoryName()))) + include_dir, include_base = os.path.split(_DropCommonSuffixes(include)) + if target_base == include_base and ( + include_dir == target_dir or + include_dir == os.path.normpath(target_dir + '/../public')): + return _LIKELY_MY_HEADER + + # If the target and include share some initial basename + # component, it's possible the target is implementing the + # include, so it's allowed to be first, but we'll never + # complain if it's not there. + target_first_component = _RE_FIRST_COMPONENT.match(target_base) + include_first_component = _RE_FIRST_COMPONENT.match(include_base) + if (target_first_component and include_first_component and + target_first_component.group(0) == + include_first_component.group(0)): + return _POSSIBLE_MY_HEADER + + return _OTHER_HEADER + + + +def CheckIncludeLine(filename, clean_lines, linenum, include_state, error): + """Check rules that are applicable to #include lines. + + Strings on #include lines are NOT removed from elided line, to make + certain tasks easier. However, to prevent false positives, checks + applicable to #include lines in CheckLanguage must be put here. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + include_state: An _IncludeState instance in which the headers are inserted. + error: The function to call with any errors found. + """ + fileinfo = FileInfo(filename) + line = clean_lines.lines[linenum] + + # "include" should use the new style "foo/bar.h" instead of just "bar.h" + # Only do this check if the included header follows google naming + # conventions. If not, assume that it's a 3rd party API that + # requires special include conventions. + # + # We also make an exception for Lua headers, which follow google + # naming convention but not the include convention. + match = Match(r'#include\s*"([^/]+\.h)"', line) + if match and not _THIRD_PARTY_HEADERS_PATTERN.match(match.group(1)): + error(filename, linenum, 'build/include', 4, + 'Include the directory when naming .h files') + + # we shouldn't include a file more than once. actually, there are a + # handful of instances where doing so is okay, but in general it's + # not. + match = _RE_PATTERN_INCLUDE.search(line) + if match: + include = match.group(2) + is_system = (match.group(1) == '<') + duplicate_line = include_state.FindHeader(include) + if duplicate_line >= 0: + error(filename, linenum, 'build/include', 4, + '"%s" already included at %s:%s' % + (include, filename, duplicate_line)) + elif (include.endswith('.cc') and + os.path.dirname(fileinfo.RepositoryName()) != os.path.dirname(include)): + error(filename, linenum, 'build/include', 4, + 'Do not include .cc files from other packages') + elif not _THIRD_PARTY_HEADERS_PATTERN.match(include): + include_state.include_list[-1].append((include, linenum)) + + # We want to ensure that headers appear in the right order: + # 1) for foo.cc, foo.h (preferred location) + # 2) c system files + # 3) cpp system files + # 4) for foo.cc, foo.h (deprecated location) + # 5) other google headers + # + # We classify each include statement as one of those 5 types + # using a number of techniques. The include_state object keeps + # track of the highest type seen, and complains if we see a + # lower type after that. + error_message = include_state.CheckNextIncludeOrder( + _ClassifyInclude(fileinfo, include, is_system)) + if error_message: + error(filename, linenum, 'build/include_order', 4, + '%s. Should be: %s.h, c system, c++ system, other.' % + (error_message, fileinfo.BaseName())) + canonical_include = include_state.CanonicalizeAlphabeticalOrder(include) + if not include_state.IsInAlphabeticalOrder( + clean_lines, linenum, canonical_include): + error(filename, linenum, 'build/include_alpha', 4, + 'Include "%s" not in alphabetical order' % include) + include_state.SetLastHeader(canonical_include) + + + +def _GetTextInside(text, start_pattern): + r"""Retrieves all the text between matching open and close parentheses. + + Given a string of lines and a regular expression string, retrieve all the text + following the expression and between opening punctuation symbols like + (, [, or {, and the matching close-punctuation symbol. This properly nested + occurrences of the punctuations, so for the text like + printf(a(), b(c())); + a call to _GetTextInside(text, r'printf\(') will return 'a(), b(c())'. + start_pattern must match string having an open punctuation symbol at the end. + + Args: + text: The lines to extract text. Its comments and strings must be elided. + It can be single line and can span multiple lines. + start_pattern: The regexp string indicating where to start extracting + the text. + Returns: + The extracted text. + None if either the opening string or ending punctuation could not be found. + """ + # TODO(unknown): Audit cpplint.py to see what places could be profitably + # rewritten to use _GetTextInside (and use inferior regexp matching today). + + # Give opening punctuations to get the matching close-punctuations. + matching_punctuation = {'(': ')', '{': '}', '[': ']'} + closing_punctuation = set(matching_punctuation.itervalues()) + + # Find the position to start extracting text. + match = re.search(start_pattern, text, re.M) + if not match: # start_pattern not found in text. + return None + start_position = match.end(0) + + assert start_position > 0, ( + 'start_pattern must ends with an opening punctuation.') + assert text[start_position - 1] in matching_punctuation, ( + 'start_pattern must ends with an opening punctuation.') + # Stack of closing punctuations we expect to have in text after position. + punctuation_stack = [matching_punctuation[text[start_position - 1]]] + position = start_position + while punctuation_stack and position < len(text): + if text[position] == punctuation_stack[-1]: + punctuation_stack.pop() + elif text[position] in closing_punctuation: + # A closing punctuation without matching opening punctuations. + return None + elif text[position] in matching_punctuation: + punctuation_stack.append(matching_punctuation[text[position]]) + position += 1 + if punctuation_stack: + # Opening punctuations left without matching close-punctuations. + return None + # punctuations match. + return text[start_position:position - 1] + + +# Patterns for matching call-by-reference parameters. +# +# Supports nested templates up to 2 levels deep using this messy pattern: +# < (?: < (?: < [^<>]* +# > +# | [^<>] )* +# > +# | [^<>] )* +# > +_RE_PATTERN_IDENT = r'[_a-zA-Z]\w*' # =~ [[:alpha:]][[:alnum:]]* +_RE_PATTERN_TYPE = ( + r'(?:const\s+)?(?:typename\s+|class\s+|struct\s+|union\s+|enum\s+)?' + r'(?:\w|' + r'\s*<(?:<(?:<[^<>]*>|[^<>])*>|[^<>])*>|' + r'::)+') +# A call-by-reference parameter ends with '& identifier'. +_RE_PATTERN_REF_PARAM = re.compile( + r'(' + _RE_PATTERN_TYPE + r'(?:\s*(?:\bconst\b|[*]))*\s*' + r'&\s*' + _RE_PATTERN_IDENT + r')\s*(?:=[^,()]+)?[,)]') +# A call-by-const-reference parameter either ends with 'const& identifier' +# or looks like 'const type& identifier' when 'type' is atomic. +_RE_PATTERN_CONST_REF_PARAM = ( + r'(?:.*\s*\bconst\s*&\s*' + _RE_PATTERN_IDENT + + r'|const\s+' + _RE_PATTERN_TYPE + r'\s*&\s*' + _RE_PATTERN_IDENT + r')') + + +def CheckLanguage(filename, clean_lines, linenum, file_extension, + include_state, nesting_state, error): + """Checks rules from the 'C++ language rules' section of cppguide.html. + + Some of these rules are hard to test (function overloading, using + uint32 inappropriately), but we do the best we can. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + file_extension: The extension (without the dot) of the filename. + include_state: An _IncludeState instance in which the headers are inserted. + nesting_state: A NestingState instance which maintains information about + the current stack of nested blocks being parsed. + error: The function to call with any errors found. + """ + # If the line is empty or consists of entirely a comment, no need to + # check it. + line = clean_lines.elided[linenum] + if not line: + return + + match = _RE_PATTERN_INCLUDE.search(line) + if match: + CheckIncludeLine(filename, clean_lines, linenum, include_state, error) + return + + # Reset include state across preprocessor directives. This is meant + # to silence warnings for conditional includes. + match = Match(r'^\s*#\s*(if|ifdef|ifndef|elif|else|endif)\b', line) + if match: + include_state.ResetSection(match.group(1)) + + # Make Windows paths like Unix. + fullname = os.path.abspath(filename).replace('\\', '/') + + # Perform other checks now that we are sure that this is not an include line + CheckCasts(filename, clean_lines, linenum, error) + CheckGlobalStatic(filename, clean_lines, linenum, error) + CheckPrintf(filename, clean_lines, linenum, error) + + if file_extension == 'h': + # TODO(unknown): check that 1-arg constructors are explicit. + # How to tell it's a constructor? + # (handled in CheckForNonStandardConstructs for now) + # TODO(unknown): check that classes declare or disable copy/assign + # (level 1 error) + pass + + # Check if people are using the verboten C basic types. The only exception + # we regularly allow is "unsigned short port" for port. + if Search(r'\bshort port\b', line): + if not Search(r'\bunsigned short port\b', line): + error(filename, linenum, 'runtime/int', 4, + 'Use "unsigned short" for ports, not "short"') + else: + match = Search(r'\b(short|long(?! +double)|long long)\b', line) + if match: + error(filename, linenum, 'runtime/int', 4, + 'Use int16/int64/etc, rather than the C type %s' % match.group(1)) + + # Check if some verboten operator overloading is going on + # TODO(unknown): catch out-of-line unary operator&: + # class X {}; + # int operator&(const X& x) { return 42; } // unary operator& + # The trick is it's hard to tell apart from binary operator&: + # class Y { int operator&(const Y& x) { return 23; } }; // binary operator& + if Search(r'\boperator\s*&\s*\(\s*\)', line): + error(filename, linenum, 'runtime/operator', 4, + 'Unary operator& is dangerous. Do not use it.') + + # Check for suspicious usage of "if" like + # } if (a == b) { + if Search(r'\}\s*if\s*\(', line): + error(filename, linenum, 'readability/braces', 4, + 'Did you mean "else if"? If not, start a new line for "if".') + + # Check for potential format string bugs like printf(foo). + # We constrain the pattern not to pick things like DocidForPrintf(foo). + # Not perfect but it can catch printf(foo.c_str()) and printf(foo->c_str()) + # TODO(unknown): Catch the following case. Need to change the calling + # convention of the whole function to process multiple line to handle it. + # printf( + # boy_this_is_a_really_long_variable_that_cannot_fit_on_the_prev_line); + printf_args = _GetTextInside(line, r'(?i)\b(string)?printf\s*\(') + if printf_args: + match = Match(r'([\w.\->()]+)$', printf_args) + if match and match.group(1) != '__VA_ARGS__': + function_name = re.search(r'\b((?:string)?printf)\s*\(', + line, re.I).group(1) + error(filename, linenum, 'runtime/printf', 4, + 'Potential format string bug. Do %s("%%s", %s) instead.' + % (function_name, match.group(1))) + + # Check for potential memset bugs like memset(buf, sizeof(buf), 0). + match = Search(r'memset\s*\(([^,]*),\s*([^,]*),\s*0\s*\)', line) + if match and not Match(r"^''|-?[0-9]+|0x[0-9A-Fa-f]$", match.group(2)): + error(filename, linenum, 'runtime/memset', 4, + 'Did you mean "memset(%s, 0, %s)"?' + % (match.group(1), match.group(2))) + + if Search(r'\busing namespace\b', line): + error(filename, linenum, 'build/namespaces', 5, + 'Do not use namespace using-directives. ' + 'Use using-declarations instead.') + + # Detect variable-length arrays. + match = Match(r'\s*(.+::)?(\w+) [a-z]\w*\[(.+)];', line) + if (match and match.group(2) != 'return' and match.group(2) != 'delete' and + match.group(3).find(']') == -1): + # Split the size using space and arithmetic operators as delimiters. + # If any of the resulting tokens are not compile time constants then + # report the error. + tokens = re.split(r'\s|\+|\-|\*|\/|<<|>>]', match.group(3)) + is_const = True + skip_next = False + for tok in tokens: + if skip_next: + skip_next = False + continue + + if Search(r'sizeof\(.+\)', tok): continue + if Search(r'arraysize\(\w+\)', tok): continue + + tok = tok.lstrip('(') + tok = tok.rstrip(')') + if not tok: continue + if Match(r'\d+', tok): continue + if Match(r'0[xX][0-9a-fA-F]+', tok): continue + if Match(r'k[A-Z0-9]\w*', tok): continue + if Match(r'(.+::)?k[A-Z0-9]\w*', tok): continue + if Match(r'(.+::)?[A-Z][A-Z0-9_]*', tok): continue + # A catch all for tricky sizeof cases, including 'sizeof expression', + # 'sizeof(*type)', 'sizeof(const type)', 'sizeof(struct StructName)' + # requires skipping the next token because we split on ' ' and '*'. + if tok.startswith('sizeof'): + skip_next = True + continue + is_const = False + break + if not is_const: + error(filename, linenum, 'runtime/arrays', 1, + 'Do not use variable-length arrays. Use an appropriately named ' + "('k' followed by CamelCase) compile-time constant for the size.") + + # Check for use of unnamed namespaces in header files. Registration + # macros are typically OK, so we allow use of "namespace {" on lines + # that end with backslashes. + if (file_extension == 'h' + and Search(r'\bnamespace\s*{', line) + and line[-1] != '\\'): + error(filename, linenum, 'build/namespaces', 4, + 'Do not use unnamed namespaces in header files. See ' + 'http://google-styleguide.googlecode.com/svn/trunk/cppguide.xml#Namespaces' + ' for more information.') + + +def CheckGlobalStatic(filename, clean_lines, linenum, error): + """Check for unsafe global or static objects. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + line = clean_lines.elided[linenum] + + # Match two lines at a time to support multiline declarations + if linenum + 1 < clean_lines.NumLines() and not Search(r'[;({]', line): + line += clean_lines.elided[linenum + 1].strip() + + # Check for people declaring static/global STL strings at the top level. + # This is dangerous because the C++ language does not guarantee that + # globals with constructors are initialized before the first access. + match = Match( + r'((?:|static +)(?:|const +))string +([a-zA-Z0-9_:]+)\b(.*)', + line) + + # Remove false positives: + # - String pointers (as opposed to values). + # string *pointer + # const string *pointer + # string const *pointer + # string *const pointer + # + # - Functions and template specializations. + # string Function(... + # string Class::Method(... + # + # - Operators. These are matched separately because operator names + # cross non-word boundaries, and trying to match both operators + # and functions at the same time would decrease accuracy of + # matching identifiers. + # string Class::operator*() + if (match and + not Search(r'\bstring\b(\s+const)?\s*\*\s*(const\s+)?\w', line) and + not Search(r'\boperator\W', line) and + not Match(r'\s*(<.*>)?(::[a-zA-Z0-9_]+)*\s*\(([^"]|$)', match.group(3))): + error(filename, linenum, 'runtime/string', 4, + 'For a static/global string constant, use a C style string instead: ' + '"%schar %s[]".' % + (match.group(1), match.group(2))) + + if Search(r'\b([A-Za-z0-9_]*_)\(\1\)', line): + error(filename, linenum, 'runtime/init', 4, + 'You seem to be initializing a member variable with itself.') + + +def CheckPrintf(filename, clean_lines, linenum, error): + """Check for printf related issues. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + line = clean_lines.elided[linenum] + + # When snprintf is used, the second argument shouldn't be a literal. + match = Search(r'snprintf\s*\(([^,]*),\s*([0-9]*)\s*,', line) + if match and match.group(2) != '0': + # If 2nd arg is zero, snprintf is used to calculate size. + error(filename, linenum, 'runtime/printf', 3, + 'If you can, use sizeof(%s) instead of %s as the 2nd arg ' + 'to snprintf.' % (match.group(1), match.group(2))) + + # Check if some verboten C functions are being used. + if Search(r'\bsprintf\s*\(', line): + error(filename, linenum, 'runtime/printf', 5, + 'Never use sprintf. Use snprintf instead.') + match = Search(r'\b(strcpy|strcat)\s*\(', line) + if match: + error(filename, linenum, 'runtime/printf', 4, + 'Almost always, snprintf is better than %s' % match.group(1)) + + +def IsDerivedFunction(clean_lines, linenum): + """Check if current line contains an inherited function. + + Args: + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + Returns: + True if current line contains a function with "override" + virt-specifier. + """ + # Scan back a few lines for start of current function + for i in xrange(linenum, max(-1, linenum - 10), -1): + match = Match(r'^([^()]*\w+)\(', clean_lines.elided[i]) + if match: + # Look for "override" after the matching closing parenthesis + line, _, closing_paren = CloseExpression( + clean_lines, i, len(match.group(1))) + return (closing_paren >= 0 and + Search(r'\boverride\b', line[closing_paren:])) + return False + + +def IsOutOfLineMethodDefinition(clean_lines, linenum): + """Check if current line contains an out-of-line method definition. + + Args: + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + Returns: + True if current line contains an out-of-line method definition. + """ + # Scan back a few lines for start of current function + for i in xrange(linenum, max(-1, linenum - 10), -1): + if Match(r'^([^()]*\w+)\(', clean_lines.elided[i]): + return Match(r'^[^()]*\w+::\w+\(', clean_lines.elided[i]) is not None + return False + + +def IsInitializerList(clean_lines, linenum): + """Check if current line is inside constructor initializer list. + + Args: + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + Returns: + True if current line appears to be inside constructor initializer + list, False otherwise. + """ + for i in xrange(linenum, 1, -1): + line = clean_lines.elided[i] + if i == linenum: + remove_function_body = Match(r'^(.*)\{\s*$', line) + if remove_function_body: + line = remove_function_body.group(1) + + if Search(r'\s:\s*\w+[({]', line): + # A lone colon tend to indicate the start of a constructor + # initializer list. It could also be a ternary operator, which + # also tend to appear in constructor initializer lists as + # opposed to parameter lists. + return True + if Search(r'\}\s*,\s*$', line): + # A closing brace followed by a comma is probably the end of a + # brace-initialized member in constructor initializer list. + return True + if Search(r'[{};]\s*$', line): + # Found one of the following: + # - A closing brace or semicolon, probably the end of the previous + # function. + # - An opening brace, probably the start of current class or namespace. + # + # Current line is probably not inside an initializer list since + # we saw one of those things without seeing the starting colon. + return False + + # Got to the beginning of the file without seeing the start of + # constructor initializer list. + return False + + +def CheckForNonConstReference(filename, clean_lines, linenum, + nesting_state, error): + """Check for non-const references. + + Separate from CheckLanguage since it scans backwards from current + line, instead of scanning forward. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + nesting_state: A NestingState instance which maintains information about + the current stack of nested blocks being parsed. + error: The function to call with any errors found. + """ + # Do nothing if there is no '&' on current line. + line = clean_lines.elided[linenum] + if '&' not in line: + return + + # If a function is inherited, current function doesn't have much of + # a choice, so any non-const references should not be blamed on + # derived function. + if IsDerivedFunction(clean_lines, linenum): + return + + # Don't warn on out-of-line method definitions, as we would warn on the + # in-line declaration, if it isn't marked with 'override'. + if IsOutOfLineMethodDefinition(clean_lines, linenum): + return + + # Long type names may be broken across multiple lines, usually in one + # of these forms: + # LongType + # ::LongTypeContinued &identifier + # LongType:: + # LongTypeContinued &identifier + # LongType< + # ...>::LongTypeContinued &identifier + # + # If we detected a type split across two lines, join the previous + # line to current line so that we can match const references + # accordingly. + # + # Note that this only scans back one line, since scanning back + # arbitrary number of lines would be expensive. If you have a type + # that spans more than 2 lines, please use a typedef. + if linenum > 1: + previous = None + if Match(r'\s*::(?:[\w<>]|::)+\s*&\s*\S', line): + # previous_line\n + ::current_line + previous = Search(r'\b((?:const\s*)?(?:[\w<>]|::)+[\w<>])\s*$', + clean_lines.elided[linenum - 1]) + elif Match(r'\s*[a-zA-Z_]([\w<>]|::)+\s*&\s*\S', line): + # previous_line::\n + current_line + previous = Search(r'\b((?:const\s*)?(?:[\w<>]|::)+::)\s*$', + clean_lines.elided[linenum - 1]) + if previous: + line = previous.group(1) + line.lstrip() + else: + # Check for templated parameter that is split across multiple lines + endpos = line.rfind('>') + if endpos > -1: + (_, startline, startpos) = ReverseCloseExpression( + clean_lines, linenum, endpos) + if startpos > -1 and startline < linenum: + # Found the matching < on an earlier line, collect all + # pieces up to current line. + line = '' + for i in xrange(startline, linenum + 1): + line += clean_lines.elided[i].strip() + + # Check for non-const references in function parameters. A single '&' may + # found in the following places: + # inside expression: binary & for bitwise AND + # inside expression: unary & for taking the address of something + # inside declarators: reference parameter + # We will exclude the first two cases by checking that we are not inside a + # function body, including one that was just introduced by a trailing '{'. + # TODO(unknown): Doesn't account for 'catch(Exception& e)' [rare]. + if (nesting_state.previous_stack_top and + not (isinstance(nesting_state.previous_stack_top, _ClassInfo) or + isinstance(nesting_state.previous_stack_top, _NamespaceInfo))): + # Not at toplevel, not within a class, and not within a namespace + return + + # Avoid initializer lists. We only need to scan back from the + # current line for something that starts with ':'. + # + # We don't need to check the current line, since the '&' would + # appear inside the second set of parentheses on the current line as + # opposed to the first set. + if linenum > 0: + for i in xrange(linenum - 1, max(0, linenum - 10), -1): + previous_line = clean_lines.elided[i] + if not Search(r'[),]\s*$', previous_line): + break + if Match(r'^\s*:\s+\S', previous_line): + return + + # Avoid preprocessors + if Search(r'\\\s*$', line): + return + + # Avoid constructor initializer lists + if IsInitializerList(clean_lines, linenum): + return + + # We allow non-const references in a few standard places, like functions + # called "swap()" or iostream operators like "<<" or ">>". Do not check + # those function parameters. + # + # We also accept & in static_assert, which looks like a function but + # it's actually a declaration expression. + whitelisted_functions = (r'(?:[sS]wap(?:<\w:+>)?|' + r'operator\s*[<>][<>]|' + r'static_assert|COMPILE_ASSERT' + r')\s*\(') + if Search(whitelisted_functions, line): + return + elif not Search(r'\S+\([^)]*$', line): + # Don't see a whitelisted function on this line. Actually we + # didn't see any function name on this line, so this is likely a + # multi-line parameter list. Try a bit harder to catch this case. + for i in xrange(2): + if (linenum > i and + Search(whitelisted_functions, clean_lines.elided[linenum - i - 1])): + return + + decls = ReplaceAll(r'{[^}]*}', ' ', line) # exclude function body + for parameter in re.findall(_RE_PATTERN_REF_PARAM, decls): + if not Match(_RE_PATTERN_CONST_REF_PARAM, parameter): + error(filename, linenum, 'runtime/references', 2, + 'Is this a non-const reference? ' + 'If so, make const or use a pointer: ' + + ReplaceAll(' *<', '<', parameter)) + + +def CheckCasts(filename, clean_lines, linenum, error): + """Various cast related checks. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + line = clean_lines.elided[linenum] + + # Check to see if they're using an conversion function cast. + # I just try to capture the most common basic types, though there are more. + # Parameterless conversion functions, such as bool(), are allowed as they are + # probably a member operator declaration or default constructor. + match = Search( + r'(\bnew\s+|\S<\s*(?:const\s+)?)?\b' + r'(int|float|double|bool|char|int32|uint32|int64|uint64)' + r'(\([^)].*)', line) + expecting_function = ExpectingFunctionArgs(clean_lines, linenum) + if match and not expecting_function: + matched_type = match.group(2) + + # matched_new_or_template is used to silence two false positives: + # - New operators + # - Template arguments with function types + # + # For template arguments, we match on types immediately following + # an opening bracket without any spaces. This is a fast way to + # silence the common case where the function type is the first + # template argument. False negative with less-than comparison is + # avoided because those operators are usually followed by a space. + # + # function // bracket + no space = false positive + # value < double(42) // bracket + space = true positive + matched_new_or_template = match.group(1) + + # Avoid arrays by looking for brackets that come after the closing + # parenthesis. + if Match(r'\([^()]+\)\s*\[', match.group(3)): + return + + # Other things to ignore: + # - Function pointers + # - Casts to pointer types + # - Placement new + # - Alias declarations + matched_funcptr = match.group(3) + if (matched_new_or_template is None and + not (matched_funcptr and + (Match(r'\((?:[^() ]+::\s*\*\s*)?[^() ]+\)\s*\(', + matched_funcptr) or + matched_funcptr.startswith('(*)'))) and + not Match(r'\s*using\s+\S+\s*=\s*' + matched_type, line) and + not Search(r'new\(\S+\)\s*' + matched_type, line)): + error(filename, linenum, 'readability/casting', 4, + 'Using deprecated casting style. ' + 'Use static_cast<%s>(...) instead' % + matched_type) + + if not expecting_function: + CheckCStyleCast(filename, clean_lines, linenum, 'static_cast', + r'\((int|float|double|bool|char|u?int(16|32|64))\)', error) + + # This doesn't catch all cases. Consider (const char * const)"hello". + # + # (char *) "foo" should always be a const_cast (reinterpret_cast won't + # compile). + if CheckCStyleCast(filename, clean_lines, linenum, 'const_cast', + r'\((char\s?\*+\s?)\)\s*"', error): + pass + else: + # Check pointer casts for other than string constants + CheckCStyleCast(filename, clean_lines, linenum, 'reinterpret_cast', + r'\((\w+\s?\*+\s?)\)', error) + + # In addition, we look for people taking the address of a cast. This + # is dangerous -- casts can assign to temporaries, so the pointer doesn't + # point where you think. + # + # Some non-identifier character is required before the '&' for the + # expression to be recognized as a cast. These are casts: + # expression = &static_cast(temporary()); + # function(&(int*)(temporary())); + # + # This is not a cast: + # reference_type&(int* function_param); + match = Search( + r'(?:[^\w]&\(([^)*][^)]*)\)[\w(])|' + r'(?:[^\w]&(static|dynamic|down|reinterpret)_cast\b)', line) + if match: + # Try a better error message when the & is bound to something + # dereferenced by the casted pointer, as opposed to the casted + # pointer itself. + parenthesis_error = False + match = Match(r'^(.*&(?:static|dynamic|down|reinterpret)_cast\b)<', line) + if match: + _, y1, x1 = CloseExpression(clean_lines, linenum, len(match.group(1))) + if x1 >= 0 and clean_lines.elided[y1][x1] == '(': + _, y2, x2 = CloseExpression(clean_lines, y1, x1) + if x2 >= 0: + extended_line = clean_lines.elided[y2][x2:] + if y2 < clean_lines.NumLines() - 1: + extended_line += clean_lines.elided[y2 + 1] + if Match(r'\s*(?:->|\[)', extended_line): + parenthesis_error = True + + if parenthesis_error: + error(filename, linenum, 'readability/casting', 4, + ('Are you taking an address of something dereferenced ' + 'from a cast? Wrapping the dereferenced expression in ' + 'parentheses will make the binding more obvious')) + else: + error(filename, linenum, 'runtime/casting', 4, + ('Are you taking an address of a cast? ' + 'This is dangerous: could be a temp var. ' + 'Take the address before doing the cast, rather than after')) + + +def CheckCStyleCast(filename, clean_lines, linenum, cast_type, pattern, error): + """Checks for a C-style cast by looking for the pattern. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + cast_type: The string for the C++ cast to recommend. This is either + reinterpret_cast, static_cast, or const_cast, depending. + pattern: The regular expression used to find C-style casts. + error: The function to call with any errors found. + + Returns: + True if an error was emitted. + False otherwise. + """ + line = clean_lines.elided[linenum] + match = Search(pattern, line) + if not match: + return False + + # Exclude lines with keywords that tend to look like casts + context = line[0:match.start(1) - 1] + if Match(r'.*\b(?:sizeof|alignof|alignas|[_A-Z][_A-Z0-9]*)\s*$', context): + return False + + # Try expanding current context to see if we one level of + # parentheses inside a macro. + if linenum > 0: + for i in xrange(linenum - 1, max(0, linenum - 5), -1): + context = clean_lines.elided[i] + context + if Match(r'.*\b[_A-Z][_A-Z0-9]*\s*\((?:\([^()]*\)|[^()])*$', context): + return False + + # operator++(int) and operator--(int) + if context.endswith(' operator++') or context.endswith(' operator--'): + return False + + # A single unnamed argument for a function tends to look like old + # style cast. If we see those, don't issue warnings for deprecated + # casts, instead issue warnings for unnamed arguments where + # appropriate. + # + # These are things that we want warnings for, since the style guide + # explicitly require all parameters to be named: + # Function(int); + # Function(int) { + # ConstMember(int) const; + # ConstMember(int) const { + # ExceptionMember(int) throw (...); + # ExceptionMember(int) throw (...) { + # PureVirtual(int) = 0; + # [](int) -> bool { + # + # These are functions of some sort, where the compiler would be fine + # if they had named parameters, but people often omit those + # identifiers to reduce clutter: + # (FunctionPointer)(int); + # (FunctionPointer)(int) = value; + # Function((function_pointer_arg)(int)) + # Function((function_pointer_arg)(int), int param) + # ; + # <(FunctionPointerTemplateArgument)(int)>; + remainder = line[match.end(0):] + if Match(r'^\s*(?:;|const\b|throw\b|final\b|override\b|[=>{),]|->)', + remainder): + # Looks like an unnamed parameter. + + # Don't warn on any kind of template arguments. + if Match(r'^\s*>', remainder): + return False + + # Don't warn on assignments to function pointers, but keep warnings for + # unnamed parameters to pure virtual functions. Note that this pattern + # will also pass on assignments of "0" to function pointers, but the + # preferred values for those would be "nullptr" or "NULL". + matched_zero = Match(r'^\s=\s*(\S+)\s*;', remainder) + if matched_zero and matched_zero.group(1) != '0': + return False + + # Don't warn on function pointer declarations. For this we need + # to check what came before the "(type)" string. + if Match(r'.*\)\s*$', line[0:match.start(0)]): + return False + + # Don't warn if the parameter is named with block comments, e.g.: + # Function(int /*unused_param*/); + raw_line = clean_lines.raw_lines[linenum] + if '/*' in raw_line: + return False + + # Passed all filters, issue warning here. + error(filename, linenum, 'readability/function', 3, + 'All parameters should be named in a function') + return True + + # At this point, all that should be left is actual casts. + error(filename, linenum, 'readability/casting', 4, + 'Using C-style cast. Use %s<%s>(...) instead' % + (cast_type, match.group(1))) + + return True + + +def ExpectingFunctionArgs(clean_lines, linenum): + """Checks whether where function type arguments are expected. + + Args: + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + + Returns: + True if the line at 'linenum' is inside something that expects arguments + of function types. + """ + line = clean_lines.elided[linenum] + return (Match(r'^\s*MOCK_(CONST_)?METHOD\d+(_T)?\(', line) or + (linenum >= 2 and + (Match(r'^\s*MOCK_(?:CONST_)?METHOD\d+(?:_T)?\((?:\S+,)?\s*$', + clean_lines.elided[linenum - 1]) or + Match(r'^\s*MOCK_(?:CONST_)?METHOD\d+(?:_T)?\(\s*$', + clean_lines.elided[linenum - 2]) or + Search(r'\bstd::m?function\s*\<\s*$', + clean_lines.elided[linenum - 1])))) + + +_HEADERS_CONTAINING_TEMPLATES = ( + ('', ('deque',)), + ('', ('unary_function', 'binary_function', + 'plus', 'minus', 'multiplies', 'divides', 'modulus', + 'negate', + 'equal_to', 'not_equal_to', 'greater', 'less', + 'greater_equal', 'less_equal', + 'logical_and', 'logical_or', 'logical_not', + 'unary_negate', 'not1', 'binary_negate', 'not2', + 'bind1st', 'bind2nd', + 'pointer_to_unary_function', + 'pointer_to_binary_function', + 'ptr_fun', + 'mem_fun_t', 'mem_fun', 'mem_fun1_t', 'mem_fun1_ref_t', + 'mem_fun_ref_t', + 'const_mem_fun_t', 'const_mem_fun1_t', + 'const_mem_fun_ref_t', 'const_mem_fun1_ref_t', + 'mem_fun_ref', + )), + ('', ('numeric_limits',)), + ('', ('list',)), + ('', ('map', 'multimap',)), + ('', ('allocator',)), + ('', ('queue', 'priority_queue',)), + ('', ('set', 'multiset',)), + ('', ('stack',)), + ('', ('char_traits', 'basic_string',)), + ('', ('tuple',)), + ('', ('pair',)), + ('', ('vector',)), + + # gcc extensions. + # Note: std::hash is their hash, ::hash is our hash + ('', ('hash_map', 'hash_multimap',)), + ('', ('hash_set', 'hash_multiset',)), + ('', ('slist',)), + ) + +_RE_PATTERN_STRING = re.compile(r'\bstring\b') + +_re_pattern_algorithm_header = [] +for _template in ('copy', 'max', 'min', 'min_element', 'sort', 'swap', + 'transform'): + # Match max(..., ...), max(..., ...), but not foo->max, foo.max or + # type::max(). + _re_pattern_algorithm_header.append( + (re.compile(r'[^>.]\b' + _template + r'(<.*?>)?\([^\)]'), + _template, + '')) + +_re_pattern_templates = [] +for _header, _templates in _HEADERS_CONTAINING_TEMPLATES: + for _template in _templates: + _re_pattern_templates.append( + (re.compile(r'(\<|\b)' + _template + r'\s*\<'), + _template + '<>', + _header)) + + +def FilesBelongToSameModule(filename_cc, filename_h): + """Check if these two filenames belong to the same module. + + The concept of a 'module' here is a as follows: + foo.h, foo-inl.h, foo.cc, foo_test.cc and foo_unittest.cc belong to the + same 'module' if they are in the same directory. + some/path/public/xyzzy and some/path/internal/xyzzy are also considered + to belong to the same module here. + + If the filename_cc contains a longer path than the filename_h, for example, + '/absolute/path/to/base/sysinfo.cc', and this file would include + 'base/sysinfo.h', this function also produces the prefix needed to open the + header. This is used by the caller of this function to more robustly open the + header file. We don't have access to the real include paths in this context, + so we need this guesswork here. + + Known bugs: tools/base/bar.cc and base/bar.h belong to the same module + according to this implementation. Because of this, this function gives + some false positives. This should be sufficiently rare in practice. + + Args: + filename_cc: is the path for the .cc file + filename_h: is the path for the header path + + Returns: + Tuple with a bool and a string: + bool: True if filename_cc and filename_h belong to the same module. + string: the additional prefix needed to open the header file. + """ + + if not filename_cc.endswith('.cc'): + return (False, '') + filename_cc = filename_cc[:-len('.cc')] + if filename_cc.endswith('_unittest'): + filename_cc = filename_cc[:-len('_unittest')] + elif filename_cc.endswith('_test'): + filename_cc = filename_cc[:-len('_test')] + filename_cc = filename_cc.replace('/public/', '/') + filename_cc = filename_cc.replace('/internal/', '/') + + if not filename_h.endswith('.h'): + return (False, '') + filename_h = filename_h[:-len('.h')] + if filename_h.endswith('-inl'): + filename_h = filename_h[:-len('-inl')] + filename_h = filename_h.replace('/public/', '/') + filename_h = filename_h.replace('/internal/', '/') + + files_belong_to_same_module = filename_cc.endswith(filename_h) + common_path = '' + if files_belong_to_same_module: + common_path = filename_cc[:-len(filename_h)] + return files_belong_to_same_module, common_path + + +def UpdateIncludeState(filename, include_dict, io=codecs): + """Fill up the include_dict with new includes found from the file. + + Args: + filename: the name of the header to read. + include_dict: a dictionary in which the headers are inserted. + io: The io factory to use to read the file. Provided for testability. + + Returns: + True if a header was successfully added. False otherwise. + """ + headerfile = None + try: + headerfile = io.open(filename, 'r', 'utf8', 'replace') + except IOError: + return False + linenum = 0 + for line in headerfile: + linenum += 1 + clean_line = CleanseComments(line) + match = _RE_PATTERN_INCLUDE.search(clean_line) + if match: + include = match.group(2) + include_dict.setdefault(include, linenum) + return True + + +def CheckForIncludeWhatYouUse(filename, clean_lines, include_state, error, + io=codecs): + """Reports for missing stl includes. + + This function will output warnings to make sure you are including the headers + necessary for the stl containers and functions that you use. We only give one + reason to include a header. For example, if you use both equal_to<> and + less<> in a .h file, only one (the latter in the file) of these will be + reported as a reason to include the . + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + include_state: An _IncludeState instance. + error: The function to call with any errors found. + io: The IO factory to use to read the header file. Provided for unittest + injection. + """ + required = {} # A map of header name to linenumber and the template entity. + # Example of required: { '': (1219, 'less<>') } + + for linenum in xrange(clean_lines.NumLines()): + line = clean_lines.elided[linenum] + if not line or line[0] == '#': + continue + + # String is special -- it is a non-templatized type in STL. + matched = _RE_PATTERN_STRING.search(line) + if matched: + # Don't warn about strings in non-STL namespaces: + # (We check only the first match per line; good enough.) + prefix = line[:matched.start()] + if prefix.endswith('std::') or not prefix.endswith('::'): + required[''] = (linenum, 'string') + + for pattern, template, header in _re_pattern_algorithm_header: + if pattern.search(line): + required[header] = (linenum, template) + + # The following function is just a speed up, no semantics are changed. + if not '<' in line: # Reduces the cpu time usage by skipping lines. + continue + + for pattern, template, header in _re_pattern_templates: + if pattern.search(line): + required[header] = (linenum, template) + + # The policy is that if you #include something in foo.h you don't need to + # include it again in foo.cc. Here, we will look at possible includes. + # Let's flatten the include_state include_list and copy it into a dictionary. + include_dict = dict([item for sublist in include_state.include_list + for item in sublist]) + + # Did we find the header for this file (if any) and successfully load it? + header_found = False + + # Use the absolute path so that matching works properly. + abs_filename = FileInfo(filename).FullName() + + # For Emacs's flymake. + # If cpplint is invoked from Emacs's flymake, a temporary file is generated + # by flymake and that file name might end with '_flymake.cc'. In that case, + # restore original file name here so that the corresponding header file can be + # found. + # e.g. If the file name is 'foo_flymake.cc', we should search for 'foo.h' + # instead of 'foo_flymake.h' + abs_filename = re.sub(r'_flymake\.cc$', '.cc', abs_filename) + + # include_dict is modified during iteration, so we iterate over a copy of + # the keys. + header_keys = include_dict.keys() + for header in header_keys: + (same_module, common_path) = FilesBelongToSameModule(abs_filename, header) + fullpath = common_path + header + if same_module and UpdateIncludeState(fullpath, include_dict, io): + header_found = True + + # If we can't find the header file for a .cc, assume it's because we don't + # know where to look. In that case we'll give up as we're not sure they + # didn't include it in the .h file. + # TODO(unknown): Do a better job of finding .h files so we are confident that + # not having the .h file means there isn't one. + if filename.endswith('.cc') and not header_found: + return + + # All the lines have been processed, report the errors found. + for required_header_unstripped in required: + template = required[required_header_unstripped][1] + if required_header_unstripped.strip('<>"') not in include_dict: + error(filename, required[required_header_unstripped][0], + 'build/include_what_you_use', 4, + 'Add #include ' + required_header_unstripped + ' for ' + template) + + +_RE_PATTERN_EXPLICIT_MAKEPAIR = re.compile(r'\bmake_pair\s*<') + + +def CheckMakePairUsesDeduction(filename, clean_lines, linenum, error): + """Check that make_pair's template arguments are deduced. + + G++ 4.6 in C++11 mode fails badly if make_pair's template arguments are + specified explicitly, and such use isn't intended in any case. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + line = clean_lines.elided[linenum] + match = _RE_PATTERN_EXPLICIT_MAKEPAIR.search(line) + if match: + error(filename, linenum, 'build/explicit_make_pair', + 4, # 4 = high confidence + 'For C++11-compatibility, omit template arguments from make_pair' + ' OR use pair directly OR if appropriate, construct a pair directly') + + +def CheckDefaultLambdaCaptures(filename, clean_lines, linenum, error): + """Check that default lambda captures are not used. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + line = clean_lines.elided[linenum] + + # A lambda introducer specifies a default capture if it starts with "[=" + # or if it starts with "[&" _not_ followed by an identifier. + match = Match(r'^(.*)\[\s*(?:=|&[^\w])', line) + if match: + # Found a potential error, check what comes after the lambda-introducer. + # If it's not open parenthesis (for lambda-declarator) or open brace + # (for compound-statement), it's not a lambda. + line, _, pos = CloseExpression(clean_lines, linenum, len(match.group(1))) + if pos >= 0 and Match(r'^\s*[{(]', line[pos:]): + error(filename, linenum, 'build/c++11', + 4, # 4 = high confidence + 'Default lambda captures are an unapproved C++ feature.') + + +def CheckRedundantVirtual(filename, clean_lines, linenum, error): + """Check if line contains a redundant "virtual" function-specifier. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + # Look for "virtual" on current line. + line = clean_lines.elided[linenum] + virtual = Match(r'^(.*)(\bvirtual\b)(.*)$', line) + if not virtual: return + + # Ignore "virtual" keywords that are near access-specifiers. These + # are only used in class base-specifier and do not apply to member + # functions. + if (Search(r'\b(public|protected|private)\s+$', virtual.group(1)) or + Match(r'^\s+(public|protected|private)\b', virtual.group(3))): + return + + # Ignore the "virtual" keyword from virtual base classes. Usually + # there is a column on the same line in these cases (virtual base + # classes are rare in google3 because multiple inheritance is rare). + if Match(r'^.*[^:]:[^:].*$', line): return + + # Look for the next opening parenthesis. This is the start of the + # parameter list (possibly on the next line shortly after virtual). + # TODO(unknown): doesn't work if there are virtual functions with + # decltype() or other things that use parentheses, but csearch suggests + # that this is rare. + end_col = -1 + end_line = -1 + start_col = len(virtual.group(2)) + for start_line in xrange(linenum, min(linenum + 3, clean_lines.NumLines())): + line = clean_lines.elided[start_line][start_col:] + parameter_list = Match(r'^([^(]*)\(', line) + if parameter_list: + # Match parentheses to find the end of the parameter list + (_, end_line, end_col) = CloseExpression( + clean_lines, start_line, start_col + len(parameter_list.group(1))) + break + start_col = 0 + + if end_col < 0: + return # Couldn't find end of parameter list, give up + + # Look for "override" or "final" after the parameter list + # (possibly on the next few lines). + for i in xrange(end_line, min(end_line + 3, clean_lines.NumLines())): + line = clean_lines.elided[i][end_col:] + match = Search(r'\b(override|final)\b', line) + if match: + error(filename, linenum, 'readability/inheritance', 4, + ('"virtual" is redundant since function is ' + 'already declared as "%s"' % match.group(1))) + + # Set end_col to check whole lines after we are done with the + # first line. + end_col = 0 + if Search(r'[^\w]\s*$', line): + break + + +def CheckRedundantOverrideOrFinal(filename, clean_lines, linenum, error): + """Check if line contains a redundant "override" or "final" virt-specifier. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + # Look for closing parenthesis nearby. We need one to confirm where + # the declarator ends and where the virt-specifier starts to avoid + # false positives. + line = clean_lines.elided[linenum] + declarator_end = line.rfind(')') + if declarator_end >= 0: + fragment = line[declarator_end:] + else: + if linenum > 1 and clean_lines.elided[linenum - 1].rfind(')') >= 0: + fragment = line + else: + return + + # Check that at most one of "override" or "final" is present, not both + if Search(r'\boverride\b', fragment) and Search(r'\bfinal\b', fragment): + error(filename, linenum, 'readability/inheritance', 4, + ('"override" is redundant since function is ' + 'already declared as "final"')) + + + + +# Returns true if we are at a new block, and it is directly +# inside of a namespace. +def IsBlockInNameSpace(nesting_state, is_forward_declaration): + """Checks that the new block is directly in a namespace. + + Args: + nesting_state: The _NestingState object that contains info about our state. + is_forward_declaration: If the class is a forward declared class. + Returns: + Whether or not the new block is directly in a namespace. + """ + if is_forward_declaration: + if len(nesting_state.stack) >= 1 and ( + isinstance(nesting_state.stack[-1], _NamespaceInfo)): + return True + else: + return False + + return (len(nesting_state.stack) > 1 and + nesting_state.stack[-1].check_namespace_indentation and + isinstance(nesting_state.stack[-2], _NamespaceInfo)) + + +def ShouldCheckNamespaceIndentation(nesting_state, is_namespace_indent_item, + raw_lines_no_comments, linenum): + """This method determines if we should apply our namespace indentation check. + + Args: + nesting_state: The current nesting state. + is_namespace_indent_item: If we just put a new class on the stack, True. + If the top of the stack is not a class, or we did not recently + add the class, False. + raw_lines_no_comments: The lines without the comments. + linenum: The current line number we are processing. + + Returns: + True if we should apply our namespace indentation check. Currently, it + only works for classes and namespaces inside of a namespace. + """ + + is_forward_declaration = IsForwardClassDeclaration(raw_lines_no_comments, + linenum) + + if not (is_namespace_indent_item or is_forward_declaration): + return False + + # If we are in a macro, we do not want to check the namespace indentation. + if IsMacroDefinition(raw_lines_no_comments, linenum): + return False + + return IsBlockInNameSpace(nesting_state, is_forward_declaration) + + +# Call this method if the line is directly inside of a namespace. +# If the line above is blank (excluding comments) or the start of +# an inner namespace, it cannot be indented. +def CheckItemIndentationInNamespace(filename, raw_lines_no_comments, linenum, + error): + line = raw_lines_no_comments[linenum] + if Match(r'^\s+', line): + error(filename, linenum, 'runtime/indentation_namespace', 4, + 'Do not indent within a namespace') + + +def ProcessLine(filename, file_extension, clean_lines, line, + include_state, function_state, nesting_state, error, + extra_check_functions=[]): + """Processes a single line in the file. + + Args: + filename: Filename of the file that is being processed. + file_extension: The extension (dot not included) of the file. + clean_lines: An array of strings, each representing a line of the file, + with comments stripped. + line: Number of line being processed. + include_state: An _IncludeState instance in which the headers are inserted. + function_state: A _FunctionState instance which counts function lines, etc. + nesting_state: A NestingState instance which maintains information about + the current stack of nested blocks being parsed. + error: A callable to which errors are reported, which takes 4 arguments: + filename, line number, error level, and message + extra_check_functions: An array of additional check functions that will be + run on each source line. Each function takes 4 + arguments: filename, clean_lines, line, error + """ + raw_lines = clean_lines.raw_lines + ParseNolintSuppressions(filename, raw_lines[line], line, error) + nesting_state.Update(filename, clean_lines, line, error) + CheckForNamespaceIndentation(filename, nesting_state, clean_lines, line, + error) + if nesting_state.InAsmBlock(): return + CheckForFunctionLengths(filename, clean_lines, line, function_state, error) + CheckForMultilineCommentsAndStrings(filename, clean_lines, line, error) + CheckStyle(filename, clean_lines, line, file_extension, nesting_state, error) + CheckLanguage(filename, clean_lines, line, file_extension, include_state, + nesting_state, error) + CheckForNonConstReference(filename, clean_lines, line, nesting_state, error) + CheckForNonStandardConstructs(filename, clean_lines, line, + nesting_state, error) + CheckVlogArguments(filename, clean_lines, line, error) + CheckPosixThreading(filename, clean_lines, line, error) + CheckInvalidIncrement(filename, clean_lines, line, error) + CheckMakePairUsesDeduction(filename, clean_lines, line, error) + CheckDefaultLambdaCaptures(filename, clean_lines, line, error) + CheckRedundantVirtual(filename, clean_lines, line, error) + CheckRedundantOverrideOrFinal(filename, clean_lines, line, error) + for check_fn in extra_check_functions: + check_fn(filename, clean_lines, line, error) + +def FlagCxx11Features(filename, clean_lines, linenum, error): + """Flag those c++11 features that we only allow in certain places. + + Args: + filename: The name of the current file. + clean_lines: A CleansedLines instance containing the file. + linenum: The number of the line to check. + error: The function to call with any errors found. + """ + line = clean_lines.elided[linenum] + + # Flag unapproved C++11 headers. + include = Match(r'\s*#\s*include\s+[<"]([^<"]+)[">]', line) + if include and include.group(1) in ('cfenv', + 'condition_variable', + 'fenv.h', + 'future', + 'mutex', + 'thread', + 'chrono', + 'ratio', + 'regex', + 'system_error', + ): + error(filename, linenum, 'build/c++11', 5, + ('<%s> is an unapproved C++11 header.') % include.group(1)) + + # The only place where we need to worry about C++11 keywords and library + # features in preprocessor directives is in macro definitions. + if Match(r'\s*#', line) and not Match(r'\s*#\s*define\b', line): return + + # These are classes and free functions. The classes are always + # mentioned as std::*, but we only catch the free functions if + # they're not found by ADL. They're alphabetical by header. + for top_name in ( + # type_traits + 'alignment_of', + 'aligned_union', + ): + if Search(r'\bstd::%s\b' % top_name, line): + error(filename, linenum, 'build/c++11', 5, + ('std::%s is an unapproved C++11 class or function. Send c-style ' + 'an example of where it would make your code more readable, and ' + 'they may let you use it.') % top_name) + + +def ProcessFileData(filename, file_extension, lines, error, + extra_check_functions=[]): + """Performs lint checks and reports any errors to the given error function. + + Args: + filename: Filename of the file that is being processed. + file_extension: The extension (dot not included) of the file. + lines: An array of strings, each representing a line of the file, with the + last element being empty if the file is terminated with a newline. + error: A callable to which errors are reported, which takes 4 arguments: + filename, line number, error level, and message + extra_check_functions: An array of additional check functions that will be + run on each source line. Each function takes 4 + arguments: filename, clean_lines, line, error + """ + lines = (['// marker so line numbers and indices both start at 1'] + lines + + ['// marker so line numbers end in a known way']) + + include_state = _IncludeState() + function_state = _FunctionState() + nesting_state = NestingState() + + ResetNolintSuppressions() + + CheckForCopyright(filename, lines, error) + + RemoveMultiLineComments(filename, lines, error) + clean_lines = CleansedLines(lines) + + if file_extension == 'h': + CheckForHeaderGuard(filename, clean_lines, error) + + for line in xrange(clean_lines.NumLines()): + ProcessLine(filename, file_extension, clean_lines, line, + include_state, function_state, nesting_state, error, + extra_check_functions) + FlagCxx11Features(filename, clean_lines, line, error) + nesting_state.CheckCompletedBlocks(filename, error) + + CheckForIncludeWhatYouUse(filename, clean_lines, include_state, error) + + # Check that the .cc file has included its header if it exists. + if file_extension == 'cc': + CheckHeaderFileIncluded(filename, include_state, error) + + # We check here rather than inside ProcessLine so that we see raw + # lines rather than "cleaned" lines. + CheckForBadCharacters(filename, lines, error) + + CheckForNewlineAtEOF(filename, lines, error) + +def ProcessConfigOverrides(filename): + """ Loads the configuration files and processes the config overrides. + + Args: + filename: The name of the file being processed by the linter. + + Returns: + False if the current |filename| should not be processed further. + """ + + abs_filename = os.path.abspath(filename) + cfg_filters = [] + keep_looking = True + while keep_looking: + abs_path, base_name = os.path.split(abs_filename) + if not base_name: + break # Reached the root directory. + + cfg_file = os.path.join(abs_path, "CPPLINT.cfg") + abs_filename = abs_path + if not os.path.isfile(cfg_file): + continue + + try: + with open(cfg_file) as file_handle: + for line in file_handle: + line, _, _ = line.partition('#') # Remove comments. + if not line.strip(): + continue + + name, _, val = line.partition('=') + name = name.strip() + val = val.strip() + if name == 'set noparent': + keep_looking = False + elif name == 'filter': + cfg_filters.append(val) + elif name == 'exclude_files': + # When matching exclude_files pattern, use the base_name of + # the current file name or the directory name we are processing. + # For example, if we are checking for lint errors in /foo/bar/baz.cc + # and we found the .cfg file at /foo/CPPLINT.cfg, then the config + # file's "exclude_files" filter is meant to be checked against "bar" + # and not "baz" nor "bar/baz.cc". + if base_name: + pattern = re.compile(val) + if pattern.match(base_name): + sys.stderr.write('Ignoring "%s": file excluded by "%s". ' + 'File path component "%s" matches ' + 'pattern "%s"\n' % + (filename, cfg_file, base_name, val)) + return False + elif name == 'linelength': + global _line_length + try: + _line_length = int(val) + except ValueError: + sys.stderr.write('Line length must be numeric.') + else: + sys.stderr.write( + 'Invalid configuration option (%s) in file %s\n' % + (name, cfg_file)) + + except IOError: + sys.stderr.write( + "Skipping config file '%s': Can't open for reading\n" % cfg_file) + keep_looking = False + + # Apply all the accumulated filters in reverse order (top-level directory + # config options having the least priority). + for filter in reversed(cfg_filters): + _AddFilters(filter) + + return True + + +def ProcessFile(filename, vlevel, extra_check_functions=[]): + """Does google-lint on a single file. + + Args: + filename: The name of the file to parse. + + vlevel: The level of errors to report. Every error of confidence + >= verbose_level will be reported. 0 is a good default. + + extra_check_functions: An array of additional check functions that will be + run on each source line. Each function takes 4 + arguments: filename, clean_lines, line, error + """ + + _SetVerboseLevel(vlevel) + _BackupFilters() + + if not ProcessConfigOverrides(filename): + _RestoreFilters() + return + + lf_lines = [] + crlf_lines = [] + try: + # Support the UNIX convention of using "-" for stdin. Note that + # we are not opening the file with universal newline support + # (which codecs doesn't support anyway), so the resulting lines do + # contain trailing '\r' characters if we are reading a file that + # has CRLF endings. + # If after the split a trailing '\r' is present, it is removed + # below. + if filename == '-': + lines = codecs.StreamReaderWriter(sys.stdin, + codecs.getreader('utf8'), + codecs.getwriter('utf8'), + 'replace').read().split('\n') + else: + lines = codecs.open(filename, 'r', 'utf8', 'replace').read().split('\n') + + # Remove trailing '\r'. + # The -1 accounts for the extra trailing blank line we get from split() + for linenum in range(len(lines) - 1): + if lines[linenum].endswith('\r'): + lines[linenum] = lines[linenum].rstrip('\r') + crlf_lines.append(linenum + 1) + else: + lf_lines.append(linenum + 1) + + except IOError: + sys.stderr.write( + "Skipping input '%s': Can't open for reading\n" % filename) + _RestoreFilters() + return + + # Note, if no dot is found, this will give the entire filename as the ext. + file_extension = filename[filename.rfind('.') + 1:] + + # When reading from stdin, the extension is unknown, so no cpplint tests + # should rely on the extension. + if filename != '-' and file_extension not in _valid_extensions: + sys.stderr.write('Ignoring %s; not a valid file name ' + '(%s)\n' % (filename, ', '.join(_valid_extensions))) + else: + ProcessFileData(filename, file_extension, lines, Error, + extra_check_functions) + + # If end-of-line sequences are a mix of LF and CR-LF, issue + # warnings on the lines with CR. + # + # Don't issue any warnings if all lines are uniformly LF or CR-LF, + # since critique can handle these just fine, and the style guide + # doesn't dictate a particular end of line sequence. + # + # We can't depend on os.linesep to determine what the desired + # end-of-line sequence should be, since that will return the + # server-side end-of-line sequence. + if lf_lines and crlf_lines: + # Warn on every line with CR. An alternative approach might be to + # check whether the file is mostly CRLF or just LF, and warn on the + # minority, we bias toward LF here since most tools prefer LF. + for linenum in crlf_lines: + Error(filename, linenum, 'whitespace/newline', 1, + 'Unexpected \\r (^M) found; better to use only \\n') + + sys.stderr.write('Done processing %s\n' % filename) + _RestoreFilters() + + +def PrintUsage(message): + """Prints a brief usage string and exits, optionally with an error message. + + Args: + message: The optional error message. + """ + sys.stderr.write(_USAGE) + if message: + sys.exit('\nFATAL ERROR: ' + message) + else: + sys.exit(1) + + +def PrintCategories(): + """Prints a list of all the error-categories used by error messages. + + These are the categories used to filter messages via --filter. + """ + sys.stderr.write(''.join(' %s\n' % cat for cat in _ERROR_CATEGORIES)) + sys.exit(0) + + +def ParseArguments(args): + """Parses the command line arguments. + + This may set the output format and verbosity level as side-effects. + + Args: + args: The command line arguments: + + Returns: + The list of filenames to lint. + """ + try: + (opts, filenames) = getopt.getopt(args, '', ['help', 'output=', 'verbose=', + 'counting=', + 'filter=', + 'root=', + 'linelength=', + 'extensions=']) + except getopt.GetoptError: + PrintUsage('Invalid arguments.') + + verbosity = _VerboseLevel() + output_format = _OutputFormat() + filters = '' + counting_style = '' + + for (opt, val) in opts: + if opt == '--help': + PrintUsage(None) + elif opt == '--output': + if val not in ('emacs', 'vs7', 'eclipse'): + PrintUsage('The only allowed output formats are emacs, vs7 and eclipse.') + output_format = val + elif opt == '--verbose': + verbosity = int(val) + elif opt == '--filter': + filters = val + if not filters: + PrintCategories() + elif opt == '--counting': + if val not in ('total', 'toplevel', 'detailed'): + PrintUsage('Valid counting options are total, toplevel, and detailed') + counting_style = val + elif opt == '--root': + global _root + _root = val + elif opt == '--linelength': + global _line_length + try: + _line_length = int(val) + except ValueError: + PrintUsage('Line length must be digits.') + elif opt == '--extensions': + global _valid_extensions + try: + _valid_extensions = set(val.split(',')) + except ValueError: + PrintUsage('Extensions must be comma seperated list.') + + if not filenames: + PrintUsage('No files were specified.') + + _SetOutputFormat(output_format) + _SetVerboseLevel(verbosity) + _SetFilters(filters) + _SetCountingStyle(counting_style) + + return filenames + + +def main(): + filenames = ParseArguments(sys.argv[1:]) + + # Change stderr to write with replacement characters so we don't die + # if we try to print something containing non-ASCII characters. + sys.stderr = codecs.StreamReaderWriter(sys.stderr, + codecs.getreader('utf8'), + codecs.getwriter('utf8'), + 'replace') + + _cpplint_state.ResetErrorCounts() + for filename in filenames: + ProcessFile(filename, _cpplint_state.verbose_level) + _cpplint_state.PrintErrorCounts() + + sys.exit(_cpplint_state.error_count > 0) + + +if __name__ == '__main__': + main() diff --git a/deps/librdkafka b/deps/librdkafka new file mode 160000 index 00000000..2213fb29 --- /dev/null +++ b/deps/librdkafka @@ -0,0 +1 @@ +Subproject commit 2213fb29f98a7a73f22da21ef85e0783f6fd67c4 diff --git a/deps/librdkafka.gyp b/deps/librdkafka.gyp new file mode 100644 index 00000000..ab6d9ac9 --- /dev/null +++ b/deps/librdkafka.gyp @@ -0,0 +1,188 @@ +{ + 'targets': [ + { + "target_name": "librdkafka_cpp", + "type": "static_library", + "include_dirs": [ + "librdkafka/src-cpp", + "librdkafka/src" + ], + "dependencies": [ + "librdkafka" + ], + 'sources': [ + "librdkafka/src-cpp/RdKafka.cpp", + "librdkafka/src-cpp/ConfImpl.cpp", + "librdkafka/src-cpp/HandleImpl.cpp", + "librdkafka/src-cpp/ConsumerImpl.cpp", + "librdkafka/src-cpp/ProducerImpl.cpp", + "librdkafka/src-cpp/KafkaConsumerImpl.cpp", + "librdkafka/src-cpp/TopicImpl.cpp", + "librdkafka/src-cpp/TopicPartitionImpl.cpp", + "librdkafka/src-cpp/MessageImpl.cpp", + "librdkafka/src-cpp/QueueImpl.cpp", + "librdkafka/src-cpp/MetadataImpl.cpp" + ], + "conditions": [ + [ + 'OS=="linux"', + { + 'cflags_cc!': [ + '-fno-rtti' + ], + 'cflags_cc' : [ + '-Wno-sign-compare', + '-Wno-missing-field-initializers', + '-Wno-empty-body', + '-g' + ], + } + ], + ['OS=="mac"', { + 'xcode_settings': { + 'OTHER_CFLAGS': [ + '-ObjC' + ], + 'MACOSX_DEPLOYMENT_TARGET': '10.7', + 'GCC_ENABLE_CPP_RTTI': 'YES', + 'OTHER_CPLUSPLUSFLAGS': [ + '-std=c++11', + '-stdlib=libc++' + ], + 'OTHER_LDFLAGS': [], + }, + 'defines': [ + 'FWD_LINKING_REQ' + ] + }] + ] + }, + { + "target_name": "librdkafka", + "type": "static_library", + 'defines': [ + 'HAVE_CONFIG_H' + ], + "include_dirs": [ + "librdkafka/src" + ], + 'cflags': [ + '-Wunused-function', + '-Wformat', + '-Wimplicit-function-declaration' + ], + "conditions": [ + [ + 'OS=="linux"', + { + 'cflags!': [ + + ], + 'cflags' : [ + '-Wno-type-limits', + '-Wno-unused-function', + '-Wno-maybe-uninitialized', + '-Wno-sign-compare', + '-Wno-missing-field-initializers', + '-Wno-empty-body', + '-Wno-old-style-declaration', + '-g' + ], + "dependencies": [ + "librdkafka_config" + ] + } + ], + [ + 'OS=="mac"', + { + 'xcode_settings': { + 'OTHER_CFLAGS' : [ + '-Wno-sign-compare', + '-Wno-missing-field-initializers', + '-ObjC', + '-Wno-implicit-function-declaration', + '-Wno-unused-function', + '-Wno-format' + ], + 'OTHER_LDFLAGS': [], + 'MACOSX_DEPLOYMENT_TARGET': '10.11', + 'libraries' : ['-lz'] + }, + "dependencies": [ + "librdkafka_config" + ] + } + ], + [ + 'OS=="win"', + { + 'msvs_settings': { + 'VCLinkerTool': { + 'SetChecksum': 'true' + } + }, + } + ] + ], + 'sources': [ + "librdkafka/src/rdgz.c", + "librdkafka/src/rdkafka.c", + "librdkafka/src/rdkafka_feature.c", + "librdkafka/src/rdkafka_broker.c", + "librdkafka/src/rdkafka_msg.c", + "librdkafka/src/rdkafka_topic.c", + "librdkafka/src/rdkafka_conf.c", + "librdkafka/src/rdkafka_timer.c", + "librdkafka/src/rdkafka_offset.c", + "librdkafka/src/rdkafka_transport.c", + "librdkafka/src/rdkafka_buf.c", + "librdkafka/src/rdkafka_queue.c", + "librdkafka/src/rdkafka_op.c", + "librdkafka/src/rdkafka_request.c", + "librdkafka/src/rdkafka_cgrp.c", + "librdkafka/src/rdkafka_pattern.c", + "librdkafka/src/rdkafka_partition.c", + "librdkafka/src/rdkafka_subscription.c", + "librdkafka/src/rdkafka_assignor.c", + "librdkafka/src/rdkafka_range_assignor.c", + "librdkafka/src/rdkafka_roundrobin_assignor.c", + "librdkafka/src/rdcrc32.c", + "librdkafka/src/rdaddr.c", + "librdkafka/src/rdrand.c", + "librdkafka/src/rdlist.c", + "librdkafka/src/tinycthread.c", + "librdkafka/src/rdlog.c", + "librdkafka/src/snappy.c" + ], + 'cflags!': [ '-fno-rtti' ], + }, + { + "target_name": "librdkafka_config", + "type": "none", + "actions": [ + { + 'action_name': 'configure_librdkafka', + 'message': 'configuring librdkafka...', + 'inputs': [ + 'librdkafka/configure', + ], + 'outputs': [ + 'librdkafka/config.h', + ], + "conditions": [ + [ + 'OS!="win"', + { + 'action': ['eval', 'cd librdkafka && chmod a+x ./configure && ./configure --disable-sasl'] + }, + { + 'action': ['echo'] + } + ] + ] + } + ] + } + ] +} diff --git a/docker-compose.yml b/docker-compose.yml new file mode 100644 index 00000000..ba3eed11 --- /dev/null +++ b/docker-compose.yml @@ -0,0 +1,14 @@ +zookeeper: + image: wurstmeister/zookeeper +kafka: + image: wurstmeister/kafka:0.9.0.1 + ports: + - 9092:9092 + links: + - zookeeper + environment: + KAFKA_ZOOKEEPER_CONNECT: zookeeper:2181 + KAFKA_CREATE_TOPICS: "test2:1:3,test:1:1,test3:1:1,test4:1:1,test5:1:1,test6:1:1" + KAFKA_ADVERTISED_HOST_NAME: 127.0.0.1 + KAFKA_ADVERTISED_PORT: 9092 + KAFKA_AUTO_CREATE_TOPICS_ENABLE: 'true' diff --git a/e2e/both.spec.js b/e2e/both.spec.js new file mode 100644 index 00000000..bff6e3d5 --- /dev/null +++ b/e2e/both.spec.js @@ -0,0 +1,189 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +var crypto = require('crypto'); +var t = require('assert'); + +var Kafka = require('../'); + +var kafkaBrokerList = process.env.KAFKA_HOST || 'localhost:9092'; + +var TestCase = require('./test-case'); +var md5hash = crypto.createHash('md5'); +md5hash.update(crypto.randomBytes(20).toString()); +var grp = 'kafka-mocha-grp-' + md5hash.digest('hex'); + +var testCase = new TestCase('Interoperability tests', function() { + + this.test('consumer can read produced messages', function(cb) { + this.timeout(50000); + var producerConnected = false; + var consumerConnected = false; + var producer; + var consumer; + var producerDisconnected = false; + var consumerDisconnected = false; + + var maybeDoneDisconnecting = function() { + if (consumerDisconnected && producerDisconnected) { + cb(); + } + }; + + var disconnect = function() { + return cb(); + /* + console.log('Disconnecting'); + if (consumer) { + console.log('DCing consumer'); + consumer.disconnect(function() { + console.log('Done dcing consumer'); + consumerDisconnected = true; + maybeDoneDisconnecting(); + }); + } else { + consumerDisconnected = true; + maybeDoneDisconnecting(); + } + + if (producer) { + console.log('DCing producer'); + producer.disconnect(function() { + console.log('Done dcing producer'); + producerDisconnected = true; + maybeDoneDisconnecting(); + }); + } else { + producerDisconnected = true; + maybeDoneDisconnecting(); + } + */ + + }; + + // Maybe Done + + var maybeDone = function() { + if (producerConnected && consumerConnected) { + var topic = 'test'; + + crypto.randomBytes(4096, function(ex, buffer) { + + consumer.on('rebalance', function(e) { + if (e.code === 500) { + setTimeout(function() { + producer.produce({ + message: buffer, + topic: topic + }, function(err) { + t.ifError(err); + }); + }, 1000); + } + }); + + var tt = setInterval(function() { + if (!producer.isConnected()) { + clearInterval(tt); + } else { + producer.poll(); + } + }, 100); + + var offset; + + producer.once('delivery-report', function(report) { + clearInterval(tt); + offset = report.offset; + }); + + consumer + .subscribe([topic]); + + var ct; + + var consumeOne = function() { + consumer.consume(function(err, message) { + if (err && (err.code === -191 || err.code === -185)) { + ct = setTimeout(consumeOne, 100); + return; + } + + clearInterval(tt); + + if (err) { + return cb(err); + } + + try { + t.equal(buffer.toString(), message.message.toString(), + 'message is not equal to buffer'); + } catch (e) { + return cb(e); + } + disconnect(); + }); + }; + + consumeOne(); + + // Consume until we get it or time out + + }); + } + }; + + producer = new Kafka.Producer({ + 'client.id': 'kafka-mocha', + 'metadata.broker.list': kafkaBrokerList, + 'fetch.wait.max.ms': 1, + + 'dr_cb': true + }); + + producer.connect({}, function(err, d) { + try { + t.ifError(err); + t.equal(typeof d, 'object', 'metadata should be returned'); + } catch (e) { + return cb(e); + } + producerConnected = true; + maybeDone(); + }); + + consumer = new Kafka.KafkaConsumer({ + 'metadata.broker.list': kafkaBrokerList, + 'group.id': grp, + 'fetch.wait.max.ms': 1000, + 'session.timeout.ms': 10000, + 'enable.auto.commit': true + // paused: true, + }, { + 'auto.offset.reset': 'largest' + }); + + consumer.connect({}, function(err, d) { + t.ifError(err); + t.equal(typeof d, 'object', 'metadata should be returned'); + consumerConnected = true; + maybeDone(); + }); + + }); + +}); + +testCase.run(function(err) { + if (!err) { + console.log('All tests passed successfully'); + } + + process.exit(); +}); diff --git a/e2e/consumer.spec.js b/e2e/consumer.spec.js new file mode 100644 index 00000000..cc0567c1 --- /dev/null +++ b/e2e/consumer.spec.js @@ -0,0 +1,60 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +var t = require('assert'); +var crypto = require('crypto'); + +var KafkaConsumer = require('../').KafkaConsumer; + +var kafkaBrokerList = process.env.KAFKA_HOST || 'localhost:9092'; + +var grp = 'kafka-mocha-grp-' + crypto.randomBytes(20).toString(); + +var TestCase = require('./test-case'); + +var testCase = new TestCase('Consumer tests', function() { + this.test('properly rebalances', function(cb) { + var consumer; + + consumer = new KafkaConsumer({ + 'group.id': grp, + 'metadata.broker.list': kafkaBrokerList, + 'fetch.wait.max.ms': 1, + 'session.timeout.ms': 2000, + }); + consumer.connect({}, function(err, metadata) { + try { + t.ifError(err); + t.equal(typeof metadata, 'object', 'metadata should be returned'); + + // Ensure it is in the correct format + t.ok(metadata.orig_broker_name, 'Broker name is not set'); + t.ok(metadata.orig_broker_id, 'Broker id is not set'); + t.equal(Array.isArray(metadata.brokers), true); + t.equal(Array.isArray(metadata.topics), true); + } catch (e) { + return cb(e); + } + + consumer.disconnect(function(err) { + cb(); + }); + }); + }); +}); + +testCase.run(function(err) { + if (err) { + return process.exit(1); + } + + console.log('All tests passed successfully'); + + process.exit(0); + +}); diff --git a/e2e/producer.spec.js b/e2e/producer.spec.js new file mode 100644 index 00000000..a2e8c34e --- /dev/null +++ b/e2e/producer.spec.js @@ -0,0 +1,156 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +var Kafka = require('../'); +var t = require('assert'); +var crypto = require('crypto'); + +var TestCase = require('./test-case'); + +var kafkaBrokerList = process.env.KAFKA_HOST || 'localhost:9092'; + +var serviceStopped = false; + +var topic = 'test'; // + crypto.randomBytes(20).toString(); +var producer; + +var testCase = new TestCase('Producer tests', function() { + + this.test('Can connect to Kafka', function(cb) { + function producerConnected() { + producer.getMetadata({}, function(err, metadata) { + try { + t.ifError(err); + t.ok(metadata); + + // Ensure it is in the correct format + t.ok(metadata.orig_broker_name, 'Broker name is not set'); + t.ok(metadata.orig_broker_id, 'Broker id is not set'); + t.equal(Array.isArray(metadata.brokers), true); + t.equal(Array.isArray(metadata.topics), true); + + } catch (err) { + cb(err); + return; + } + + producer.disconnect(function() { + producer = null; + cb(); + }); + }); + } + + + producer = new Kafka.Producer({ + 'client.id': 'kafka-test', + 'metadata.broker.list': kafkaBrokerList, + 'dr_cb': true + }); + producer.connect({}, function(err) { + if (err) { + cb(err); return; + } + + producerConnected(); + }); + }); + + this.test('Producer gets 100% deliverability', function(cb) { + producer = new Kafka.Producer({ + 'client.id': 'kafka-mocha', + 'metadata.broker.list': kafkaBrokerList, + 'dr_cb': true + }); + producer.connect(); + + var total = 0; + var totalSent = 0; + var max = 10000; + var errors = 0; + var started = Date.now(); + + var sendMessage = function() { + if (totalSent < 1) { + // topic = producer.Topic(topic, {}); + } + var ret = producer.produce({ + topic: topic, + message: new Buffer('message ' + total) + }, function(err) { + total++; + totalSent++; + if (err) { + return cb(err); + } else { + if (total >= max) { + } else { + sendMessage(); + } + } + }); + + }; + + var verified_received = 0; + var exitNextTick = false; + var errorsArr = []; + + var tt = setInterval(function() { + producer.poll(); + + if (exitNextTick) { + clearInterval(tt); + if (errors > 0) { + return cb(errorsArr[0]); + } + producer.disconnect(function() { + cb(); + }); + + return; + } + + if (verified_received + errors === max) { + exitNextTick = true; + } + + }, 1000).unref(); + + producer + .on('delivery-report', function(report) { + try { + t.ok(report !== undefined); + t.ok(typeof report.topic_name === 'string'); + t.ok(typeof report.partition === 'number'); + t.ok(typeof report.offset === 'number'); + } catch (e) { + cb(e); + } + verified_received++; + }).on('ready', sendMessage); + }); + +}); + +testCase.run(function(err) { + if (err) { + process.exitCode = 1; + } else { + console.log('All tests passed successfully'); + } + + // let x = process._getActiveHandles(); + // let y = process._getActiveRequests(); + + // console.log(x); + // console.log(y); + process.exit(); + +}); diff --git a/e2e/test-case.js b/e2e/test-case.js new file mode 100644 index 00000000..b96b0223 --- /dev/null +++ b/e2e/test-case.js @@ -0,0 +1,115 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +module.exports = TestCase; + +function TestCase(name, test) { + + this.name = name; + this.test = test; + +} + +TestCase.prototype.execute = function(test, cb) { + var t; + var fakeCb = function(err) { + // Reset it so we don't get the callback executed twice + if (!err) { + err = new Error('Callback executed after failure'); + } + + console.log(err); + }; + + try { + test.bind({ + timeout: function(int) { + if (int) { + t = setTimeout(function() { + cb(new Error('Test timed out')); + cb = fakeCb; + }, int); + } + } + })(function(err) { + clearTimeout(t); + if (err) { + cb(err); + } else { + cb(); + } + }); + } catch (e) { + cb(e); + cb = fakeCb; + } +}; + +TestCase.prototype.run = function(cb) { + + console.log('Running test case: ' + this.name); + + var self = this; + var t; + + var cases = {}; + try { + + + this.test.call({ + test: function(name, testCase) { + cases[name] = { + case: testCase + }; + } + }); + + var names = Object.keys(cases); + var counter = -1; + + var runNextCase = function(err) { + if (err) { + err.case = names[counter]; + err.suite = self.name; + TestCase.printError(err); + process.exitCode = 1; + return cb(err); + } + counter++; + + if (counter > 0) { + TestCase.printSuccess({ suite: self.name, case: names[counter-1] }); + } + + var testCase = cases[names[counter]]; + + if (!testCase) { + // Done. + cb(); + } else { + self.execute(testCase.case, runNextCase); + } + }; + + runNextCase(); + + } catch (err) { + cb(err); + } +}; + +TestCase.printError = function(err) { + // First log the test case + console.error(err.suite + ': ' + err.case + ' - ERROR! ' + err.message); + console.error(err.stack); +}; + +TestCase.printSuccess = function(obj) { + console.info(obj.suite + ': ' + obj.case + ' - passed'); +}; diff --git a/examples/consumer.md b/examples/consumer.md new file mode 100644 index 00000000..80dabaa7 --- /dev/null +++ b/examples/consumer.md @@ -0,0 +1,56 @@ +Connecting to a Kafka Consumer is easy. Let's try to connect to one using +the Stream implementation + +```js +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +var Transform = require('stream').Transform; + +var Kafka = require('../'); + +var consumer = new Kafka.KafkaConsumer({ + 'metadata.broker.list': 'localhost:9092', + 'group.id': 'librd-test', + 'socket.keepalive.enable': true, + 'enable.auto.commit': false +}); + +var stream = consumer.getReadStream('test', { + waitInterval: 0 +}); + +stream.on('error', function() { + process.exit(1); +}); + +consumer.on('rebalance', function(e) { + + if (e.code === Kafka.CODES.REBALANCE.PARTITION_ASSIGNMENT) { + console.log('Partition assignment'); + } else { + console.log('Partition unassignment'); + } +}); + +stream + .pipe(new Transform({ + objectMode: true, + transform: function(data, encoding, callback) { + // consumer.commit(data, function(err) {}); + // We want to force it to run async + callback(null, data.message); + } + })) + .pipe(process.stdout); + +consumer.on('error', function(err) { + console.log(err); +}); +``` diff --git a/examples/metadata.md b/examples/metadata.md new file mode 100644 index 00000000..391a1ad1 --- /dev/null +++ b/examples/metadata.md @@ -0,0 +1,27 @@ +```js +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +var Kafka = require('../'); + +var producer = new Kafka.Producer({ + 'metadata.broker.list': 'localhost:9092', + 'client.id': 'hey', + 'compression.codec': 'snappy' +}); + +producer.connect() + .on('ready', function(i) { + console.log(i); + console.log(producer.getMetadata()); + }) + .on('error', function(err) { + console.log(err); + }); +``` diff --git a/examples/producer-cluster.md b/examples/producer-cluster.md new file mode 100644 index 00000000..3debbfb8 --- /dev/null +++ b/examples/producer-cluster.md @@ -0,0 +1,101 @@ +```js +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +var cluster = require('cluster'); +var numCPUs = 6; +var Kafka = require('../'); + +if (cluster.isMaster) { + // Fork workers. + for (var i = 0; i < numCPUs; i++) { + cluster.fork(); + } + + var exited_workers = 0; + + cluster.on('exit', function(worker, code, signal) { + exited_workers++; + if (exited_workers === numCPUs - 1) { + process.exit(); + } + }); +} else { + // Configure client + var producer = new Kafka.Producer({ + 'client.id': 'kafka', + 'metadata.broker.list': 'localhost:9092', + 'compression.codec': 'none', + 'retry.backoff.ms': 200, + 'message.send.max.retries': 10, + 'socket.keepalive.enable': true, + 'queue.buffering.max.messages': 100000, + 'queue.buffering.max.ms': 1000, + 'batch.num.messages': 1000000, + 'dr_cb': true + }); + + var total = 0; + var totalSent = 0; + var max = 20000; + var errors = 0; + var started = Date.now(); + + var sendMessage = function() { + var ret = producer.sendMessage({ + topic: 'librdtesting-01', + message: new Buffer('message ' + total) + }, function() { + }); + total++; + if (total >= max) { + } else { + setImmediate(sendMessage); + } + }; + + var verified_received = 0; + var exitNextTick = false; + var errorsArr = []; + + var t = setInterval(function() { + producer.poll(); + + if (exitNextTick) { + clearInterval(t); + return setTimeout(function() { + console.log('[%d] Received: %d, Errors: %d, Total: %d', process.pid, verified_received, errors, total); + // console.log('[%d] Finished sending %d in %d seconds', process.pid, total, parseInt((Date.now() - started) / 1000)); + if (errors > 0) { + console.error(errorsArr[0]); + return process.exitCode = 1; + } + process.exitCode = 0; + setTimeout(process.exit, 1000); + }, 2000); + } + + if (verified_received + errors === max) { + exitNextTick = true; + } + + }, 1000); + producer.connect() + .on('error', function(e) { + errors++; + errorsArr.push(e); + }) + .on('delivery-report', function() { + verified_received++; + }) + .on('ready', sendMessage); + + +} +``` diff --git a/examples/producer-multiple.md b/examples/producer-multiple.md new file mode 100644 index 00000000..9faf278d --- /dev/null +++ b/examples/producer-multiple.md @@ -0,0 +1,37 @@ +```js +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +var Kafka = require('../'); + +var producer = Kafka.Producer({}); + +function multiWrite(buffer, streams, cb) { + var length = streams.length; + var completed = 0; + + var done = function() { + completed++; + if (completed === length) { + cb(); + } + }; + + streams.forEach(function(stream) { + stream.write(buffer, null, done); + }); +} + +var topic1 = producer.getWriteStream('topic1'); +var topic2 = producer.getWriteStream('topic2'); + +multiWrite(new Buffer('test'), [topic1, topic2], function() { + console.log('all done'); +}); +``` diff --git a/examples/producer.md b/examples/producer.md new file mode 100644 index 00000000..676aebf8 --- /dev/null +++ b/examples/producer.md @@ -0,0 +1,85 @@ +```js +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +var Kafka = require('../'); + + var producer = new Kafka.Producer({ + 'client.id': 'kafka', + 'metadata.broker.list': 'localhost:9092', + 'compression.codec': 'gzip', + 'retry.backoff.ms': 200, + 'message.send.max.retries': 10, + 'socket.keepalive.enable': true, + 'queue.buffering.max.messages': 100000, + 'queue.buffering.max.ms': 1000, + 'batch.num.messages': 1000000, + 'dr_cb': true + }); + + var total = 0; + var totalSent = 0; + var max = 200000; + var errors = 0; + var started = Date.now(); + + var sendMessage = function() { + var ret = producer.sendMessage({ + topic: 'test', + message: new Buffer('message ' + total) + }, function() {}); + total++; + if (total >= max) { + console.log('Done sending'); + } else { + setImmediate(sendMessage); + } + }; + + var verified_received = 0; + var exitNextTick = false; + var errorsArr = []; + + var t = setInterval(function() { + producer.poll(); + + if (exitNextTick) { + clearInterval(t); + return setTimeout(function() { + console.log('[%d] Received: %d, Errors: %d, Total: %d', process.pid, verified_received, errors, total); + console.log('[%d] Finished sending %d in %d seconds', process.pid, total, parseInt((Date.now() - started) / 1000)); + if (errors > 0) { + console.error(errorsArr[0]); + process.exitCode = 1; + } + producer.disconnect(function() { + process.exit(process.exitCode); + }); + }, 2000); + } + + if (verified_received + errors === max) { + exitNextTick = true; + } + + }, 1000); + producer.connect({}, function(err) { + if (err) { + return process.exit(1); + } + }) + .on('error', function(e) { + errors++; + errorsArr.push(e); + }) + .on('delivery-report', function() { + verified_received++; + }) + .on('ready', sendMessage); +``` diff --git a/index.js b/index.js new file mode 100644 index 00000000..d9993708 --- /dev/null +++ b/index.js @@ -0,0 +1,12 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +var Kafka; + +module.exports = Kafka = require('./lib'); diff --git a/kafka-native.js b/kafka-native.js new file mode 100644 index 00000000..4198da07 --- /dev/null +++ b/kafka-native.js @@ -0,0 +1,12 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +var kafka = require('bindings')('node-librdkafka'); + +module.exports = kafka; diff --git a/lib/client.js b/lib/client.js new file mode 100644 index 00000000..9fda5a62 --- /dev/null +++ b/lib/client.js @@ -0,0 +1,362 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +module.exports = Client; + +var Emitter = require('events').EventEmitter; +var util = require('util'); +var Kafka = require('../kafka-native.js'); +var assert = require('assert'); + +var LibrdKafkaError = require('./error'); + +util.inherits(Client, Emitter); + +/** + * Base class for Consumer and Producer + * + * This should not be created independently, but rather is + * the base class on which both producer and consumer + * get their common functionality. + * + * @param {object} globalConf - Global configuration in key value pairs. + * @param {function} SubClientType - The function representing the subclient + * type. In C++ land this needs to be a class that inherits from Connection. + * @param {object} topicConf - Topic configuration in key value pairs + * @constructor + * @extends Emitter + */ +function Client(globalConf, SubClientType, topicConf) { + if (!(this instanceof Client)) { + return new Client(globalConf, SubClientType, topicConf); + } + + Emitter.call(this); + + // This superclass must be initialized with the Kafka.{Producer,Consumer} + // @example var client = new Client({}, Kafka.Producer); + // remember this is a superclass so this will get taken care of in + // the producer and consumer main wrappers + + var no_event_cb = globalConf.event_cb === false; + topicConf = topicConf || {}; + + // delete this because librdkafka will complain since this particular + // key is a real conf value + delete globalConf.event_cb; + + // Enforce a default timeout + if (!globalConf.hasOwnProperty('socket.timeout.ms')) { + globalConf['socket.timeout.ms'] = 250; + } + + this._client = new SubClientType(globalConf, topicConf); + + // primitive callbacks from c++ wrapper + // need binds on them if we want to broadcast the event via the emitter + + // However, the C++ wrapper lazy processes callbacks. + // If we want to improve performance, removing this line is a good way to do it. + // we need to do this on creation of the object + // thats why we did some processing on config + + // Self required because we are inside an async callback function scope + if (!no_event_cb) { + this._client.onEvent(function eventHandler(eventType, eventData) { + switch (eventType) { + case 'error': + self.emit('error', LibrdKafkaError.create(eventData)); + break; + case 'stats': + self.emit('event.stats', eventData); + break; + case 'log': + self.emit('event.log', eventData); + break; + default: + self.emit('event.event', eventData); + self.emit('event.' + eventType, eventData); + } + }); + } + + this.metrics = {}; + this._isConnected = false; + this.errorCounter = 0; + + /** + * Metadata object. Starts out empty but will be filled with information after + * the initial connect. + * + * @type {Client~Metadata} + */ + this._metadata = {}; + + var self = this; + + this.on('ready', function(info) { + self.metrics.connectionOpened = Date.now(); + self._isConnected = true; + self.name = info.name; + }) + .on('disconnected', function() { + // reset metrics + self.metrics = {}; + self._isConnected = false; + // keep the metadata. it still may be useful + }) + .on('error', function(err) { + self.lastError = err; + ++self.errorCounter; + }); + +} + +/** + * Connect to the broker and receive its metadata. + * + * Connects to a broker by establishing the client and fetches its metadata. + * + * @param {object} metadataOptions - Options to be sent to the metadata. + * @param {Client~connectionCallback} cb - Callback that indicates we are + * done connecting. + * @return {Client} - Returns itself. + */ +Client.prototype.connect = function(metadataOptions, cb) { + var self = this; + + var next = function(err, data) { + self._isConnecting = false; + if (cb) { + cb(err, data); + } + }; + + if (this._isConnected) { + setImmediate(next); + return self; + } + + if (this._isConnecting) { + this.once('ready', function() { + next(null, this._metadata); + }); + return self; + } + + this._isConnecting = true; + + var fail = function(err, emit) { + if (emit) { + self.emit('error', LibrdKafkaError.create(err)); + } + self._isConnected = false; + next(err); return; + }; + + this._client.connect(function(err, info) { + if (err) { + fail(err, true); return; + } + + self._isConnected = true; + + // Otherwise we are successful + self.getMetadata(metadataOptions || {}, function(err, metadata) { + if (err) { + fail(LibrdKafkaError.create(err)); return; + } + + self._isConnecting = false; + // We got the metadata otherwise. It is set according to above param + + /** + * Ready event. Called when the Client connects successfully + * + * @event Client#ready + * @type {object} + * @property {string} name - the name of the broker. + */ + self.emit('ready', info); + next(null, metadata); return; + + }); + + }); + + return self; + +}; + +/** + * Get the native Kafka client. + * + * You probably shouldn't use this, but if you want to execute methods directly + * on the c++ wrapper you can do it here. + * + * @see connection.cc + * @return {Conection} - The native Kafka client. + */ +Client.prototype.getClient = function() { + return this._client; +}; + +/** + * Find out how long we have been connected to Kafka. + * + * @return {number} - Milliseconds since the connection has been established. + */ +Client.prototype.connectedTime = function() { + if (!this.isConnected()) { + return 0; + } + return Date.now() - this.metrics.connectionOpened(); +}; + +/** + * Whether or not we are connected to Kafka. + * + * @return {boolean} - Whether we are connected. + */ +Client.prototype.isConnected = function() { + return this._isConnected && this._client; +}; + +/** + * Get the last error emitted if it exists. + * + * @return {LibrdKafkaError} - Returns the LibrdKafkaError or null if + * one hasn't been thrown. + */ +Client.prototype.getLastError = function() { + return this.lastError || null; +}; + +/** + * Disconnect from the Kafka client. + * + * This method will disconnect us from Kafka unless we are already in a + * disconnecting state. Use this when you're done reading or producing messages + * on a given client. + * + * It will also emit the disconnected event. + * + * @fires Client#disconnected + * @return {function} - Callback to call when disconnection is complete. + */ +Client.prototype.disconnect = function(cb) { + var self = this; + + if (!this._isDisconnecting && this._client) { + this._isDisconnecting = true; + this._client.disconnect(function() { + // this take 5000 milliseconds. Librdkafka needs to make sure the memory + // has been cleaned up before we delete things. @see RdKafka::wait_destroyed + + // Broadcast metrics. Gives people one last chance to do something with them + self._isDisconnecting = false; + /** + * Disconnect event. Called after disconnection is finished. + * + * @event Client#disconnected + * @type {object} + * @property {date} connectionOpened - when the connection was opened. + */ + self.emit('disconnected', self.metrics); + if (cb) { + cb(self.metrics); + } + + }); + + } + + return self; +}; + +/** + * Get client metadata. + * + * @param {object} metadataOptions - Metadata options to pass to the client. + * @param {string} metadataOptions.topic - Topic string for which to fetch + * metadata + * @param {number} metadataOptions.timeout - Max time, in ms, to try to fetch + * metadata before timing out. Defaults to 3000. + * @param {Client~metadataCallback} cb - Callback to fire with the metadata. + */ +Client.prototype.getMetadata = function(metadataOptions, cb) { + if (!this.isConnected()) { + return cb(new Error('Client is disconnected')); + } + + var self = this; + + this._client.getMetadata(metadataOptions || {}, function(err, metadata) { + if (err) { + self.emit('error', LibrdKafkaError.create(err)); + if (cb) { + cb(err); + } + return; + } + + // No error otherwise + self._metadata = metadata; + + if (cb) { + cb(null, metadata); + } + + }); + +}; + +/** + * This callback is used to pass metadata or an error after a successful + * connection + * + * @callback Client~connectionCallback + * @param {Error} err - An error, if one occurred while connecting. + * @param {Client~Metadata} metadata - Metadata object. + */ + +/** + * @typedef {object} Client~MetadataBroker + * @property {number} id - Broker ID + * @property {string} host - Broker host + * @property {number} port - Broker port. + */ + +/** + * @typedef {object} Client~MetadataTopic + * @property {string} name - Topic name + * @property {Client~MetadataPartition[]} partitions - Array of partitions + */ + +/** + * @typedef {object} Client~MetadataPartition + * @property {number} id - Partition id + * @property {number} leader - Broker ID for the partition leader + * @property {number[]} replicas - Array of replica IDs + * @property {number[]} isrs - Arrqay of ISRS ids +*/ + +/** + * Metadata object. + * + * This is the representation of Kafka metadata in JavaScript. + * + * @typedef {object} Client~Metadata + * @property {number} orig_broker_id - The broker ID of the original bootstrap + * broker. + * @property {string} orig_broker_name - The name of the original bootstrap + * broker. + * @property {Client~MetadataBroker[]} brokers - An array of broker objects + * @property {Client~MetadataTopic[]} topics - An array of topics. + */ diff --git a/lib/error.js b/lib/error.js new file mode 100644 index 00000000..833f1dab --- /dev/null +++ b/lib/error.js @@ -0,0 +1,233 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +module.exports = LibrdKafkaError; + +var util = require('util'); + +util.inherits(LibrdKafkaError, Error); + +LibrdKafkaError.create = createLibrdkafkaError; + +/** + * Enum for identifying errors reported by the library + * + * You can find this list in the C++ code at + * https://github.com/edenhill/librdkafka/blob/master/src-cpp/rdkafkacpp.h#L148 + * + * @readonly + * @enum {number} + * @constant + */ +LibrdKafkaError.codes = { + + /** Begin internal error codes */ + ERR__BEGIN: -200, + /** Received message is incorrect */ + ERR__BAD_MSG: -199, + /** Bad/unknown compression */ + ERR__BAD_COMPRESSION: -198, + /** Broker is going away */ + ERR__DESTROY: -197, + /** Generic failure */ + ERR__FAIL: -196, + /** Broker transport failure */ + ERR__TRANSPORT: -195, + /** Critical system resource */ + ERR__CRIT_SYS_RESOURCE: -194, + /** Failed to resolve broker */ + ERR__RESOLVE: -193, + /** Produced message timed out */ + ERR__MSG_TIMED_OUT: -192, + /** Reached the end of the topic+partition queue on the broker. Not really an error. */ + ERR__PARTITION_EOF: -191, + /** Permanent: Partition does not exist in cluster. */ + ERR__UNKNOWN_PARTITION: -190, + /** File or filesystem error */ + ERR__FS: -189, + /** Permanent: Topic does not exist in cluster. */ + ERR__UNKNOWN_TOPIC: -188, + /** All broker connections are down. */ + ERR__ALL_BROKERS_DOWN: -187, + /** Invalid argument, or invalid configuration */ + ERR__INVALID_ARG: -186, + /** Operation timed out */ + ERR__TIMED_OUT: -185, + /** Queue is full */ + ERR__QUEUE_FULL: -184, + /** ISR count < required.acks */ + ERR__ISR_INSUFF: -183, + /** Broker node update */ + ERR__NODE_UPDATE: -182, + /** SSL error */ + ERR__SSL: -181, + /** Waiting for coordinator to become available. */ + ERR__WAIT_COORD: -180, + /** Unknown client group */ + ERR__UNKNOWN_GROUP: -179, + /** Operation in progress */ + ERR__IN_PROGRESS: -178, + /** Previous operation in progress, wait for it to finish. */ + ERR__PREV_IN_PROGRESS: -177, + /** This operation would interfere with an existing subscription */ + ERR__EXISTING_SUBSCRIPTION: -176, + /** Assigned partitions (rebalance_cb) */ + ERR__ASSIGN_PARTITIONS: -175, + /** Revoked partitions (rebalance_cb) */ + ERR__REVOKE_PARTITIONS: -174, + /** Conflicting use */ + ERR__CONFLICT: -173, + /** Wrong state */ + ERR__STATE: -172, + /** Unknown protocol */ + ERR__UNKNOWN_PROTOCOL: -171, + /** Not implemented */ + ERR__NOT_IMPLEMENTED: -170, + /** Authentication failure */ + ERR__AUTHENTICATION: -169, + /** No stored offset */ + ERR__NO_OFFSET: -168, + /** Outdated */ + ERR__OUTDATED: -167, + /** End internal error codes */ + ERR__END: -100, + + /** Unknown broker error */ + ERR_UNKNOWN: -1, + /** Success */ + ERR_NO_ERROR: 0, + /** Offset out of range */ + ERR_OFFSET_OUT_OF_RANGE: 1, + /** Invalid message */ + ERR_INVALID_MSG: 2, + /** Unknown topic or partition */ + ERR_UNKNOWN_TOPIC_OR_PART: 3, + /** Invalid message size */ + ERR_INVALID_MSG_SIZE: 4, + /** Leader not available */ + ERR_LEADER_NOT_AVAILABLE: 5, + /** Not leader for partition */ + ERR_NOT_LEADER_FOR_PARTITION: 6, + /** Request timed out */ + ERR_REQUEST_TIMED_OUT: 7, + /** Broker not available */ + ERR_BROKER_NOT_AVAILABLE: 8, + /** Replica not available */ + ERR_REPLICA_NOT_AVAILABLE: 9, + /** Message size too large */ + ERR_MSG_SIZE_TOO_LARGE: 10, + /** StaleControllerEpochCode */ + ERR_STALE_CTRL_EPOCH: 11, + /** Offset metadata string too large */ + ERR_OFFSET_METADATA_TOO_LARGE: 12, + /** Broker disconnected before response received */ + ERR_NETWORK_EXCEPTION: 13, + /** Group coordinator load in progress */ + ERR_GROUP_LOAD_IN_PROGRESS: 14, + /** Group coordinator not available */ + ERR_GROUP_COORDINATOR_NOT_AVAILABLE: 15, + /** Not coordinator for group */ + ERR_NOT_COORDINATOR_FOR_GROUP: 16, + /** Invalid topic */ + ERR_TOPIC_EXCEPTION: 17, + /** Message batch larger than configured server segment size */ + ERR_RECORD_LIST_TOO_LARGE: 18, + /** Not enough in-sync replicas */ + ERR_NOT_ENOUGH_REPLICAS: 19, + /** Message(s) written to insufficient number of in-sync replicas */ + ERR_NOT_ENOUGH_REPLICAS_AFTER_APPEND: 20, + /** Invalid required acks value */ + ERR_INVALID_REQUIRED_ACKS: 21, + /** Specified group generation id is not valid */ + ERR_ILLEGAL_GENERATION: 22, + /** Inconsistent group protocol */ + ERR_INCONSISTENT_GROUP_PROTOCOL: 23, + /** Invalid group.id */ + ERR_INVALID_GROUP_ID: 24, + /** Unknown member */ + ERR_UNKNOWN_MEMBER_ID: 25, + /** Invalid session timeout */ + ERR_INVALID_SESSION_TIMEOUT: 26, + /** Group rebalance in progress */ + ERR_REBALANCE_IN_PROGRESS: 27, + /** Commit offset data size is not valid */ + ERR_INVALID_COMMIT_OFFSET_SIZE: 28, + /** Topic authorization failed */ + ERR_TOPIC_AUTHORIZATION_FAILED: 29, + /** Group authorization failed */ + ERR_GROUP_AUTHORIZATION_FAILED: 30, + /** Cluster authorization failed */ + ERR_CLUSTER_AUTHORIZATION_FAILED: 31 +}; + +/** + * Representation of a librdkafka error + * + * This can be created by giving either another error + * to piggy-back on. In this situation it tries to parse + * the error string to figure out the intent. However, more usually, + * it is constructed by an error object created by a C++ Baton. + * + * @param {object|error} e - An object or error to wrap + * @property {string} message - The error message + * @property {number} code - The error code. + * @property {string} origin - The origin, whether it is local or remote + * @constructor + */ +function LibrdKafkaError(e) { + if (!(this instanceof LibrdKafkaError)) { + return new LibrdKafkaError(e); + } + + // This is the better way + if (!util.isError(e)) { + this.message = e.message; + this.code = e.code; + this.errno = e.code; + if (e.code >= LibrdKafkaError.codes.ERR__END) { + this.origin = 'local'; + } else { + this.origin = 'kafka'; + } + Error.captureStackTrace(this, this.constructor); + } else { + var message = e.message; + var parsedMessage = message.split(': '); + + var origin, msg; + + if (parsedMessage.length > 1) { + origin = parsedMessage[0].toLowerCase(); + msg = parsedMessage[1].toLowerCase(); + } else { + origin = 'unknown'; + msg = message.toLowerCase(); + } + + // special cases + if (msg === 'consumer is disconnected' || msg === 'producer is disconnected') { + this.origin = 'local'; + this.code = LibrdKafkaError.codes.ERR__STATE; + this.errno = this.code; + this.message = msg; + } else { + this.origin = origin; + this.message = msg; + this.code = -1; + this.errno = this.code; + this.stack = e.stack; + } + + } + +} + +function createLibrdkafkaError(e) { + return new LibrdKafkaError(e); +} diff --git a/lib/index.js b/lib/index.js new file mode 100644 index 00000000..a73cffe7 --- /dev/null +++ b/lib/index.js @@ -0,0 +1,26 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +var KafkaConsumer = require('./kafka-consumer'); +var Producer = require('./producer'); +var error = require('./error'); +var util = require('util'); + +module.exports = { + Consumer: util.deprecate('Use KafkaConsumer instead. This may be changed in a later version', KafkaConsumer), + Producer: Producer, + KafkaConsumer: KafkaConsumer, + CODES: { + REBALANCE: { + PARTITION_ASSIGNMENT: 500, + PARTITION_UNASSIGNMENT: 501 + }, + ERRORS: error.codes + } +}; diff --git a/lib/kafka-consumer.js b/lib/kafka-consumer.js new file mode 100644 index 00000000..fc884dcc --- /dev/null +++ b/lib/kafka-consumer.js @@ -0,0 +1,455 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ +'use strict'; + +module.exports = KafkaConsumer; + +var Client = require('./client'); +var util = require('util'); +var Kafka = require('../kafka-native.js'); +var TopicReadable = require('./util/topicReadable'); +var LibrdKafkaError = require('./error'); + +util.inherits(KafkaConsumer, Client); + +/** + * KafkaConsumer class for reading messages from Kafka + * + * This is the main entry point for reading data from Kafka. You + * configure this like you do any other client, with a global + * configuration and default topic configuration. + * + * Once you instantiate this object, connecting will open a socket. + * Data will not be read until you tell the consumer what topics + * you want to read from. + * + * @param {object} conf - Key value pairs to configure the consumer + * @param {object} topicConf - Key value pairs to create a default + * topic configuration + * @extends Client + * @constructor + */ +function KafkaConsumer(conf, topicConf) { + if (!(this instanceof KafkaConsumer)) { + return new KafkaConsumer(conf, topicConf); + } + + /** + * KafkaConsumer message. + * + * This is the representation of a message read from Kafka. + * + * @typedef {object} KafkaConsumer~Message + * @property {buffer} message - the message buffer from Kafka. + * @property {string} topic - the topic name + * @property {number} partition - the partition on the topic the + * message was on + * @property {number} offset - the offset of the message + * @property {string} key - the message key + * @property {number} size - message size, in bytes. + */ + + Client.call(this, conf, Kafka.KafkaConsumer, topicConf); + + var self = this; + + /** + * Rebalance event. Called when the KafkaConsumer is rebalancing. + * + * @event KafkaConsumer#rebalance + * @type {object} + * @property {number} code - whether the rebalance was an assignment or + * an unassignment + */ + this._client.onRebalance(function(e) { + self.emit('rebalance', e); + }); + + this.globalConfig = conf; + this.topicConfig = topicConf; +} + +/** + * Get a stream representation of this KafkaConsumer + * + * @see TopicReadable + * @example + * var consumer = new Kafka.KafkaConsumer({ + * 'metadata.broker.list': 'localhost:9092', + * 'group.id': 'librd-test', + * 'socket.keepalive.enable': true, + * 'enable.auto.commit': false + * }); + * + * var stream = consumer.getReadStream('test'); + * @param {array} topics - An array of topics to subscribe to. + * @param {object} options - Configuration for the stream. + * @return {TopicReadable} - Readable stream that receives messages + * when new ones become available. + */ +KafkaConsumer.prototype.getReadStream = function(topics, options) { + return new TopicReadable(this, topics, options); +}; + + +/** + * Assign the consumer specific partitions and topics + * + * @param {array} assignments - Assignments array. Should contain + * objects with topic and partition set. + * @return {Client} - Returns itself + */ + +KafkaConsumer.prototype.assign = function(assignments) { + this._client.assign(assignments); + return this; +}; + +/** + * Subscribe to an array of topics synchronously. + * + * This operation is pretty fast because it just sets + * an assignment in librdkafka. This is the recommended + * way to deal with subscriptions in a situation where you + * will be reading across multiple files or as part of + * your configure-time initialization. + * + * This is also a good way to do it for streams. + * + * @param {array} topics - An array of topics to listen to + * @return {KafkaConsumer} - Returns itself. + */ +KafkaConsumer.prototype.subscribe = function(topics) { + this._client.subscribeSync(topics); + return this; +}; + +/** + * Unsubscribe from all currently subscribed topics + * + * Before you subscribe to new topics you need to unsubscribe + * from the old ones, if there is an active subscription. + * Otherwise, you will get an error because there is an + * existing subscription. + * + * @return {KafkaConsumer} - Returns itself. + */ +KafkaConsumer.prototype.unsubscribe = function(cb) { + if (!cb) { + cb = function() {}; + } + return this._client.unsubscribe(cb); +}; + +/** + * Read a single message from Kafka. + * + * This method reads one message from Kafka. It emits that + * data through the emitting interface, so make sure not + * to call this method and operate on both. + * + * Some errors are more innocuous than others, depending on + * your situation. This method does not filter any errors + * reported by Librdkafka. Any error, including ERR__PARTITION_EOF + * will be reported down the pipeline. It is your responsibility + * to make sure they are "handled" + * + * @param {KafkaConsumer~readCallback} cb - Callback to return when the work is + * done + * @return {KafkaConsumer} - Returns itself + *//** + * Read a number of messages from Kafka. + * + * This method is similar to the main one, except that it reads a number + * of messages before calling back. This may get better performance than + * reading a single message each time in stream implementations. + * + * This will keep going until it gets ERR__PARTITION_EOF or ERR__TIMED_OUT + * so the array may not be the same size you ask for. The size is advisory, + * but we will not exceed it. + * + * @param {number} size - Number of messages to read + * @param {KafkaConsumer~readCallback} cb - Callback to return when work is done. + *//** + * Read messages from Kafka as fast as possible + * + * This method keeps a background thread running to fetch the messages + * as quickly as it can, sleeping only in between EOF and broker timeouts. + * + * Use this to get the maximum read performance if you don't care about the + * stream backpressure. + * @param {Array|string} topics - Array of topics or a topic string to + * subscribe to. + * @param {KafkaConsumer~readCallback} cb - Callback to return when a message + * is fetched. + */ +KafkaConsumer.prototype.consume = function(topics, cb) { + var self = this; + // If topics is set and is an array or a string, or if we have both + // parameters, run it like this. + if ((topics && (Array.isArray(topics) || typeof topics === 'string'))) { + + if (typeof topics === 'string') { + topics = [topics]; + } else if (!Array.isArray(topics)) { + throw new TypeError('"topics" must be an array or a string'); + } + + if (cb === undefined) { + cb = function() {}; + } + + this._consumeLoop(topics, cb); + + } else if ((topics && typeof topics === 'number') || (topics && cb)) { + + var numMessages = topics; + if (cb === undefined) { + cb = function() {}; + } else if (typeof cb !== 'function') { + throw new TypeError('Callback must be a function'); + } + + this._consumeNum(numMessages, cb); + + } else { + if (topics === undefined) { + cb = function() {}; + } else if (typeof topics !== 'function') { + throw new TypeError('Callback must be a function'); + } else { + cb = topics; + } + + this._consumeOne(cb); + } + return this; +}; + +/** + * Open a background thread and keep getting messages as fast + * as we can. Should not be called directly, and instead should + * be called using consume. + * + * @private + * @see consume + */ +KafkaConsumer.prototype._consumeLoop = function(topics, cb) { + var self = this; + + this._client.subscribe(topics, function subscribeCallback(err) { + if (err) { + err = new LibrdKafkaError(err); + self.emit('error', err); + cb(err); + } else { + + self._client.consumeLoop(function readCallback(err, message) { + + if (err) { + // A few different types of errors here + // but the two we do NOT care about are + // time outs and broker no more messages + // at least now + err = new LibrdKafkaError(err); + if (err.code !== LibrdKafkaError.codes.ERR__TIMED_OUT && + err.code !== LibrdKafkaError.codes.ERR__PARTITION_EOF) { + self.emit('error', err); + } + cb(err); + } else { + /** + * Data event. called whenever a message is received. + * + * @event KafkaConsumer#data + * @type {KafkaConsumer~Message} + */ + self.emit('data', message); + cb(err, message); + } + }); + } + }); +}; + +/** + * Consume a number of messages and wrap in a try catch with + * proper error reporting. Should not be called directly, + * and instead should be called using consume. + * + * @private + * @see consume + */ +KafkaConsumer.prototype._consumeNum = function(numMessages, cb) { + var self = this; + + try { + this._client.consume(numMessages, function(err, messages) { + if (err) { + err = new LibrdKafkaError(err); + if (cb) { + cb(err); + } + self.emit('error', err); + return; + } + + for (var i = 0; i < messages.length; i++) { + self.emit('data', messages[i]); + } + + if (cb) { + cb(null, messages); + } + + }); + } catch (e) { + // This either means we are disconnected, or we were never subscribed. + // This is basically non-recoverable. + var error = new LibrdKafkaError(e); + self.emit('error', error); + + if (cb) { + setImmediate(function() { + cb(error); + }); + } + } +}; + +/** + * Consume a single message and wrap in a try catch with + * proper error reporting. Should not be called directly, + * and instead should be called using consume. + * + * @private + * @see consume + */ +KafkaConsumer.prototype._consumeOne = function(cb) { + // Otherwise, we run this method + var self = this; + + try { + this._client.consume(function(err, message) { + if (err) { + err = new LibrdKafkaError(err); + if (cb) { + cb(err); + } + self.emit('error', err); + return; + } + /** + * Data event. Receives the message just like the callback + * + * @event KafkaConsumer#data + * @type {KafkaConsumer~Message} + */ + self.emit('data', message); + if (cb) { + cb(null, message); + } + }); + } catch (e) { + // This either means we are disconnected, or we were never subscribed. + // This is basically non-recoverable. + var error = new LibrdKafkaError(e); + self.emit('error', error); + + if (cb) { + setImmediate(function() { + cb(error); + }); + } + } +}; + +/** + * Subscribe to an array of topics. + * @param {array} topics - An array of topics to subscribe to. + * @return {KafkaConsumer} - Returns itself + * @deprecated + */ +KafkaConsumer.prototype.subscribeSync = util.deprecate(KafkaConsumer.prototype.consume, + 'subscribeSync: use subscribe instead'); + + /** + * @param {KafkaConsumer~readCallback} cb - Callback to return when the work is + * done + * @return {KafkaConsumer} - Returns itself + * @deprecated + */ +KafkaConsumer.prototype.read = util.deprecate(KafkaConsumer.prototype.consume, + 'this.read: use this.consume instead'); + +/** + * This callback returns the message read from Kafka. + * + * @callback KafkaConsumer~readCallback + * @param {LibrdKafkaError} err - An error, if one occurred while reading + * the data. + * @param {KafkaConsumer~Message} message + */ + +/** + * Commit a message or all messages that have been read + * + * If you provide a message, it will commit that message. Otherwise, + * it will commit all read offsets. + * + * @param {object|function} - Message object to commit or callback + * if you are committing all messages + * @param {function} - Callback to fire when committing is done. + * + * @return {KafkaConsumer} - returns itself. + */ +KafkaConsumer.prototype.commit = function() { + var cb; + var msg = null; + + if (arguments.length === 0) { + cb = function() {}; + } else if (arguments.length === 1) { + cb = arguments[0]; + if (typeof cb !== 'function') { + msg = cb; + cb = function() {}; + } + } else if (arguments.length === 2) { + msg = arguments[0]; + cb = arguments[1]; + } + + if (msg) { + this._client.commit(msg, cb); + } else { + this._client.commit(cb); + } + + return this; +}; + +/** + * Commit a message (or all messages) synchronously + * + * Commit messages or offsets to Kafka so they will not be + * read by other clients. + * + * @see KafkaConsumer#commit + * @param {object} msg - A message object to commit. If none + * is provided, we will commit all read offsets. + * @return {KafkaConsumer} - returns itself. + */ +KafkaConsumer.prototype.commitSync = function(msg) { + if (msg) { + return this._client.commitSync(msg); + } else { + return this._client.commitSync(); + } +}; diff --git a/lib/producer.js b/lib/producer.js new file mode 100644 index 00000000..cc1a76d4 --- /dev/null +++ b/lib/producer.js @@ -0,0 +1,320 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +module.exports = Producer; + +var Client = require('./client'); + +var util = require('util'); +var Kafka = require('../kafka-native.js'); +var TopicWritable = require('./util/topicWritable'); +var LibrdKafkaError = require('./error'); + +util.inherits(Producer, Client); + +/** + * Producer class for sending messages to Kafka + * + * This is the main entry point for writing data to Kafka. You + * configure this like you do any other client, with a global + * configuration and default topic configuration. + * + * Once you instantiate this object, you need to connect to it first. + * This allows you to get the metadata and make sure the connection + * can be made before you depend on it. After that, problems with + * the connection will by brought down by using poll, which automatically + * runs when a transaction is made on the object. + * + * @param {object} conf - Key value pairs to configure the consumer + * @param {object} topicConf - Key value pairs to create a default + * topic configuration + * @extends Client + * @constructor + */ +function Producer(conf, topicConf) { + if (!(this instanceof Producer)) { + return new Producer(conf, topicConf); + } + + /** + * Producer message. This is sent to the wrapper, not received from it + * + * @typedef {object} Producer~Message + * @property {string|buffer} message - The buffer to send to Kafka. + * @property {Topic} topic - The Kafka topic to produce to. + * @property {number} partition - The partition to produce to. Defaults to + * the partitioner + * @property {string} key - The key string to use for the message. + * @see Consumer~Message + */ + + var gTopic = conf.topic || false; + var gPart = conf.partition || null; + var dr_cb = conf.dr_cb || null; + var dr_msg_cb = conf.dr_msg_cb || null; + + // delete keys we don't want to pass on + delete conf.topic; + delete conf.partition; + + delete conf.dr_cb; + delete conf.dr_msg_cb; + + // client is an initialized consumer object + // @see NodeKafka::Consumer::Init + Client.call(this, conf, Kafka.Producer, topicConf); + var self = this; + + // Delete these keys after saving them in vars + this.globalConfig = conf; + this.topicConfig = topicConf; + this.defaultTopic = gTopic || null; + this.defaultPartition = gPart || null; + + this.outstandingMessages = 0; + this.sentMessages = 0; + + this.createdTopics = {}; + + if (dr_msg_cb) { + self.on('delivery-message-report', dr_msg_cb); + } + + if (dr_cb) { + this._client.onDeliveryReport(function onDeliveryReport(err, report) { + if (err) { + self.emit('error', LibrdKafkaError.create(err)); + } else { + self.outstandingMessages--; + self.emit('delivery-report', report); + } + }); + + if (typeof dr_cb === 'function') { + self.on('delivery-report', dr_cb); + } + + } + +} + +/** + * Create a topic if it does not exist. Otherwise, create it + * + * @private + */ +function maybeTopic(name, config) { + // create or get topic + var topic; + + if (typeof(name) === 'string') { + + // If config is an empty object, reuse the ones we have in cache + if (config) { + topic = this.Topic(name, config); + } else { + // if config is an empty object + if (this.createdTopics[name]) { + topic = this.createdTopics[name]; + } else { + topic = this.createdTopics[name] = this.Topic(name, {}); + } + } + + return topic; + } else { + // this may be what we want + return name; + } +}; + +/** + * Create a topic by topic name and config + * + * This is a v8 managed object. It does this to avoid creating + * a topic configuration every time we send a message, which, + * in our case, is tens of thousands of times per second. + * + * Rather than that, since topics generally don't change much, + * we just manage them in memory for however many we are using. + * They are pretty small and it is easy since the name can act + * as a key. + * + * This is a good place to start in an upgrade across the board + * to v4 + * @version 0.12 + * + * @param {string} name - the name of the topic to create + * @param {object} config - a topic configuration. + * @return {Topic} - a new Kafka Topic. + */ +Producer.prototype.Topic = function(name, config) { + if (!this._isConnected) { + throw new Error('Producer not connected'); + } + return new Kafka.Topic(name, config, this._client); +}; + +/** + * Produce a message to Kafka synchronously. + * + * This is the method mainly used in this class. Use it to produce + * a message to Kafka. The first and only parameter of the synchronous + * variant is the message object. + * + * When this is sent off, there is no guarantee it is delivered. If you need + * guaranteed delivery, change your *acks* settings, or use delivery reports. + * + * @param {Producer~Message} msg - The message object. + * @return {error|boolean} - returns an error if it failed, or true if not + * @see Producer#produce + */ +Producer.prototype.produceSync = function(msg) { + if (!this._isConnected) { + throw new Error('Producer not connected'); + } + + this.outstandingMessages++; + this.sentMessages++; + + var topic = msg.topic || false; + var partition = msg.partition || this.defaultPartition; + + if (!topic) { + throw new TypeError('"topic" needs to be set'); + } + + topic = maybeTopic.call(this, topic, msg.config || false); + + var produceOpts = { + topic: topic, + partition: partition, + message: !Buffer.isBuffer(msg.message) ? new Buffer(msg.message) : msg.message, + key: msg.key || null, + opaque: msg.opaque || null + }; + + var ret = this._client.produceSync(produceOpts, topic); + if (util.isError(ret)) { + var err = LibrdKafkaError.create(err); + this.emit('error', err); + return err; + } + + return true; + +}; + +/** + * Alias for produceSync + * + * @see Producer#produceSync + * @deprecated + */ +Producer.prototype.sendMessageSync = util.deprecate(this.produceSync, + 'this.sendMessageSync: use this.produceSync instead'); + +/** + * Produce a message to Kafka. + * + * This is the method mainly used in this class. Use it to produce + * a message to Kafka. + * + * When this is sent off, there is no guarantee it is delivered. If you need + * guaranteed delivery, change your *acks* settings, or use delivery reports. + * The reason this is not guaranteed is because the library does internal + * batching to ensure we write when necessary and not on every call. + * + * @param {Producer~Message} msg - The message object. + * @param {function} cb - The callback to call when the message has entered + * the librdkafka queue. + * @return {Producer} - returns itself. + * @see Producer#produceSync + */ +Producer.prototype.produce = function(msg, cb) { + if (!this._isConnected) { + throw new Error('Producer not connected'); + } + + var self = this; + + this.outstandingMessages++; + this.sentMessages++; + + var topic = msg.topic || false; + + var partition = msg.partition || this.defaultPartition; + + if (!topic) { + throw new TypeError('"topic" needs to be set'); + } + + topic = maybeTopic.call(this, topic, msg.config || false); + + var produceOpts = { + partition: partition, + message: !Buffer.isBuffer(msg.message) ? new Buffer(msg.message) : msg.message, + key: msg.key || null, + opaque: msg.opaque || null + }; + + this._client.produce(produceOpts, topic, function clientProduce(err) { + if (err) { + self.emit('error', LibrdKafkaError.create(err)); + } + if (cb) { + cb(err); return; + } + }); + + return this; + +}; + +/** + * Alias for produce + * + * @see Producer#produce + * @deprecated + */ +Producer.prototype.sendMessage = util.deprecate(Producer.prototype.produce, + 'this.sendMessage: use this.produce instead'); + +/** + * Get a write stream for the given topic. + * + * This stream does not run in object mode. It only takes buffers of data. + * + * @param {string} topic - String topic name. + * @param {object} options - Stream options + * @param {number} options.highWaterMark - Controls how many messages are + * read at a time. + * @return {TopicWritable} - returns the write stream for writing to Kafka. + */ +Producer.prototype.getWriteStream = function(topic, options) { + return new TopicWritable(this, topic, options); +}; + +/** + * Poll for events + * + * We need to run poll in order to learn about new events that have occurred. + * This is done automatically every time we produce a message, but if you'd + * like to do it more because Kafka does not have enough traffic or because + * you need to know at certain points, you can manually poll for changes. + * + * @return {Producer} - returns itself. + */ +Producer.prototype.poll = function() { + if (!this._isConnected) { + throw new Error('Producer not connected'); + } + this._client.poll(); + return this; +}; diff --git a/lib/util/topicReadable.js b/lib/util/topicReadable.js new file mode 100644 index 00000000..663b47fd --- /dev/null +++ b/lib/util/topicReadable.js @@ -0,0 +1,206 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +'use strict'; + +module.exports = TopicReadable; + +var Readable = require('stream').Readable; +var util = require('util'); + +util.inherits(TopicReadable, Readable); + +/** + * ReadableStream integrating with the Kafka Consumer. + * + * This class is used to read data off of Kafka in a streaming way. It is + * useful if you'd like to have a way to pipe Kafka into other systems. You + * should generally not make this class yourself, as it is not even exposed + * as part of module.exports. Instead, you should getReadStream off of a Kafka + * Consumer. + * + * The stream implementation is slower than the continuous subscribe callback. + * If you don't care so much about backpressure and would rather squeeze + * out performance, use that method. Using the stream will ensure you read only + * as fast as you write. + * + * The stream detects if Kafka is already connected. If it is, it will begin + * reading. If it is not, it will connect and read when it is ready. + * + * This stream operates in objectMode. It streams {Consumer~Message} + * + * @param {Consumer} consumer - The Kafka Consumer object. + * @param {array} topics - Array of topics + * @param {object} options - Options to configure the stream. + * @param {number} options.waitInterval - Number of ms to wait if Kafka reports + * that it has timed out or that we are out of messages (right now). + * @constructor + * @extends stream.Readable + * @see Consumer~Message + */ +function TopicReadable(consumer, topics, options) { + if (!(this instanceof TopicReadable)) { + return new TopicReadable(consumer, topics, options); + } + + if (options === undefined) { + options = { waitInterval: 1000 }; + } else if (typeof options === 'number') { + options = { waitInterval: options }; + } else if (options === null || typeof options !== 'object') { + throw new TypeError('"options" argument must be a number or an object'); + } + + if (!Array.isArray(topics)) { + if (typeof topics !== 'string') { + throw new TypeError('"topics" argument must be a string or an array'); + } else { + topics = [topics]; + } + } + + options = Object.create(options); + options.objectMode = true; + + Readable.call(this, options); + + this.consumer = consumer; + this.topics = topics; + this.autoClose = options.autoClose === undefined ? true : !!options.autoClose; + this.waitInterval = options.waitInterval === undefined ? 1000 : options.waitInterval; + this.rebalanced = undefined; + this.connected = false; + this.fetchSize = options.fetchSize || 1; + + var self = this; + + this.consumer + .once('rebalance', function() { + self.rebalanced = true; + }) + .on('disconnected', function() { + self.rebalanced = undefined; + self.connected = false; + }); + + if (!this.consumer.isConnected()) { + this.connect(); + } else { + this.consumer.subscribe(self.topics); + this.connected = true; + } + + this.on('end', function() { + if (this.autoClose) { + this.destroy(); + } + }); + +} + +/** + * Internal stream read method. + * @param {number} size - This parameter is ignored for our cases. + * @private + */ +TopicReadable.prototype._read = function(size) { + if (!this.connected) { + return this.once('connected', function() { + // This is the way Node.js does it + // https://github.com/nodejs/node/blob/master/lib/fs.js#L1733 + this._read(size); + }); + } + + if (this.destroyed) { + return; + } + + var self = this; + + this.consumer.consume(this.fetchSize, onread); + + function onread(err, messages) { + if (err) { + // bad error + if (self.autoClose) { + self.destroy(); + } + self.emit('error', err); + return; + } + + // If there are no messages it means we reached EOF or a timeout. + // Do what we used to do + + if (messages.length < 1) { + + if (!self.waitInterval) { + setImmediate(function() { + self._read(size); + }); + } else { + setTimeout(function() { + self._read(size); + }, self.waitInterval).unref(); + } + + return; + } else if (messages.length === 1) { + self.push(messages[0]); + } else { + self.push(messages); + } + + } +}; + +TopicReadable.prototype.connect = function() { + var self = this; + + self.consumer.connect({}, function(err, metadata) { + if (err) { + self.emit('error', err); + self.destroy(); + return; + } + + self.connected = true; + // Subscribe to the topics as well so we will be ready + self.consumer.subscribe(self.topics); + + self.emit('connected'); + // start the flow of data + self.read(); + }); + +}; + +TopicReadable.prototype.destroy = function() { + if (this.destroyed) { + return; + } + this.destroyed = true; + this.close(); +}; + +TopicReadable.prototype.close = function(cb) { + var self = this; + if (cb) { + this.once('close', cb); + } + + self.connected = false; + close(); + + function close() { + self.consumer.unsubscribe(); + self.emit('closed'); + } +}; diff --git a/lib/util/topicWritable.js b/lib/util/topicWritable.js new file mode 100644 index 00000000..20e52ec1 --- /dev/null +++ b/lib/util/topicWritable.js @@ -0,0 +1,210 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +'use strict'; + +module.exports = TopicWritable; + +var Writable = require('stream').Writable; +var util = require('util'); + +util.inherits(TopicWritable, Writable); + +/** + * Writable stream integrating with the Kafka Producer. + * + * This class is used to write data to Kafka in a streaming way. It takes + * buffers of data and puts them into the appropriate Kafka topic. If you need + * finer control over partitions or keys, this is probably not the class for + * you. In that situation just use the Producer itself. + * + * The stream detects if Kafka is already connected. You can safely begin + * writing right away. + * + * This stream does not operate in Object mode and can only be given buffers. + * + * @param {Producer} producer - The Kafka Producer object. + * @param {array} topics - Array of topics + * @param {object} options - Topic configuration. + * @constructor + * @extends stream.Writable + */ +function TopicWritable(producer, topicName, options) { + if (!(this instanceof TopicWritable)) { + return new TopicWritable(topicName, producer, options); + } + + if (options === undefined) { + options = {}; + } else if (typeof options === 'string') { + options = { encoding: options }; + } else if (options === null || typeof options !== 'object') { + throw new TypeError('"options" argument must be a string or an object'); + } + + options = Object.create(options); + + if (options.topic) { + delete options.topic; + } + + Writable.call(this, options); + + this.producer = producer; + this.topicName = topicName; + this.topicOptions = options.topic || {}; + this.topic = undefined; + this.autoClose = options.autoClose === undefined ? true : !!options.autoClose; + this.connected = false; + + if (options.encoding) { + this.setDefaultEncoding(options.encoding); + } + + if (this.producer.isConnected()) { + this.connected = true; + } else { + this.connect(); + } + + this.once('finish', function() { + if (this.autoClose) { + this.close(); + } + }); + +} + +TopicWritable.prototype.connect = function() { + this.producer.connect({}, function(err, data) { + if (err) { + this.emit('error', err); + return; + } + + this.connected = true; + this.emit('connected', data); + + }.bind(this)); +}; + +/** + * Internal stream write method for TopicWritable. + * + * This method should never be called externally. It has some recursion to + * handle cases where the producer is not yet connected. + * + * @param {buffer} chunk - Chunk to write. + * @param {string} encoding - Encoding for the buffer + * @param {Function} cb - Callback to call when the stream is done processing + * the data. + * @private + * @see https://github.com/nodejs/node/blob/master/lib/fs.js#L1901 + */ +TopicWritable.prototype._write = function(data, encoding, cb) { + if (!(data instanceof Buffer)) { + this.emit('error', new Error('Invalid data. Can only produce buffers')); + return; + } + + if (!this.connected) { + this.once('connected', function() { + this._write(data, encoding, cb); + }); + return; + } + + var self = this; + + if (self.topic === undefined) { + self.topic = self.producer.Topic(self.topicName, self.topicOptions); + } + + return this.producer.produce({ + topic: self.topic, + message: data, + }, function streamSendMessageCallback(err) { + if (err) { + if (self.autoClose) { + self.close(); + } + } + return cb(err); + }); +}; + +function writev(producer, topic, chunks, cb) { + + // @todo maybe a produce batch method? + var doneCount = 0; + var err = null; + + function maybeDone(e) { + if (e) { + err = e; + } + doneCount ++; + if (doneCount === chunks.length) { + cb(err); + } + } + + for (var i = 0; i < chunks.length; i++) { + producer.produce({ + topic: topic, + message: chunks[i] + }, maybeDone); + } + +} + +TopicWritable.prototype._writev = function(data, cb) { + if (!this.connected) { + this.once('connected', function() { + this._writev(data, cb); + }); + return; + } + + var self = this; + var len = data.length; + var chunks = new Array(len); + var size = 0; + + for (var i = 0; i < len; i++) { + var chunk = data[i].chunk; + + chunks[i] = chunk; + size += chunk.length; + } + + writev(this.producer, this.topic, chunks, function(err) { + if (err) { + self.close(); + cb(err); + return; + } + cb(); + }); + +}; + +TopicWritable.prototype.close = function(cb) { + var self = this; + if (cb) { + this.once('close', cb); + } + + self.connected = false; + setImmediate(close); + + function close() { + self.emit('closed'); + } +}; diff --git a/make_docs.sh b/make_docs.sh new file mode 100755 index 00000000..0511bcfa --- /dev/null +++ b/make_docs.sh @@ -0,0 +1,47 @@ +#!/bin/bash + +if [[ `git status --porcelain` ]]; then + # changes + >&2 echo "You have unstaged changes. Please commit before you run this." + exit 1 +fi + +make docs + +# Get package version and save to variable + +PACKAGE=$(node -pe 'require("./package.json").name.split("/")[1]') +VERSION=$(node -pe 'require("./package.json").version') + +# Make a temporary folder + +TEMPDIR=$(mktemp -d) + +VERSIONDIR="$TEMPDIR/$VERSION" +cp -r docs $VERSIONDIR + +# Now, checkout the gh-pages, but first get current checked out branch +# + +CURRENT_BRANCH=$(git rev-parse --symbolic-full-name --abbrev-ref HEAD) + +COMMIT_MESSAGE=$(git log --pretty='format:%B' -1) +COMMIT_AUTHOR=$(git log --pretty='format:%aN <%aE>' -1) + +if [[ `git checkout --quiet gh-pages` ]]; then + >&2 echo "Could not checkout gh-pages" + exit 1 +fi + +rm -rf current +rm -rf $VERSION + +cp -r $VERSIONDIR $VERSION +cp -r $VERSIONDIR current + +git add --all +git commit --author="$COMMIT_AUTHOR" -m "Updated docs for '$COMMIT_MESSAGE'" + +rm -rf $TEMPDIR + +git checkout $CURRENT_BRANCH diff --git a/package.json b/package.json new file mode 100644 index 00000000..23685538 --- /dev/null +++ b/package.json @@ -0,0 +1,44 @@ +{ + "name": "node-rdkafka", + "version": "0.3.0", + "description": "Node.js bindings for librdkafka", + "main": "lib/index.js", + "scripts": { + "configure": "node-gyp configure", + "build": "node-gyp build", + "test": "make test" + }, + "keywords": [ + "kafka", + "librdkafka" + ], + "repository": { + "type": "git", + "url": "git@github.com:Blizzard/node-rdkafka.git" + }, + "contributors": [ + { + "name": "Stephen Parente", + "email": "webmakersteve@gmail.com" + }, + { + "name": "Matt Gollob", + "email": "mattness@users.noreply.github.com" + } + ], + "license": "MIT", + "devDependencies": { + "jshint": "2.x", + "mocha": "2.x", + "node-gyp": "3.x" + }, + "dependencies": { + "bindings": "1.x", + "jsdoc": "^3.4.0", + "nan": "2.x", + "toolkit-jsdoc": "^1.0.0" + }, + "engines": { + "npm": "^2.7.3" + } +} diff --git a/src/binding.cc b/src/binding.cc new file mode 100644 index 00000000..e0b43a6e --- /dev/null +++ b/src/binding.cc @@ -0,0 +1,41 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +#include +#include "src/binding.h" + +using NodeKafka::Producer; +using NodeKafka::Consumer; +using NodeKafka::Topic; + +using node::AtExit; + +static void RdKafkaCleanup(void*) { // NOLINT + /* + * Wait for RdKafka to decommission. + * This is not strictly needed but + * allows RdKafka to clean up all its resources before the application + * exits so that memory profilers such as valgrind wont complain about + * memory leaks. + */ + + RdKafka::wait_destroyed(5000); +} + +void Init(v8::Local exports, v8::Local module) { + AtExit(RdKafkaCleanup); + Consumer::Init(exports); + Producer::Init(exports); + Topic::Init(exports); + + exports->Set(Nan::New("librdkafkaVersion").ToLocalChecked(), + Nan::New(RdKafka::version_str().c_str()).ToLocalChecked()); +} + +NODE_MODULE(kafka, Init) diff --git a/src/binding.h b/src/binding.h new file mode 100644 index 00000000..e6c70a0f --- /dev/null +++ b/src/binding.h @@ -0,0 +1,23 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +#ifndef SRC_BINDING_H_ +#define SRC_BINDING_H_ + +#include +#include "deps/librdkafka/src-cpp/rdkafkacpp.h" +#include "src/common.h" +#include "src/errors.h" +#include "src/config.h" +#include "src/connection.h" +#include "src/consumer.h" +#include "src/producer.h" +#include "src/topic.h" + +#endif // SRC_BINDING_H_ diff --git a/src/callbacks.cc b/src/callbacks.cc new file mode 100644 index 00000000..3bada551 --- /dev/null +++ b/src/callbacks.cc @@ -0,0 +1,510 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +#include +#include + +#include "src/callbacks.h" +#include "src/consumer.h" + +using v8::Local; +using v8::Value; +using v8::Object; +using v8::String; +using v8::Array; +using v8::Number; + +namespace NodeKafka { +namespace Callbacks { + +Dispatcher::Dispatcher() { + async = new uv_async_t; + uv_async_init(uv_default_loop(), async, AsyncMessage_); + + async->data = this; + + uv_mutex_init(&async_lock); +} + +Dispatcher::~Dispatcher() { + if (callbacks.size() < 1) return; + + for (size_t i=0; i < callbacks.size(); i++) { + callbacks[i].Reset(); + } + + // async->data = this; + + uv_mutex_destroy(&async_lock); +} + +bool Dispatcher::HasCallbacks() { + return callbacks.size() > 0; +} + +void Dispatcher::Execute() { + uv_async_send(async); +} + +void Dispatcher::Dispatch(const int _argc, Local _argv[]) { + // This should probably be an array of v8 values + if (!HasCallbacks()) { + return; + } + + for (size_t i=0; i < callbacks.size(); i++) { + v8::Local f = Nan::New(callbacks[i]); + Nan::Callback cb(f); + cb.Call(_argc, _argv); + } +} + +void Dispatcher::AddCallback(v8::Local func) { + Nan::Persistent > value(func); + // PersistentCopyableFunction value(func); + callbacks.push_back(value); +} + +event_t::event_t(const RdKafka::Event &event) { + message = ""; + fac = ""; + + type = event.type(); + + switch (type = event.type()) { + case RdKafka::Event::EVENT_ERROR: + message = RdKafka::err2str(event.err()); + break; + case RdKafka::Event::EVENT_STATS: + message = event.str(); + break; + case RdKafka::Event::EVENT_LOG: + severity = event.severity(); + fac = event.fac(); + message = event.str(); + break; + case RdKafka::Event::EVENT_THROTTLE: + message = RdKafka::err2str(event.err()); + throttle_time = event.throttle_time(); + broker_name = event.broker_name(); + broker_id = static_cast(event.broker_id()); + break; + default: + message = event.str(); + break; + } +} +event_t::~event_t() {} + +// Event callback +Event::Event(): + dispatcher() {} + +Event::~Event() {} + +void Event::event_cb(RdKafka::Event &event) { + // Second parameter is going to be an object with properties to + // represent the others. + + if (!dispatcher.HasCallbacks()) { + return; + } + + event_t e(event); + + dispatcher.Add(e); + dispatcher.Execute(); +} + +EventDispatcher::EventDispatcher() {} +EventDispatcher::~EventDispatcher() {} + +void EventDispatcher::Add(const event_t &e) { + scoped_mutex_lock lock(async_lock); + events.push_back(e); +} + +void EventDispatcher::Flush() { + Nan::HandleScope scope; + // Iterate through each of the currently stored events + // generate a callback object for each, setting to the members + // then + if (events.size() < 1) return; + + const unsigned int argc = 2; + + std::vector _events; + { + scoped_mutex_lock lock(async_lock); + events.swap(_events); + } + + for (size_t i=0; i < _events.size(); i++) { + Local argv[argc] = {}; + Local jsobj = Nan::New(); + + switch (_events[i].type) { + case RdKafka::Event::EVENT_ERROR: + argv[0] = Nan::New("error").ToLocalChecked(); + argv[1] = Nan::Error(_events[i].message.c_str()); + + // if (event->err() == RdKafka::ERR__ALL_BROKERS_DOWN). Stop running + // This may be better suited to the node side of things + break; + case RdKafka::Event::EVENT_STATS: + argv[0] = Nan::New("stats").ToLocalChecked(); + + Nan::Set(jsobj, Nan::New("message").ToLocalChecked(), + Nan::New(_events[i].message.c_str()).ToLocalChecked()); + + break; + case RdKafka::Event::EVENT_LOG: + argv[0] = Nan::New("log").ToLocalChecked(); + + Nan::Set(jsobj, Nan::New("severity").ToLocalChecked(), + Nan::New(_events[i].severity)); + Nan::Set(jsobj, Nan::New("fac").ToLocalChecked(), + Nan::New(_events[i].fac.c_str()).ToLocalChecked()); + Nan::Set(jsobj, Nan::New("message").ToLocalChecked(), + Nan::New(_events[i].message.c_str()).ToLocalChecked()); + + break; + case RdKafka::Event::EVENT_THROTTLE: + argv[0] = Nan::New("throttle").ToLocalChecked(); + + Nan::Set(jsobj, Nan::New("message").ToLocalChecked(), + Nan::New(_events[i].message.c_str()).ToLocalChecked()); + + Nan::Set(jsobj, Nan::New("throttleTime").ToLocalChecked(), + Nan::New(_events[i].throttle_time)); + Nan::Set(jsobj, Nan::New("brokerName").ToLocalChecked(), + Nan::New(_events[i].broker_name).ToLocalChecked()); + Nan::Set(jsobj, Nan::New("brokerId").ToLocalChecked(), + Nan::New(_events[i].broker_id)); + + break; + default: + argv[0] = Nan::New("event").ToLocalChecked(); + + Nan::Set(jsobj, Nan::New("message").ToLocalChecked(), + Nan::New(events[i].message.c_str()).ToLocalChecked()); + + break; + } + + if (_events[i].type != RdKafka::Event::EVENT_ERROR) { + // error would be assigned already + argv[1] = jsobj; + } + + Dispatch(argc, argv); + } +} + + +ConsumeDispatcher::ConsumeDispatcher() {} +ConsumeDispatcher::~ConsumeDispatcher() {} + +void ConsumeDispatcher::Add(NodeKafka::Message *m) { + scoped_mutex_lock lock(async_lock); + events.push_back(m); +} + +void ConsumeDispatcher::Flush() { + Nan::HandleScope scope; + // Iterate through each of the currently stored events + // generate a callback object for each, setting to the members + // then + if (events.size() < 1) return; + + const unsigned int argc = 2; + + std::vector _events; + { + scoped_mutex_lock lock(async_lock); + events.swap(_events); + } + + for (size_t i=0; i < _events.size(); i++) { + Local argv[argc] = {}; + + switch (_events[i]->errcode()) { + case RdKafka::ERR_NO_ERROR: + /* Real message */ + argv[0] = Nan::Null(); + argv[1] = _events[i]->Pack(); + + break; + + case RdKafka::ERR__TIMED_OUT: + argv[0] = Nan::Error("Message timed out"); + break; + + case RdKafka::ERR__PARTITION_EOF: + /* Last message */ + argv[0] = Nan::Error("EOF reached for all partition(s)"); + break; + + case RdKafka::ERR__UNKNOWN_TOPIC: + case RdKafka::ERR__UNKNOWN_PARTITION: + argv[0] = _events[i]->GetErrorObject(); + break; + + default: + argv[0] = _events[i]->GetErrorObject(); + } + + if (_events[i]->errcode() != RdKafka::ERR_NO_ERROR) { + argv[1] = Nan::Null(); + delete _events[i]; + } + + Dispatch(argc, argv); + } +} + +// Consume callback + +Consume::Consume(): + dispatcher() {} +Consume::~Consume() {} + +void Consume::consume_cb(RdKafka::Message &msg, void *opaque) { + NodeKafka::Message * m = new NodeKafka::Message(&msg); + dispatcher.Add(m); + dispatcher.Execute(); +} + +DeliveryReportDispatcher::DeliveryReportDispatcher() {} +DeliveryReportDispatcher::~DeliveryReportDispatcher() {} + +void DeliveryReportDispatcher::Add(const delivery_report_t &e) { + scoped_mutex_lock lock(async_lock); + events.push_back(e); +} + +void DeliveryReportDispatcher::Flush() { + Nan::HandleScope scope; + // + if (events.size() < 1) return; + + const unsigned int argc = 2; + + std::vector _events; + { + scoped_mutex_lock lock(async_lock); + events.swap(_events); + } + + for (size_t i=0; i < _events.size(); i++) { + v8::Local argv[argc] = {}; + + if (_events[i].is_error) { + // If it is an error we need the first argument to be set + argv[0] = Nan::Error(_events[i].error_string.c_str()); + argv[1] = Nan::Null(); + } else { + argv[0] = Nan::Null(); + Local jsobj(Nan::New()); + + Nan::Set(jsobj, Nan::New("topic_name").ToLocalChecked(), + Nan::New(_events[i].topic_name).ToLocalChecked()); + Nan::Set(jsobj, Nan::New("partition").ToLocalChecked(), + Nan::New(_events[i].partition)); + Nan::Set(jsobj, Nan::New("offset").ToLocalChecked(), + Nan::New(_events[i].offset)); + if (_events[i].key.empty()) { + Nan::Set(jsobj, Nan::New("key").ToLocalChecked(), Nan::Null()); + } else { + Nan::Set(jsobj, Nan::New("key").ToLocalChecked(), + Nan::New(_events[i].key).ToLocalChecked()); + } + + argv[1] = jsobj; + } + + Dispatch(argc, argv); + } +} + +delivery_report_t::delivery_report_t(RdKafka::Message &message) { + if (message.err() == RdKafka::ERR_NO_ERROR) { + is_error = false; + topic_name = message.topic_name(); + partition = message.partition(); + offset = message.offset(); + if (message.key_len() > 0) { + key = *message.key(); + } else { + key = ""; + } + } else { + is_error = true; + error_code = message.err(); + error_string = message.errstr(); + } +} + +delivery_report_t::~delivery_report_t() {} + +// Delivery Report + +Delivery::Delivery(): + dispatcher() {} +Delivery::~Delivery() {} + +void Delivery::dr_cb(RdKafka::Message &message) { + if (!dispatcher.HasCallbacks()) { + return; + } + + delivery_report_t msg(message); + dispatcher.Add(msg); + dispatcher.Execute(); +} + +// Rebalance CB + +RebalanceDispatcher::RebalanceDispatcher() {} +RebalanceDispatcher::~RebalanceDispatcher() {} + +void RebalanceDispatcher::Add(const rebalance_event_t &e) { + scoped_mutex_lock lock(async_lock); + events.push_back(e); +} + +void RebalanceDispatcher::Flush() { + Nan::HandleScope scope; + // Iterate through each of the currently stored events + // generate a callback object for each, setting to the members + // then + + if (events.size() < 1) return; + + const unsigned int argc = 1; + + std::vector _events; + { + scoped_mutex_lock lock(async_lock); + events.swap(_events); + } + + for (size_t i=0; i < _events.size(); i++) { + v8::Local argv[argc] = {}; + Local jsobj(Nan::New()); + + switch (_events[i].err) { + case RdKafka::ERR__ASSIGN_PARTITIONS: + Nan::Set(jsobj, Nan::New("code").ToLocalChecked(), + Nan::New(500)); + break; + default: + Nan::Set(jsobj, Nan::New("code").ToLocalChecked(), + Nan::New(501)); + break; + } + + argv[0] = jsobj; + + Dispatch(argc, argv); + } +} + +Rebalance::~Rebalance() {} +Rebalance::Rebalance(NodeKafka::Consumer* that) : + that_(that) { + eof_cnt = 0; +} + +void Rebalance::rebalance_cb(RdKafka::KafkaConsumer *consumer, + RdKafka::ErrorCode err, + std::vector &partitions) { + if (err == RdKafka::ERR__ASSIGN_PARTITIONS) { + that_->Assign(partitions); + } else { + that_->Unassign(); + } + + dispatcher.Add(rebalance_event_t(err, partitions)); + dispatcher.Execute(); + + eof_cnt = 0; +} + +// Partitioner callback + +Partitioner::Partitioner() {} +Partitioner::~Partitioner() {} + +int32_t Partitioner::partitioner_cb(const RdKafka::Topic *topic, + const std::string *key, + int32_t partition_cnt, + void *msg_opaque) { + // Send this and get teh callback and parse the int + if (callback.IsEmpty()) { + // default behavior + return random(topic, partition_cnt); + } + + Local argv[3] = {}; + + argv[0] = Nan::New(topic->name().c_str()).ToLocalChecked(); + if (key->empty()) { + argv[1] = Nan::Null(); + } else { + argv[1] = Nan::New(key->c_str()).ToLocalChecked(); + } + + argv[2] = Nan::New(partition_cnt); + + v8::Local return_value = callback.Call(3, argv); + + Nan::Maybe partition_return = Nan::To(return_value); + + int32_t chosen_partition; + + if (partition_return.IsNothing()) { + chosen_partition = RdKafka::Topic::PARTITION_UA; + } else { + chosen_partition = partition_return.FromJust(); + } + + if (!topic->partition_available(chosen_partition)) { + return RdKafka::Topic::PARTITION_UA; + } + + return chosen_partition; +} + +unsigned int Partitioner::djb_hash(const char *str, size_t len) { + unsigned int hash = 5381; + for (size_t i = 0 ; i < len ; i++) + hash = ((hash << 5) + hash) + str[i]; + return hash; +} + +unsigned int Partitioner::random(const RdKafka::Topic *topic, int32_t max) { + int32_t random_partition = rand() % max; // NOLINT + + if (topic->partition_available(random_partition)) { + return random_partition; + } else { + return RdKafka::Topic::PARTITION_UA; + } +} + +void Partitioner::SetCallback(v8::Local cb) { + callback(cb); +} + + +} // end namespace Callbacks + +} // End namespace NodeKafka diff --git a/src/callbacks.h b/src/callbacks.h new file mode 100644 index 00000000..c24f39cf --- /dev/null +++ b/src/callbacks.h @@ -0,0 +1,191 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +#ifndef SRC_CALLBACKS_H_ +#define SRC_CALLBACKS_H_ + +#include +#include + +#include + +#include "deps/librdkafka/src-cpp/rdkafkacpp.h" +#include "src/common.h" +#include "src/message.h" + +typedef Nan::Persistent > PersistentCopyableFunction; +typedef std::vector CopyableFunctionList; + +namespace NodeKafka { + +class Consumer; + +namespace Callbacks { + +class Dispatcher { + public: + Dispatcher(); + ~Dispatcher(); + void Dispatch(const int, v8::Local []); + void AddCallback(v8::Local); + bool HasCallbacks(); + virtual void Flush() = 0; + void Execute(); + protected: + std::vector > > callbacks; // NOLINT + + uv_mutex_t async_lock; + private: + NAN_INLINE static NAUV_WORK_CB(AsyncMessage_) { + Dispatcher *dispatcher = + static_cast(async->data); + dispatcher->Flush(); + } + + uv_async_t *async; +}; + +struct event_t { + RdKafka::Event::Type type; + std::string message; + + RdKafka::Event::Severity severity; + std::string fac; + + std::string broker_name; + int throttle_time; + int broker_id; + + explicit event_t(const RdKafka::Event &); + ~event_t(); +}; + +class EventDispatcher : public Dispatcher { + public: + EventDispatcher(); + ~EventDispatcher(); + void Add(const event_t &); + void Flush(); + protected: + std::vector events; +}; + +class Event : public RdKafka::EventCb { + public: + Event(); + ~Event(); + void event_cb(RdKafka::Event&); + EventDispatcher dispatcher; +}; + +class ConsumeDispatcher : public Dispatcher { + public: + ConsumeDispatcher(); + ~ConsumeDispatcher(); + void Add(NodeKafka::Message *); + void Flush(); + protected: + std::vector events; +}; + +class Consume : public RdKafka::ConsumeCb { + public: + Consume(); + ~Consume(); + void consume_cb(RdKafka::Message &, void *); + ConsumeDispatcher dispatcher; +}; + +struct delivery_report_t { + // If it is an error these will be set + bool is_error; + std::string error_string; + RdKafka::ErrorCode error_code; + + // If it is not + std::string topic_name; + int32_t partition; + int64_t offset; + std::string key; + + explicit delivery_report_t(RdKafka::Message &); + ~delivery_report_t(); +}; + +class DeliveryReportDispatcher : public Dispatcher { + public: + DeliveryReportDispatcher(); + ~DeliveryReportDispatcher(); + void Flush(); + void Add(const delivery_report_t &); + protected: + std::vector events; +}; + +class Delivery : public RdKafka::DeliveryReportCb { + public: + Delivery(); + ~Delivery(); + void dr_cb(RdKafka::Message&); + DeliveryReportDispatcher dispatcher; +}; + +// Rebalance dispatcher + +struct rebalance_event_t { + RdKafka::ErrorCode err; + std::vector partitions; + + rebalance_event_t(RdKafka::ErrorCode _err, + std::vector _partitions): + err(_err), + partitions(_partitions) {} +}; + +class RebalanceDispatcher : public Dispatcher { + public: + RebalanceDispatcher(); + ~RebalanceDispatcher(); + void Add(const rebalance_event_t &); + void Flush(); + protected: + std::vector events; +}; + +class Rebalance : public RdKafka::RebalanceCb { + public: + explicit Rebalance(NodeKafka::Consumer* that); + ~Rebalance(); + // NAN_DISALLOW_ASSIGN_COPY_MOVE? + NodeKafka::Consumer* const that_; + + void rebalance_cb(RdKafka::KafkaConsumer *, RdKafka::ErrorCode, + std::vector &); + RebalanceDispatcher dispatcher; + private: + int eof_cnt; +}; + +class Partitioner : public RdKafka::PartitionerCb { + public: + Partitioner(); + ~Partitioner(); + int32_t partitioner_cb( const RdKafka::Topic*, const std::string*, int32_t, void*); // NOLINT + Nan::Callback callback; // NOLINT + void SetCallback(v8::Local); + private: + static unsigned int djb_hash(const char*, size_t); + static unsigned int random(const RdKafka::Topic*, int32_t); +}; + +} // namespace Callbacks + +} // namespace NodeKafka + +#endif // SRC_CALLBACKS_H_ diff --git a/src/common.cc b/src/common.cc new file mode 100644 index 00000000..e9b4887b --- /dev/null +++ b/src/common.cc @@ -0,0 +1,117 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +#include +#include + +#include "src/common.h" + +namespace NodeKafka { + +void Log(std::string str) { + std::cerr << "% " << str.c_str() << std::endl; +} + +template +T GetParameter(v8::Local object, std::string field_name, T def) { + v8::Local field = Nan::New(field_name.c_str()).ToLocalChecked(); + if (Nan::Has(object, field).FromMaybe(false)) { + Nan::Maybe maybeT = Nan::To(Nan::Get(object, field).ToLocalChecked()); + if (maybeT.IsNothing()) { + return def; + } else { + return maybeT.FromJust(); + } + } + return def; +} + +template<> +int64_t GetParameter(v8::Local object, + std::string field_name, int64_t def) { + v8::Local field = Nan::New(field_name.c_str()).ToLocalChecked(); + if (Nan::Has(object, field).FromMaybe(false)) { + Nan::Maybe maybeInt = Nan::To( + Nan::Get(object, field).ToLocalChecked()); + if (maybeInt.IsNothing()) { + return def; + } else { + return maybeInt.FromJust(); + } + } + return def; +} + +template<> +int GetParameter(v8::Local object, + std::string field_name, int def) { + v8::Local field = Nan::New(field_name.c_str()).ToLocalChecked(); + if (Nan::Has(object, field).FromMaybe(false)) { + Nan::Maybe maybeInt = Nan::To( + Nan::Get(object, field).ToLocalChecked()); + if (!maybeInt.IsNothing()) { + return static_cast(maybeInt.FromJust()); + } + } + return def; +} + +template<> +std::string GetParameter(v8::Local object, + std::string field_name, + std::string def) { + v8::Local field = Nan::New(field_name.c_str()).ToLocalChecked(); + if (Nan::Has(object, field).FromMaybe(false)) { + Nan::MaybeLocal parameter = + Nan::To(Nan::Get(object, field).ToLocalChecked()); + + if (!parameter.IsEmpty()) { + Nan::Utf8String parameterValue(parameter.ToLocalChecked()); + std::string parameterString(*parameterValue); + + return parameterString; + } + } + return def; +} + +template<> +std::vector GetParameter >( + v8::Local object, std::string field_name, + std::vector def) { + v8::Local field = Nan::New(field_name.c_str()).ToLocalChecked(); + + if (Nan::Has(object, field).FromMaybe(false)) { + v8::Local maybeArray = Nan::Get(object, field).ToLocalChecked(); + if (maybeArray->IsArray()) { + v8::Local parameter = maybeArray.As(); + return v8ArrayToStringVector(parameter); + } + } + return def; +} + +std::vector v8ArrayToStringVector(v8::Local parameter) { + std::vector newItem; + + if (parameter->Length() >= 1) { + for (unsigned int i = 0; i < parameter->Length(); i++) { + Nan::MaybeLocal p = Nan::To(parameter->Get(i)); + if (p.IsEmpty()) { + continue; + } + Nan::Utf8String pVal(p.ToLocalChecked()); + std::string pString(*pVal); + newItem.push_back(pString); + } + } + return newItem; +} + +} // namespace NodeKafka diff --git a/src/common.h b/src/common.h new file mode 100644 index 00000000..6f776d69 --- /dev/null +++ b/src/common.h @@ -0,0 +1,50 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +#ifndef SRC_COMMON_H_ +#define SRC_COMMON_H_ + +#include + +#include +#include +#include + +#include "deps/librdkafka/src-cpp/rdkafkacpp.h" + +namespace NodeKafka { + +void Log(std::string); + +template T GetParameter(v8::Local, std::string, T); +template<> std::string GetParameter( + v8::Local, std::string, std::string); +template<> std::vector GetParameter >( + v8::Local, std::string, std::vector); +// template int GetParameter(v8::Local v8ArrayToStringVector(v8::Local); + +class scoped_mutex_lock { + public: + explicit scoped_mutex_lock(uv_mutex_t& lock_) : // NOLINT + async_lock(lock_) { + uv_mutex_lock(&async_lock); + } + + ~scoped_mutex_lock() { + uv_mutex_unlock(&async_lock); + } + + private: + uv_mutex_t &async_lock; +}; + +} // namespace NodeKafka + +#endif // SRC_COMMON_H_ diff --git a/src/config.cc b/src/config.cc new file mode 100644 index 00000000..adbe6e78 --- /dev/null +++ b/src/config.cc @@ -0,0 +1,77 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +#include +#include + +#include "src/config.h" + +using RdKafka::Conf; +using Nan::MaybeLocal; +using Nan::Maybe; +using v8::Local; +using v8::String; +using v8::Object; +using std::cout; +using std::endl; + +namespace NodeKafka { + +namespace Config { + +template +void LoadParameter(v8::Local object, std::string field, const T &to) { // NOLINT + to = GetParameter(object, field, to); +} + +std::string GetValue(RdKafka::Conf* rdconf, const std::string name) { + std::string value; + if (rdconf->get(name, value) == RdKafka::Conf::CONF_UNKNOWN) { + return std::string(); + } + + return value; +} + +RdKafka::Conf* Create(RdKafka::Conf::ConfType type, v8::Local object, std::string &errstr) { // NOLINT + RdKafka::Conf* rdconf = RdKafka::Conf::create(type); + + v8::Local property_names = object->GetOwnPropertyNames(); + + for (unsigned int i = 0; i < property_names->Length(); ++i) { + std::string string_value; + std::string string_key; + + v8::Local key = property_names->Get(i); + v8::Local value = object->Get(key); + + if (key->IsString()) { + Nan::Utf8String utf8_key(key); + string_key = std::string(*utf8_key); + } else { + continue; + } + + if (!value->IsFunction()) { + Nan::Utf8String utf8_value(value.As()); + string_value = std::string(*utf8_value); + if (rdconf->set(string_key, string_value, errstr) + != Conf::CONF_OK) { + delete rdconf; + return NULL; + } + } + } + + return rdconf; +} + +} // namespace Config + +} // namespace NodeKafka diff --git a/src/config.h b/src/config.h new file mode 100644 index 00000000..210140a2 --- /dev/null +++ b/src/config.h @@ -0,0 +1,31 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +#ifndef SRC_CONFIG_H_ +#define SRC_CONFIG_H_ + +#include +#include +#include + +#include "deps/librdkafka/src-cpp/rdkafkacpp.h" +#include "src/common.h" + +namespace NodeKafka { +namespace Config { + +template void LoadParameter(v8::Local, std::string, T &); // NOLINT +std::string GetValue(RdKafka::Conf*, const std::string); +RdKafka::Conf* Create(RdKafka::Conf::ConfType, v8::Local, std::string &); // NOLINT + +} // namespace Config + +} // namespace NodeKafka + +#endif // SRC_CONFIG_H_ diff --git a/src/connection.cc b/src/connection.cc new file mode 100644 index 00000000..13474142 --- /dev/null +++ b/src/connection.cc @@ -0,0 +1,197 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +#include +#include + +#include "src/connection.h" +#include "src/workers.h" + +using RdKafka::Conf; + +namespace NodeKafka { + +/** + * @brief Connection v8 wrapped object. + * + * Wraps the RdKafka::Handle object with compositional inheritence and + * provides sensible defaults for exposing callbacks to node + * + * This object can't itself expose methods to the prototype directly, as far + * as I can tell. But it can provide the NAN_METHODS that just need to be added + * to the prototype. Since connections, etc. are managed differently based on + * whether it is a producer or consumer, they manage that. This base class + * handles some of the wrapping functionality and more importantly, the + * configuration of callbacks + * + * Any callback available to both consumers and producers, like logging or + * events will be handled in here. + * + * @sa RdKafka::Handle + * @sa NodeKafka::Client + */ + +Connection::Connection(RdKafka::Conf* gconfig, RdKafka::Conf* tconfig): + m_event_cb(), + m_gconfig(gconfig), + m_tconfig(tconfig) { + std::string errstr; + + m_client = NULL; + m_is_closing = false; + uv_mutex_init(&m_connection_lock); + + // Try to set the event cb. Shouldn't be an error here, but if there + // is, it doesn't get reported. + // + // Perhaps node new methods should report this as an error? But there + // isn't anything the user can do about it. + m_gconfig->set("event_cb", &m_event_cb, errstr); + } + +Connection::~Connection() { + uv_mutex_destroy(&m_connection_lock); + if (m_gconfig) { + delete m_gconfig; + } + + if (m_tconfig) { + delete m_tconfig; + } +} + +RdKafka::TopicPartition* Connection::GetPartition(std::string &topic) { + return RdKafka::TopicPartition::create(topic, RdKafka::Topic::PARTITION_UA); +} + +RdKafka::TopicPartition* Connection::GetPartition(std::string &topic, int partition) { // NOLINT + return RdKafka::TopicPartition::create(topic, partition); +} + +bool Connection::IsConnected() { + return !m_is_closing && m_client != NULL; +} + +void Connection::DumpConfig(std::list *dump) { + for (std::list::iterator it = dump->begin(); + it != dump->end(); ) { + std::cout << *it << " = "; + it++; + std::cout << *it << std::endl; + it++; + } + std::cout << std::endl; +} + +RdKafka::Handle* Connection::GetClient() { + return m_client; +} + +Baton Connection::CreateTopic(std::string topic_name) { + RdKafka::Topic* topic = + RdKafka::Topic::create(m_client, topic_name, NULL, m_errstr); + + if (!m_errstr.empty()) { + return Baton(RdKafka::ErrorCode::ERR_TOPIC_EXCEPTION); + } + + // Maybe do it this way later? Then we don't need to do static_cast + // + return Baton(topic); +} + +Baton Connection::GetMetadata(std::string topic_name, int timeout_ms) { + RdKafka::Topic* topic = NULL; + RdKafka::ErrorCode err; + + std::string errstr; + + if (!topic_name.empty()) { + Baton b = CreateTopic(topic_name); + if (b.err() == RdKafka::ErrorCode::ERR_NO_ERROR) { + topic = b.data(); + } + } + + RdKafka::Metadata* metadata = NULL; + + if (!errstr.empty()) { + return Baton(RdKafka::ERR_TOPIC_EXCEPTION); + } + + if (IsConnected()) { + scoped_mutex_lock lock(m_connection_lock); + if (IsConnected()) { + err = m_client->metadata(topic != NULL, topic, &metadata, timeout_ms); + } else { + err = RdKafka::ERR__STATE; + } + } else { + err = RdKafka::ERR__STATE; + } + + if (err == RdKafka::ERR_NO_ERROR) { + return Baton(metadata); + } else { + // metadata is not set here + // @see https://github.com/edenhill/librdkafka/blob/master/src-cpp/rdkafkacpp.h#L860 + return Baton(err); + } +} + +// NAN METHODS + +NAN_METHOD(Connection::NodeGetMetadata) { + Nan::HandleScope scope; + + Connection* obj = ObjectWrap::Unwrap(info.This()); + + v8::Local config; + if (info[0]->IsObject()) { + config = info[0].As(); + } else { + config = Nan::New(); + } + + if (!info[1]->IsFunction()) { + Nan::ThrowError("Second parameter must be a callback"); + return; + } + + v8::Local cb = info[1].As(); + + std::string topic = GetParameter(config, "topic", ""); + int timeout_ms = GetParameter(config, "timeout", 3000); + + Nan::Callback *callback = new Nan::Callback(cb); + + Nan::AsyncQueueWorker(new Workers::ConnectionMetadata( + callback, obj, topic, timeout_ms)); + + info.GetReturnValue().Set(Nan::Null()); +} + +// Node methods +NAN_METHOD(Connection::NodeOnEvent) { + Nan::HandleScope scope; + + if (info.Length() < 1 || !info[0]->IsFunction()) { + // Just throw an exception + return Nan::ThrowError("Need to specify a callback"); + } + + Connection* obj = ObjectWrap::Unwrap(info.This()); + + v8::Local cb = info[0].As(); + obj->m_event_cb.dispatcher.AddCallback(cb); + + info.GetReturnValue().Set(Nan::True()); +} + +} // namespace NodeKafka diff --git a/src/connection.h b/src/connection.h new file mode 100644 index 00000000..c375257a --- /dev/null +++ b/src/connection.h @@ -0,0 +1,88 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +#ifndef SRC_CONNECTION_H_ +#define SRC_CONNECTION_H_ + +#include +#include +#include +#include + +#include "deps/librdkafka/src-cpp/rdkafkacpp.h" + +#include "src/common.h" +#include "src/errors.h" +#include "src/config.h" +#include "src/callbacks.h" + +namespace NodeKafka { + +/** + * @brief Connection v8 wrapped object. + * + * Wraps the RdKafka::Handle object with compositional inheritence and + * provides sensible defaults for exposing callbacks to node + * + * This object can't itself expose methods to the prototype directly, as far + * as I can tell. But it can provide the NAN_METHODS that just need to be added + * to the prototype. Since connections, etc. are managed differently based on + * whether it is a producer or consumer, they manage that. This base class + * handles some of the wrapping functionality and more importantly, the + * configuration of callbacks + * + * Any callback available to both consumers and producers, like logging or + * events will be handled in here. + * + * @sa RdKafka::Handle + * @sa NodeKafka::Client + */ + +class Connection : public Nan::ObjectWrap { + public: + bool IsConnected(); + + // Baton + Baton CreateTopic(std::string); + Baton GetMetadata(std::string, int); + + RdKafka::Handle* GetClient(); + + static RdKafka::TopicPartition* GetPartition(std::string &); + static RdKafka::TopicPartition* GetPartition(std::string &, int); + + Callbacks::Event m_event_cb; + + protected: + Connection(RdKafka::Conf*, RdKafka::Conf*); + ~Connection(); + + void DumpConfig(std::list *); + + static Nan::Persistent constructor; + static void New(const Nan::FunctionCallbackInfo& info); + + bool m_has_been_disconnected; + bool m_is_closing; + + RdKafka::Conf* m_gconfig; + RdKafka::Conf* m_tconfig; + std::string m_errstr; + + uv_mutex_t m_connection_lock; + + RdKafka::Handle* m_client; + + static NAN_METHOD(NodeOnEvent); + static NAN_METHOD(NodeGetMetadata); +}; + +} // namespace NodeKafka + +#endif // SRC_CONNECTION_H_ diff --git a/src/consumer.cc b/src/consumer.cc new file mode 100644 index 00000000..f83d8e1e --- /dev/null +++ b/src/consumer.cc @@ -0,0 +1,801 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +#include +#include + +#include "src/consumer.h" +#include "src/workers.h" + +using Nan::FunctionCallbackInfo; + +namespace NodeKafka { + +consumer_commit_t::consumer_commit_t(std::string topic_name, int partition, int64_t offset) { // NOLINT + _topic_name = topic_name; + _partition = partition; + _offset = offset; +} + +consumer_commit_t::consumer_commit_t() { + _topic_name = ""; +} + +/** + * @brief Consumer v8 wrapped object. + * + * Specializes the connection to wrap a consumer object through compositional + * inheritence. Establishes its prototype in node through `Init` + * + * @sa RdKafka::Handle + * @sa NodeKafka::Client + */ + +Consumer::Consumer(RdKafka::Conf* gconfig, RdKafka::Conf* tconfig): + Connection(gconfig, tconfig), + m_consume_cb(), + m_rebalance_cb(this) { + m_is_subscribed = false; + + std::string errstr; + + m_gconfig->set("rebalance_cb", &m_rebalance_cb, errstr); + m_gconfig->set("default_topic_conf", m_tconfig, errstr); + } + +Consumer::~Consumer() { + // We only want to run this if it hasn't been run already + Disconnect(); +} + +Baton Consumer::Connect() { + if (IsConnected()) { + return Baton(RdKafka::ERR_NO_ERROR); + } + + std::string errstr; + + m_client = RdKafka::KafkaConsumer::create(m_gconfig, errstr); + + if (!m_client || !errstr.empty()) { + return Baton(RdKafka::ERR__STATE); + } + + if (m_partitions.size() > 0) { + m_client->resume(m_partitions); + } + + return Baton(RdKafka::ERR_NO_ERROR); +} + +bool Consumer::HasAssignedPartitions() { + return !m_partitions.empty(); +} +int Consumer::AssignedPartitionCount() { + return m_partition_cnt; +} + +void Consumer::part_list_print(const std::vector &partitions) { // NOLINT + for (unsigned int i = 0 ; i < partitions.size() ; i++) + std::cerr << partitions[i]->topic() << + "[" << partitions[i]->partition() << "], "; + std::cerr << std::endl; +} + +void Consumer::Assign(std::vector partitions) { + RdKafka::KafkaConsumer* consumer = + dynamic_cast(m_client); + consumer->assign(partitions); + m_partition_cnt = partitions.size(); + m_partitions.swap(partitions); +} + +void Consumer::Unassign() { + RdKafka::KafkaConsumer* consumer = + dynamic_cast(m_client); + consumer->unassign(); + m_partitions.empty(); + m_partition_cnt = 0; + // is_connected = false; +} + +Baton Consumer::Commit(std::string topic_name, int partition, int64_t offset) { + if (!IsConnected()) { + return Baton(RdKafka::ERR__STATE); + } + + RdKafka::KafkaConsumer* consumer = + dynamic_cast(m_client); + + RdKafka::TopicPartition* topic = + RdKafka::TopicPartition::create(topic_name, partition); + topic->set_offset(offset); + + // Need to put topic in a vector for it to work + std::vector offsets = {topic}; + + RdKafka::ErrorCode err = consumer->commitSync(offsets); + + // We are done. Clean up our mess + delete topic; + + return Baton(err); +} + +Baton Consumer::Commit() { + // sets an error message + if (!IsConnected()) { + return Baton(RdKafka::ERR__STATE); + } + + RdKafka::KafkaConsumer* consumer = + dynamic_cast(m_client); + + RdKafka::ErrorCode err = consumer->commitSync(); + + return Baton(err); +} + +Baton Consumer::Unsubscribe() { + if (IsConnected() && IsSubscribed()) { + RdKafka::KafkaConsumer* consumer = + dynamic_cast(m_client); + consumer->unsubscribe(); + m_is_subscribed = false; + return Baton(RdKafka::ERR_NO_ERROR); + } else { + return Baton(RdKafka::ERR__STATE); + } +} + +Baton Consumer::Subscribe(std::vector topics) { + if (!IsConnected()) { + return Baton(RdKafka::ERR__STATE); + } + + RdKafka::KafkaConsumer* consumer = + dynamic_cast(m_client); + + RdKafka::ErrorCode errcode = consumer->subscribe(topics); + if (errcode != RdKafka::ERR_NO_ERROR) { + return Baton(errcode); + } + + m_is_subscribed = true; + + return Baton(RdKafka::ERR_NO_ERROR); +} + +NodeKafka::Message* Consumer::Consume() { + NodeKafka::Message* m; + + if (IsConnected() && IsSubscribed()) { + scoped_mutex_lock lock(m_connection_lock); + if (!IsConnected() && IsSubscribed()) { + m = new NodeKafka::Message(RdKafka::ERR__STATE); + } else { + RdKafka::KafkaConsumer* consumer = + dynamic_cast(m_client); + m = new NodeKafka::Message(consumer->consume(1000)); + + if (m->ConsumerShouldStop()) { + Unsubscribe(); + } + } + } else { + m = new NodeKafka::Message(RdKafka::ERR__STATE); + } + + return m; +} + +Baton Consumer::RefreshAssignments() { + if (!IsConnected()) { + return Baton(RdKafka::ERR__STATE); + } + + if (m_partitions.size() < 1) { + // If the assigned partition is not set it means we haven't + // rebalanced or been manually assigned yet. + return Baton(RdKafka::ERR__STATE); + } + + RdKafka::KafkaConsumer* consumer = + dynamic_cast(m_client); + + std::vector partition_list; + RdKafka::ErrorCode err = consumer->assignment(partition_list); + + switch (err) { + case RdKafka::ERR_NO_ERROR: + m_partition_cnt = partition_list.size(); + m_partitions.swap(partition_list); + return Baton(RdKafka::ERR_NO_ERROR); + break; + default: + return Baton(err); + break; + } +} + +std::string Consumer::Name() { + if (!IsConnected()) { + return std::string(""); + } + return std::string(m_client->name()); +} + +Nan::Persistent Consumer::constructor; + +void Consumer::Init(v8::Local exports) { + Nan::HandleScope scope; + + v8::Local tpl = Nan::New(New); + tpl->SetClassName(Nan::New("KafkaConsumer").ToLocalChecked()); + tpl->InstanceTemplate()->SetInternalFieldCount(1); + + /* + * Lifecycle events inherited from NodeKafka::Connection + * + * @sa NodeKafka::Connection + */ + + Nan::SetPrototypeMethod(tpl, "onEvent", NodeOnEvent); + + /* + * @brief Methods to do with establishing state + */ + + Nan::SetPrototypeMethod(tpl, "connect", NodeConnect); + Nan::SetPrototypeMethod(tpl, "disconnect", NodeDisconnect); + Nan::SetPrototypeMethod(tpl, "getMetadata", NodeGetMetadata); + // Nan::SetPrototypeMethod(tpl, "poll", NodePoll); + + /* + * Lifecycle events specifically designated for RdKafka::KafkaConsumer + * + * @sa RdKafka::KafkaConsumer + */ + + Nan::SetPrototypeMethod(tpl, "onConsume", NodeOnConsume); + Nan::SetPrototypeMethod(tpl, "onRebalance", NodeOnRebalance); + + /* + * @brief Methods exposed to do with message retrieval + */ + + Nan::SetPrototypeMethod(tpl, "subscribe", NodeSubscribe); + Nan::SetPrototypeMethod(tpl, "subscribeSync", NodeSubscribeSync); + Nan::SetPrototypeMethod(tpl, "unsubscribe", NodeUnsubscribe); + Nan::SetPrototypeMethod(tpl, "consumeLoop", NodeConsumeLoop); + Nan::SetPrototypeMethod(tpl, "consume", NodeConsume); + + /* + * @brief Methods to do with partition assignment / rebalancing + */ + + Nan::SetPrototypeMethod(tpl, "assign", NodeAssign); + Nan::SetPrototypeMethod(tpl, "unassign", NodeUnassign); + Nan::SetPrototypeMethod(tpl, "getAssignments", NodeGetAssignments); + + Nan::SetPrototypeMethod(tpl, "commit", NodeCommit); + Nan::SetPrototypeMethod(tpl, "commitSync", NodeCommitSync); + + constructor.Reset(tpl->GetFunction()); + exports->Set(Nan::New("KafkaConsumer").ToLocalChecked(), tpl->GetFunction()); +} + +void Consumer::New(const Nan::FunctionCallbackInfo& info) { + if (!info.IsConstructCall()) { + return Nan::ThrowError("non-constructor invocation not supported"); + } + + if (info.Length() < 2) { + return Nan::ThrowError("You must supply global and topic configuration"); + } + + if (!info[0]->IsObject()) { + return Nan::ThrowError("Global configuration data must be specified"); + } + + if (!info[1]->IsObject()) { + return Nan::ThrowError("Topic configuration must be specified"); + } + + std::string errstr; + + RdKafka::Conf* gconfig = + Config::Create(RdKafka::Conf::CONF_GLOBAL, info[0]->ToObject(), errstr); + + if (!gconfig) { + return Nan::ThrowError(errstr.c_str()); + } + + RdKafka::Conf* tconfig = + Config::Create(RdKafka::Conf::CONF_TOPIC, info[1]->ToObject(), errstr); + + if (!tconfig) { + delete gconfig; + return Nan::ThrowError(errstr.c_str()); + } + + Consumer* consumer = new Consumer(gconfig, tconfig); + + // Wrap it + consumer->Wrap(info.This()); + + // Then there is some weird initialization that happens + // basically it sets the configuration data + // we don't need to do that because we lazy load it + + info.GetReturnValue().Set(info.This()); +} + +v8::Local Consumer::NewInstance(v8::Local arg) { + Nan::EscapableHandleScope scope; + + const unsigned argc = 1; + + v8::Local argv[argc] = { arg }; + v8::Local cons = Nan::New(constructor); + v8::Local instance = cons->NewInstance(argc, argv); + + return scope.Escape(instance); +} + +Baton Consumer::Disconnect() { + // Only close client if it is connected + RdKafka::ErrorCode err = RdKafka::ERR_NO_ERROR; + + if (IsConnected()) { + m_is_closing = true; + { + scoped_mutex_lock lock(m_connection_lock); + + RdKafka::KafkaConsumer* consumer = + dynamic_cast(m_client); + err = consumer->close(); + + delete m_client; + m_client = NULL; + + RdKafka::wait_destroyed(1000); + } + } + + m_is_closing = false; + + return Baton(err); +} + +bool Consumer::IsSubscribed() { + if (!IsConnected()) { + return false; + } + + if (!m_is_subscribed) { + return false; + } + + return true; +} + +/* Node exposed methods */ + +NAN_METHOD(Consumer::NodeOnConsume) { + if (info.Length() < 1 || !info[0]->IsFunction()) { + // Just throw an exception + return Nan::ThrowError("Need to specify a callback"); + } + + Consumer* obj = ObjectWrap::Unwrap(info.This()); + + v8::Local cb = info[0].As(); + obj->m_consume_cb.dispatcher.AddCallback(cb); + + info.GetReturnValue().Set(Nan::True()); +} + +NAN_METHOD(Consumer::NodeOnRebalance) { + if (info.Length() < 1 || !info[0]->IsFunction()) { + // Just throw an exception + return Nan::ThrowError("Need to specify a callback"); + } + + Consumer* consumer = ObjectWrap::Unwrap(info.This()); + + v8::Local cb = info[0].As(); + + consumer->m_rebalance_cb.dispatcher.AddCallback(cb); + + info.GetReturnValue().Set(Nan::True()); +} + +NAN_METHOD(Consumer::NodeGetAssignments) { + Nan::HandleScope scope; + + Consumer* consumer = ObjectWrap::Unwrap(info.This()); + + if (!consumer->IsConnected()) { + Nan::ThrowError("Consumer is disconnected"); + return; + } + + std::vector::iterator it; + + v8::Local returnArray = Nan::New(); + + unsigned int i = 0; + Baton b = consumer->RefreshAssignments(); + + if (b.err() != RdKafka::ERR_NO_ERROR) { + Nan::ThrowError(RdKafka::err2str(b.err()).c_str()); + return; + } + + if (consumer->m_partitions.size() < 1) { + info.GetReturnValue().Set(returnArray); + return; + } + + for (it = consumer->m_partitions.begin(); it != consumer->m_partitions.end(); + ++it, ++i) { + RdKafka::TopicPartition* part = *it; + + v8::Local jsobj = Nan::New(); + + Nan::Set(jsobj, Nan::New("topic").ToLocalChecked(), + Nan::New(part->topic()).ToLocalChecked()); + Nan::Set(jsobj, Nan::New("partition").ToLocalChecked(), + Nan::New(part->partition())); + + returnArray->Set(i, jsobj); + } + + info.GetReturnValue().Set(returnArray); +} + +NAN_METHOD(Consumer::NodeAssign) { + Nan::HandleScope scope; + + if (info.Length() < 1 || !info[0]->IsArray()) { + // Just throw an exception + return Nan::ThrowError("Need to specify an array of partitions"); + } + + v8::Local partitions = info[0].As(); + std::vector _partitions; + + for (unsigned int i = 0; i < partitions->Length(); ++i) { + v8::Local partition_obj_value = partitions->Get(i); + if (!partition_obj_value->IsObject()) { + Nan::ThrowError("Must pass objects"); + } + + v8::Local partition_obj = partition_obj_value.As(); + + // Got the object + int64_t partition = GetParameter(partition_obj, "partition", -1); + std::string topic = GetParameter(partition_obj, "topic", ""); + + if (!topic.empty()) { + RdKafka::TopicPartition* part; + + if (partition < 0) { + part = Connection::GetPartition(topic); + } else { + part = Connection::GetPartition(topic, partition); + } + + _partitions.push_back(part); + } + } + Consumer* consumer = ObjectWrap::Unwrap(info.This()); + + if (!consumer->IsConnected()) { + Nan::ThrowError("Consumer is disconnected"); + return; + } + + consumer->Assign(_partitions); + + // i dont know who manages the memory at this point + // i have to assume it does because it is asking for pointers + + info.GetReturnValue().Set(Nan::True()); +} + +NAN_METHOD(Consumer::NodeUnassign) { + Nan::HandleScope scope; + + Consumer* consumer = ObjectWrap::Unwrap(info.This()); + + if (!consumer->IsConnected()) { + Nan::ThrowError("Consumer is disconnected"); + return; + } + + consumer->Unassign(); + + info.GetReturnValue().Set(Nan::True()); +} + +void Consumer::NodeUnsubscribe(const FunctionCallbackInfo& info) { + Nan::HandleScope scope; + + Consumer* consumer = ObjectWrap::Unwrap(info.This()); + // Don't check if we are subscribed. If we aren't just leave it be + + if (info.Length() < 1 || !info[0]->IsFunction()) { + Nan::ThrowError("First argument must be a function"); + } + + v8::Local cb = info[1].As(); + + Nan::Callback *callback = new Nan::Callback(cb); + + // These workers likely need to be tracked so we can stop them when we + // disconnect or run unubscribe + Nan::AsyncQueueWorker(new Workers::ConsumerUnsubscribe(callback, consumer)); + + info.GetReturnValue().Set(Nan::Null()); +} + +void Consumer::NodeUnsubscribeSync(const FunctionCallbackInfo& info) { // NOLINT + Nan::HandleScope scope; + + Consumer* consumer = ObjectWrap::Unwrap(info.This()); + + // Do it sync i guess + Baton b = consumer->Unsubscribe(); + + if (b.err() != RdKafka::ERR_NO_ERROR) { + info.GetReturnValue().Set(Nan::Error(RdKafka::err2str(b.err()).c_str())); + return; + } + + info.GetReturnValue().Set(Nan::True()); +} + +NAN_METHOD(Consumer::NodeCommitSync) { + Nan::HandleScope scope; + + Consumer* consumer = ObjectWrap::Unwrap(info.This()); + + if (!consumer->IsConnected()) { + Nan::ThrowError("Consumer is disconnected"); + return; + } + + // If we are provided a message object + if (info.Length() >= 1 && info[0]->IsObject()) { + v8::Local params = info[0].As(); + + // This one is a buffer + std::string topic_name = GetParameter(params, "topic", ""); + int partition = GetParameter(params, "partition", 0); + int64_t offset = GetParameter(params, "offset", -1); + + // Do it sync i guess + Baton b = consumer->Commit(topic_name, partition, offset); + + if (b.err() != RdKafka::ERR_NO_ERROR) { + info.GetReturnValue().Set(b.ToObject()); + return; + } + + } else { + Baton b = consumer->Commit(); + + if (b.err() != RdKafka::ERR_NO_ERROR) { + info.GetReturnValue().Set(b.ToObject()); + return; + } + + info.GetReturnValue().Set(Nan::True()); + } +} + +NAN_METHOD(Consumer::NodeCommit) { + Nan::HandleScope scope; + + Consumer* consumer = ObjectWrap::Unwrap(info.This()); + + if (info.Length() >= 1 && info[0]->IsObject()) { + v8::Local params = info[0].As(); + + // This one is a buffer + std::string topic_name = GetParameter(params, "topic", ""); + int partition = GetParameter(params, "partition", 0); + int64_t offset = GetParameter(params, "offset", -1); + + consumer_commit_t commit_request(topic_name, partition, offset); + v8::Local cb = info[1].As(); + + Nan::Callback *callback = new Nan::Callback(cb); + + Nan::AsyncQueueWorker( + new Workers::ConsumerCommit(callback, consumer, commit_request)); + + info.GetReturnValue().Set(Nan::Null()); + + } else { + v8::Local cb = info[0].As(); + + Nan::Callback *callback = new Nan::Callback(cb); + // These workers likely need to be tracked so we can stop them when we + // disconnect or run unubscribe + Nan::AsyncQueueWorker(new Workers::ConsumerCommit(callback, consumer)); + + info.GetReturnValue().Set(Nan::Null()); + } +} + +NAN_METHOD(Consumer::NodeSubscribe) { + Nan::HandleScope scope; + + if (info.Length() < 2) { + // Just throw an exception + return Nan::ThrowError("Invalid number of parameters"); + } + + if (!info[0]->IsArray()) { + return Nan::ThrowError("First parameter to subscribe must be an array"); + } + + if (!info[1]->IsFunction()) { + // Just throw an exception + return Nan::ThrowError("Need to specify a callback"); + } + + Consumer* consumer = ObjectWrap::Unwrap(info.This()); + + v8::Local topicsArray = info[0].As(); + v8::Local cb = info[1].As(); + + std::vector topics = v8ArrayToStringVector(topicsArray); + + if (topics.empty()) { + Nan::ThrowError("Please provide an array of topics to subscribe to"); + return; + } + + Nan::Callback *callback = new Nan::Callback(cb); + + // These workers likely need to be tracked so we can stop them when we + // disconnect or run unubscribe + Nan::AsyncQueueWorker( + new Workers::ConsumerSubscribe(callback, consumer, topics)); + + info.GetReturnValue().Set(Nan::Null()); +} + +NAN_METHOD(Consumer::NodeSubscribeSync) { + Nan::HandleScope scope; + + if (info.Length() < 1 || !info[0]->IsArray()) { + // Just throw an exception + return Nan::ThrowError("First parameter must be an array"); + } + + Consumer* consumer = ObjectWrap::Unwrap(info.This()); + + v8::Local topicsArray = info[0].As(); + std::vector topics = v8ArrayToStringVector(topicsArray); + + Baton b = consumer->Subscribe(topics); + + if (b.err() != RdKafka::ERR_NO_ERROR) { + info.GetReturnValue().Set(b.ToObject()); + } else { + info.GetReturnValue().Set(Nan::True()); + } +} + +NAN_METHOD(Consumer::NodeConsumeLoop) { + Nan::HandleScope scope; + + if (info.Length() < 1) { + // Just throw an exception + return Nan::ThrowError("Invalid number of parameters"); + } + + if (!info[0]->IsFunction()) { + return Nan::ThrowError("Need to specify a callback"); + } + + Consumer* consumer = ObjectWrap::Unwrap(info.This()); + + v8::Local cb = info[0].As(); + + Nan::Callback *callback = new Nan::Callback(cb); + Nan::AsyncQueueWorker(new Workers::ConsumerConsumeLoop(callback, consumer)); + + info.GetReturnValue().Set(Nan::Null()); +} + +NAN_METHOD(Consumer::NodeConsume) { + Nan::HandleScope scope; + + if (info.Length() < 1) { + // Just throw an exception + return Nan::ThrowError("Invalid number of parameters"); + } + + if (info[0]->IsNumber()) { + if (!info[1]->IsFunction()) { + return Nan::ThrowError("Need to specify a callback"); + } + + v8::Local numMessagesNumber = info[0].As(); + Nan::Maybe numMessagesMaybe = Nan::To(numMessagesNumber); // NOLINT + + uint32_t numMessages; + if (numMessagesMaybe.IsNothing()) { + return Nan::ThrowError("Parameter must be a number over 0"); + } else { + numMessages = numMessagesMaybe.FromJust(); + } + + Consumer* consumer = ObjectWrap::Unwrap(info.This()); + + v8::Local cb = info[1].As(); + Nan::Callback *callback = new Nan::Callback(cb); + Nan::AsyncQueueWorker( + new Workers::ConsumerConsumeNum(callback, consumer, numMessages)); + + } else { + if (!info[0]->IsFunction()) { + return Nan::ThrowError("Need to specify a callback"); + } + + Consumer* consumer = ObjectWrap::Unwrap(info.This()); + + v8::Local cb = info[0].As(); + Nan::Callback *callback = new Nan::Callback(cb); + Nan::AsyncQueueWorker(new Workers::ConsumerConsume(callback, consumer)); + } + + info.GetReturnValue().Set(Nan::Null()); +} + +NAN_METHOD(Consumer::NodeConnect) { + Nan::HandleScope scope; + + if (info.Length() < 1 || !info[0]->IsFunction()) { + // Just throw an exception + return Nan::ThrowError("Need to specify a callback"); + } + + Consumer* consumer = ObjectWrap::Unwrap(info.This()); + + Nan::Callback *callback = new Nan::Callback(info[0].As()); + Nan::AsyncQueueWorker(new Workers::ConsumerConnect(callback, consumer)); + + info.GetReturnValue().Set(Nan::Null()); +} + +NAN_METHOD(Consumer::NodeDisconnect) { + Nan::HandleScope scope; + + if (info.Length() < 1 || !info[0]->IsFunction()) { + // Just throw an exception + return Nan::ThrowError("Need to specify a callback"); + } + + v8::Local cb = info[0].As(); + Nan::Callback *callback = new Nan::Callback(cb); + Consumer* consumer = ObjectWrap::Unwrap(info.This()); + + Nan::AsyncQueueWorker(new Workers::ConsumerDisconnect(callback, consumer)); + info.GetReturnValue().Set(Nan::Null()); +} + + +} // namespace NodeKafka diff --git a/src/consumer.h b/src/consumer.h new file mode 100644 index 00000000..2ae85b8c --- /dev/null +++ b/src/consumer.h @@ -0,0 +1,115 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +#ifndef SRC_CONSUMER_H_ +#define SRC_CONSUMER_H_ + +#include +#include +#include +#include +#include + +#include "deps/librdkafka/src-cpp/rdkafkacpp.h" + +#include "src/common.h" +#include "src/connection.h" +#include "src/callbacks.h" +#include "src/message.h" + +namespace NodeKafka { + +struct consumer_commit_t { + int64_t _offset; + int _partition; + std::string _topic_name; + + consumer_commit_t(std::string, int, int64_t); + consumer_commit_t(); +}; + + +/** + * @brief Consumer v8 wrapped object. + * + * Specializes the connection to wrap a consumer object through compositional + * inheritence. Establishes its prototype in node through `Init` + * + * @sa RdKafka::Handle + * @sa NodeKafka::Client + */ + +class Consumer : public Connection { + public: + friend class NodeKafka::Callbacks::Rebalance; + + static void Init(v8::Local); + static v8::Local NewInstance(v8::Local); + + Baton Connect(); + Baton Disconnect(); + + Baton Unsubscribe(); + bool IsSubscribed(); + + Baton Commit(std::string, int, int64_t); + Baton Commit(); + + Baton RefreshAssignments(); + + bool HasAssignedPartitions(); + int AssignedPartitionCount(); + + void Assign(std::vector); + void Unassign(); + + std::string Name(); + + Baton Subscribe(std::vector); + NodeKafka::Message* Consume(); + + protected: + static Nan::Persistent constructor; + static void New(const Nan::FunctionCallbackInfo& info); + + Consumer(RdKafka::Conf *, RdKafka::Conf *); + ~Consumer(); + + private: + static void part_list_print(const std::vector&); + + std::vector m_partitions; + int m_partition_cnt; + bool m_is_subscribed; + + Callbacks::Consume m_consume_cb; + Callbacks::Rebalance m_rebalance_cb; + + // Node methods + static NAN_METHOD(NodeOnConsume); + static NAN_METHOD(NodeOnRebalance); + static NAN_METHOD(NodeConnect); + static NAN_METHOD(NodeSubscribe); + static NAN_METHOD(NodeSubscribeSync); + static NAN_METHOD(NodeDisconnect); + static NAN_METHOD(NodeAssign); + static NAN_METHOD(NodeUnassign); + static NAN_METHOD(NodeGetAssignments); + static NAN_METHOD(NodeUnsubscribe); + static NAN_METHOD(NodeUnsubscribeSync); + static NAN_METHOD(NodeCommit); + static NAN_METHOD(NodeCommitSync); + + static NAN_METHOD(NodeConsumeLoop); + static NAN_METHOD(NodeConsume); +}; + +} // namespace NodeKafka + +#endif // SRC_CONSUMER_H_ diff --git a/src/errors.cc b/src/errors.cc new file mode 100644 index 00000000..02f4055b --- /dev/null +++ b/src/errors.cc @@ -0,0 +1,46 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +#include + +#include "src/errors.h" + +namespace NodeKafka { + +v8::Local RdKafkaError(const RdKafka::ErrorCode &err) { + // + int code = static_cast(err); + + v8::Local ret = Nan::New(); + + ret->Set(Nan::New("message").ToLocalChecked(), + Nan::New(RdKafka::err2str(err)).ToLocalChecked()); + ret->Set(Nan::New("code").ToLocalChecked(), + Nan::New(code)); + + return ret; +} + +Baton::Baton(const RdKafka::ErrorCode &code) { + m_err = code; +} + +Baton::Baton(void* _data) { + m_err = RdKafka::ERR_NO_ERROR; + m_data = _data; +} + +v8::Local Baton::ToObject() { + return RdKafkaError(m_err); +} + +RdKafka::ErrorCode Baton::err() { + return m_err; +} +} // namespace NodeKafka diff --git a/src/errors.h b/src/errors.h new file mode 100644 index 00000000..cc770ce0 --- /dev/null +++ b/src/errors.h @@ -0,0 +1,45 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +#ifndef SRC_ERRORS_H_ +#define SRC_ERRORS_H_ + +#include +#include +#include + +#include "deps/librdkafka/src-cpp/rdkafkacpp.h" + +#include "src/common.h" + +namespace NodeKafka { + +class Baton { + public: + explicit Baton(const RdKafka::ErrorCode &); + explicit Baton(void* data); + + template T data() { + return static_cast(m_data); + } + + RdKafka::ErrorCode err(); + + v8::Local ToObject(); + + private: + void* m_data; + RdKafka::ErrorCode m_err; +}; + +v8::Local RdKafkaError(const RdKafka::ErrorCode &); + +} // namespace NodeKafka + +#endif // SRC_ERRORS_H_ diff --git a/src/message.cc b/src/message.cc new file mode 100644 index 00000000..a9d1ec26 --- /dev/null +++ b/src/message.cc @@ -0,0 +1,135 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +#include + +#include "src/message.h" + +namespace NodeKafka { + +v8::Local Message::Pack() { + if (IsError()) { + return GetErrorObject(); + } + + v8::Local pack = Nan::New(); + + Nan::Set(pack, Nan::New("message").ToLocalChecked(), ToBuffer()); + Nan::Set(pack, Nan::New("size").ToLocalChecked(), + Nan::New(size)); + Nan::Set(pack, Nan::New("topic").ToLocalChecked(), + Nan::New(topic_name).ToLocalChecked()); + Nan::Set(pack, Nan::New("offset").ToLocalChecked(), + Nan::New(offset)); + Nan::Set(pack, Nan::New("partition").ToLocalChecked(), + Nan::New(partition)); + + return pack; +} + +RdKafka::Message* Message::GetMessage() { + return m_message; +} + +RdKafka::ErrorCode Message::errcode() { + return m_errcode; +} + +Message::Message(const RdKafka::ErrorCode &err) { + stop_running = false; + m_errcode = err; + m_message = NULL; +} + +Message::Message(RdKafka::Message *message): + m_message(message) { + m_errcode = message->err(); + stop_running = false; + + // Starts polling before the partitioner is ready. + // This may be a problem we want to keep record of + switch (m_errcode) { + case RdKafka::ERR_NO_ERROR: + /* Real message */ + size = message->len(); + offset = message->offset(); + + partition = message->partition(); + + topic_name = message->topic_name(); + payload = message->payload(); + + break; + + case RdKafka::ERR__TIMED_OUT: + errstr = message->errstr(); + break; + + case RdKafka::ERR__PARTITION_EOF: + errstr = message->errstr(); + // exit_eof && ++eof_cnt == partition_cnt + break; + + case RdKafka::ERR__UNKNOWN_TOPIC: + case RdKafka::ERR__UNKNOWN_PARTITION: + errstr = message->errstr(); + stop_running = true; + break; + + default: + /* Errors */ + errstr = message->errstr(); + stop_running = true; // We don't know what this is so back out + } +} + +Message::~Message() { + // @todo I think I'd rather check if _message is NULL + if (m_message) { + delete m_message; + } +} + +size_t Message::Size() { + return size; +} + +void Message::Free(char * data, void * hint) { + Message* m = static_cast(hint); + // @note Am I responsible for freeing data as well? + delete m; +} + +v8::Local Message::ToBuffer() { + Nan::MaybeLocal buff = + Nan::NewBuffer(static_cast(Payload()), + static_cast(Size()), + Message::Free, + this); + + return buff.ToLocalChecked(); +} + +bool Message::IsError() { + return m_errcode != RdKafka::ERR_NO_ERROR; +} + +v8::Local Message::GetErrorObject() { + return RdKafkaError(m_errcode); +} + +char* Message::Payload() { + return static_cast(payload); +} + +bool Message::ConsumerShouldStop() { + return stop_running; +} + +} // namespace NodeKafka diff --git a/src/message.h b/src/message.h new file mode 100644 index 00000000..f031d559 --- /dev/null +++ b/src/message.h @@ -0,0 +1,64 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +#ifndef SRC_MESSAGE_H_ +#define SRC_MESSAGE_H_ + +#include +#include + +#include "deps/librdkafka/src-cpp/rdkafkacpp.h" + +#include "src/common.h" +#include "src/errors.h" + +namespace NodeKafka { + +class Message { + public: + explicit Message(RdKafka::Message*); + explicit Message(const RdKafka::ErrorCode &); + ~Message(); + + bool ConsumerShouldStop(); + bool IsSubscribed(); + + char* Payload(); + size_t Size(); + + bool IsError(); + v8::Local GetErrorObject(); + + RdKafka::Message* GetMessage(); + v8::Local Pack(); + + RdKafka::ErrorCode errcode(); + + static void Free(char *, void *); + v8::Local ToBuffer(); + + private: + RdKafka::Message* m_message; + RdKafka::ErrorCode m_errcode; + + size_t size; + int64_t offset; + std::string topic_name; + std::string key; + void* payload; + int partition; + + bool stop_running; + + std::string errstr; +}; + +} // namespace NodeKafka + +#endif // SRC_MESSAGE_H_ diff --git a/src/producer.cc b/src/producer.cc new file mode 100644 index 00000000..0d972553 --- /dev/null +++ b/src/producer.cc @@ -0,0 +1,499 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +#include +#include + +#include "src/producer.h" +#include "src/workers.h" + +namespace NodeKafka { + +/** + * @brief Producer v8 wrapped object. + * + * Wraps the RdKafka::Producer object with compositional inheritence and + * provides methods for interacting with it exposed to node. + * + * The base wrappable RdKafka::Handle deals with most of the wrapping but + * we still need to declare its prototype. + * + * @sa RdKafka::Producer + * @sa NodeKafka::Connection + */ + +Producer::Producer(RdKafka::Conf* gconfig, RdKafka::Conf* tconfig): + Connection(gconfig, tconfig), + m_dr_cb(), + m_partitioner_cb() { + std::string errstr; + + m_gconfig->set("dr_cb", &m_dr_cb, errstr); + } + +Producer::~Producer() { + Disconnect(); +} + +Nan::Persistent Producer::constructor; + +void Producer::Init(v8::Local exports) { + Nan::HandleScope scope; + + v8::Local tpl = Nan::New(New); + tpl->SetClassName(Nan::New("Producer").ToLocalChecked()); + tpl->InstanceTemplate()->SetInternalFieldCount(1); + + /* + * Lifecycle events inherited from NodeKafka::Connection + * + * @sa NodeKafka::Connection + */ + + Nan::SetPrototypeMethod(tpl, "onEvent", NodeOnEvent); + + /* + * @brief Methods to do with establishing state + */ + + Nan::SetPrototypeMethod(tpl, "connect", NodeConnect); + Nan::SetPrototypeMethod(tpl, "disconnect", NodeDisconnect); + Nan::SetPrototypeMethod(tpl, "getMetadata", NodeGetMetadata); + Nan::SetPrototypeMethod(tpl, "poll", NodePoll); + + /* + * Lifecycle events specifically designated for RdKafka::Producer + * + * @sa RdKafka::Producer + */ + + Nan::SetPrototypeMethod(tpl, "onDeliveryReport", NodeOnDelivery); + + /* + * @brief Methods exposed to do with message production + */ + + Nan::SetPrototypeMethod(tpl, "setPartitioner", NodeSetPartitioner); + Nan::SetPrototypeMethod(tpl, "produce", NodeProduce); + Nan::SetPrototypeMethod(tpl, "produceSync", NodeProduceSync); + + // connect. disconnect. resume. pause. get meta data + constructor.Reset(tpl->GetFunction()); + + exports->Set(Nan::New("Producer").ToLocalChecked(), tpl->GetFunction()); +} + +void Producer::New(const Nan::FunctionCallbackInfo& info) { + if (!info.IsConstructCall()) { + return Nan::ThrowError("non-constructor invocation not supported"); + } + + if (info.Length() < 2) { + return Nan::ThrowError("You must supply global and topic configuration"); + } + + if (!info[0]->IsObject()) { + return Nan::ThrowError("Global configuration data must be specified"); + } + + if (!info[1]->IsObject()) { + return Nan::ThrowError("Topic configuration must be specified"); + } + + std::string errstr; + + RdKafka::Conf* gconfig = + Config::Create(RdKafka::Conf::CONF_GLOBAL, info[0]->ToObject(), errstr); + + if (!gconfig) { + return Nan::ThrowError(errstr.c_str()); + } + + RdKafka::Conf* tconfig = + Config::Create(RdKafka::Conf::CONF_TOPIC, info[1]->ToObject(), errstr); + + if (!tconfig) { + // No longer need this since we aren't instantiating anything + delete gconfig; + return Nan::ThrowError(errstr.c_str()); + } + + Producer* producer = new Producer(gconfig, gconfig); + + // Wrap it + producer->Wrap(info.This()); + + // Then there is some weird initialization that happens + // basically it sets the configuration data + // we don't need to do that because we lazy load it + + info.GetReturnValue().Set(info.This()); +} + +v8::Local Producer::NewInstance(v8::Local arg) { + Nan::EscapableHandleScope scope; + + const unsigned argc = 1; + + v8::Local argv[argc] = { arg }; + v8::Local cons = Nan::New(constructor); + v8::Local instance = cons->NewInstance(argc, argv); + + return scope.Escape(instance); +} + + +std::string Producer::Name() { + if (!IsConnected()) { + return std::string(""); + } + return std::string(m_client->name()); +} + +Baton Producer::Connect() { + if (IsConnected()) { + return Baton(RdKafka::ERR_NO_ERROR); + } + + std::string errstr; + + m_client = RdKafka::Producer::create(m_gconfig, errstr); + + if (!m_client) { + // @todo implement errstr into this somehow + return Baton(RdKafka::ERR__STATE); + } + + return Baton(RdKafka::ERR_NO_ERROR); +} + +void Producer::Disconnect() { + if (IsConnected()) { + scoped_mutex_lock lock(m_connection_lock); + // @todo look at hanging + delete m_client; + m_client = NULL; + } + + RdKafka::wait_destroyed(1000); +} + +Baton Producer::Produce(ProducerMessage* msg) { + return Produce(msg->Payload(), msg->Size(), msg->GetTopic(), + msg->partition, msg->key); +} + +Baton Producer::Produce(void* message, size_t size, std::string topic_name, + int32_t partition, std::string *key) { + std::string errstr; + + RdKafka::Topic* topic = + RdKafka::Topic::create(m_client, topic_name, m_tconfig, errstr); + + if (errstr.empty()) { + // Cede ownership of the pointer to this function + return Produce(message, size, topic, partition, key); + } + + // We own the pointer here so we need to free it + free(message); + return Baton(RdKafka::ERR__INVALID_ARG); +} + +Baton Producer::Produce(void* message, size_t size, RdKafka::Topic* topic, + int32_t partition, std::string *key) { + RdKafka::ErrorCode response_code; + + if (IsConnected()) { + scoped_mutex_lock lock(m_connection_lock); + if (IsConnected()) { + RdKafka::Producer* producer = dynamic_cast(m_client); + + response_code = producer->produce(topic, partition, + RdKafka::Producer::RK_MSG_FREE, message, size, NULL, NULL); + + Poll(); + } else { + response_code = RdKafka::ERR__STATE; + } + } else { + response_code = RdKafka::ERR__STATE; + } + + // These topics actually link to the configuration + // they are made from. It's so we can reuse topic configurations + // That means if we delete it here and librd thinks its still linked, + // producing to the same topic will try to reuse it and it will die. + // + // Honestly, we may need to make configuration a first class object + // @todo(Conf needs to be a first class object that is passed around) + // delete topic; + + if (response_code != RdKafka::ERR_NO_ERROR) { + return Baton(response_code); + } + + return Baton(RdKafka::ERR_NO_ERROR); +} + +void Producer::Poll() { + m_client->poll(0); +} + +/* Node exposed methods */ +NAN_METHOD(Producer::NodeProduceSync) { + Nan::HandleScope scope; + + // Need to extract the message data here. + if (info.Length() < 2 || !info[0]->IsObject() || !info[1]->IsObject()) { + // Just throw an exception + return Nan::ThrowError("Need to specify message data and topic"); + } + + v8::Local obj = info[0].As(); + + // Second parameter is a topic config + Topic* topic = ObjectWrap::Unwrap(info[1].As()); + + ProducerMessage* message = new ProducerMessage(obj, topic); + if (message->IsEmpty()) { + if (message->errstr.empty()) { + return Nan::ThrowError("Need to specify a message to send"); + } else { + return Nan::ThrowError(message->errstr.c_str()); + } + } + + Producer* producer = ObjectWrap::Unwrap(info.This()); + + // Make a fake callback for this function to call. + Baton b = producer->Produce(message); + + if (b.err() != RdKafka::ERR_NO_ERROR) { + info.GetReturnValue().Set(b.ToObject()); + } else { + info.GetReturnValue().Set(Nan::True()); + } +} + +NAN_METHOD(Producer::NodeProduce) { + Nan::HandleScope scope; + + // This needs to be offloaded to libuv + // Need to extract the message data here. + if (info.Length() < 2 || !info[0]->IsObject() || !info[1]->IsObject()) { + // Just throw an exception + return Nan::ThrowError("Need to specify message data and topic"); + } + + v8::Local obj = info[0].As(); + + // Second parameter is a topic config + Topic* topic = ObjectWrap::Unwrap(info[1].As()); + + ProducerMessage* message = new ProducerMessage(obj, topic); + if (message->IsEmpty()) { + if (message->errstr.empty()) { + return Nan::ThrowError("Need to specify a message to send"); + } else { + return Nan::ThrowError(message->errstr.c_str()); + } + } + + Producer* producer = ObjectWrap::Unwrap(info.This()); + + if (info.Length() < 3 || !info[2]->IsFunction()) { + return Nan::ThrowError("You must provide a callback"); + } + + v8::Local cb = info[2].As(); + + Nan::Callback * callback = new Nan::Callback(cb); + + Nan::AsyncQueueWorker( + new Workers::ProducerProduce(callback, producer, message)); + info.GetReturnValue().Set(Nan::Null()); +} + +NAN_METHOD(Producer::NodeOnDelivery) { + Nan::HandleScope scope; + + if (info.Length() < 1 || !info[0]->IsFunction()) { + // Just throw an exception + return Nan::ThrowError("Need to specify a callback"); + } + + Producer* producer = ObjectWrap::Unwrap(info.This()); + v8::Local cb = info[0].As(); + + producer->m_dr_cb.dispatcher.AddCallback(cb); + info.GetReturnValue().Set(Nan::True()); +} + +NAN_METHOD(Producer::NodeSetPartitioner) { + Nan::HandleScope scope; + + if (info.Length() < 1 || !info[0]->IsFunction()) { + // Just throw an exception + return Nan::ThrowError("Need to specify a callback"); + } + + Producer* producer = ObjectWrap::Unwrap(info.This()); + v8::Local cb = info[0].As(); + producer->m_partitioner_cb.SetCallback(cb); + info.GetReturnValue().Set(Nan::True()); +} + +NAN_METHOD(Producer::NodeConnect) { + Nan::HandleScope scope; + + if (info.Length() < 1 || !info[0]->IsFunction()) { + // Just throw an exception + return Nan::ThrowError("Need to specify a callback"); + } + + // This needs to be offloaded to libuv + v8::Local cb = info[0].As(); + Nan::Callback *callback = new Nan::Callback(cb); + + Producer* producer = ObjectWrap::Unwrap(info.This()); + Nan::AsyncQueueWorker(new Workers::ProducerConnect(callback, producer)); + + info.GetReturnValue().Set(Nan::Null()); +} + +NAN_METHOD(Producer::NodePoll) { + Nan::HandleScope scope; + + Producer* producer = ObjectWrap::Unwrap(info.This()); + + if (!producer->IsConnected()) { + Nan::ThrowError("Producer is disconnected"); + } else { + producer->Poll(); + info.GetReturnValue().Set(Nan::True()); + } +} + +NAN_METHOD(Producer::NodeDisconnect) { + Nan::HandleScope scope; + + if (info.Length() < 1 || !info[0]->IsFunction()) { + // Just throw an exception + return Nan::ThrowError("Need to specify a callback"); + } + + + v8::Local cb = info[0].As(); + Nan::Callback *callback = new Nan::Callback(cb); + + Producer* producer = ObjectWrap::Unwrap(info.This()); + Nan::AsyncQueueWorker(new Workers::ProducerDisconnect(callback, producer)); + + info.GetReturnValue().Set(Nan::Null()); +} + +/** + * @brief Producer message object. + * + * Handles conversions between v8::Values for topic config and message config + * + * Rather than have the Producer itself worry about v8 I'd rather it be decoupled. + * Similarly to the way we have topics deal with conversions of v8 objects + * rather than the Producer itself, we have this message object deal with + * validating object parameters and getting it ready to send. + * + * @sa RdKafka::Producer::send + * @sa NodeKafka::Consumer::Produce + */ + +ProducerMessage::ProducerMessage(v8::Local obj, Topic * topic): + topic_(topic), + is_empty(true) { + // We have this bad boy now + v8::Local partitionField = Nan::New("partition").ToLocalChecked(); + if (Nan::Has(obj, partitionField).FromMaybe(false)) { + v8::Local partitionVal = + Nan::Get(obj, partitionField).ToLocalChecked(); + + if (!partitionVal->IsNumber()) { + partition = -1; + } else { + partition = Nan::To(partitionVal).FromJust(); + } + + } else { + partition = -1; + } + + if (partition < 0) { + partition = RdKafka::Topic::PARTITION_UA; // this is just -1 + } + + // This one is a buffer + v8::Local field = Nan::New("message").ToLocalChecked(); + if (Nan::Has(obj, field).FromMaybe(false)) { + Nan::MaybeLocal buffer_pre_object = + Nan::Get(obj, field); + + if (buffer_pre_object.IsEmpty()) { + // this is an error object then + // errstr = "Missing message parameter"; + return; + } + + v8::Local buffer_value = buffer_pre_object.ToLocalChecked(); + + if (!node::Buffer::HasInstance(buffer_value)) { + return; + } + + v8::Local buffer_object = buffer_value->ToObject(); + + // v8 handles the garbage collection here so we need to make a copy of + // the buffer or assign the buffer to a persistent handle. + + // I'm not sure which would be the more performant option. I assume + // the persistent handle would be but for now we'll try this one + // which should be more memory-efficient and allow v8 to dispose of the + // buffer sooner + + buffer_length = node::Buffer::Length(buffer_object); + buffer_data = malloc(buffer_length); + memcpy(buffer_data, node::Buffer::Data(buffer_object), buffer_length); + } else { + return; + } + + is_empty = false; + + // partition = GetParameter(obj, "partition", 0); + std::string key_string = GetParameter(obj, "key", ""); + key = &key_string; +} + +ProducerMessage::~ProducerMessage() {} + +bool ProducerMessage::IsEmpty() { + return is_empty; +} + +void* ProducerMessage::Payload() { + return buffer_data; +} + +RdKafka::Topic* ProducerMessage::GetTopic() { + return topic_->toRDKafkaTopic(); +} + +size_t ProducerMessage::Size() { + return buffer_length; +} + +} // namespace NodeKafka diff --git a/src/producer.h b/src/producer.h new file mode 100644 index 00000000..6da5c070 --- /dev/null +++ b/src/producer.h @@ -0,0 +1,85 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +#ifndef SRC_PRODUCER_H_ +#define SRC_PRODUCER_H_ + +#include +#include +#include +#include + +#include "deps/librdkafka/src-cpp/rdkafkacpp.h" + +#include "src/common.h" +#include "src/connection.h" +#include "src/callbacks.h" +#include "src/topic.h" + +namespace NodeKafka { + +class ProducerMessage { + public: + explicit ProducerMessage(v8::Local, NodeKafka::Topic*); + ~ProducerMessage(); + + void* Payload(); + size_t Size(); + bool IsEmpty(); + RdKafka::Topic * GetTopic(); + + std::string errstr; + + Topic * topic_; + int32_t partition; + std::string* key; + + void* buffer_data; + size_t buffer_length; + + bool is_empty; +}; + +class Producer : public Connection { + public: + static void Init(v8::Local); + static v8::Local NewInstance(v8::Local); + + Baton Connect(); + void Disconnect(); + void Poll(); + + Baton Produce(ProducerMessage* msg); + Baton Produce(void*, size_t, std::string, int32_t, std::string*); + Baton Produce(void*, size_t, RdKafka::Topic*, int32_t, std::string*); + std::string Name(); + + protected: + static Nan::Persistent constructor; + static void New(const Nan::FunctionCallbackInfo&); + + Producer(RdKafka::Conf*, RdKafka::Conf*); + ~Producer(); + + private: + static NAN_METHOD(NodeProduceSync); + static NAN_METHOD(NodeProduce); + static NAN_METHOD(NodeOnDelivery); + static NAN_METHOD(NodeSetPartitioner); + static NAN_METHOD(NodeConnect); + static NAN_METHOD(NodeDisconnect); + static NAN_METHOD(NodePoll); + + Callbacks::Delivery m_dr_cb; + Callbacks::Partitioner m_partitioner_cb; +}; + +} // namespace NodeKafka + +#endif // SRC_PRODUCER_H_ diff --git a/src/topic.cc b/src/topic.cc new file mode 100644 index 00000000..fd16feb4 --- /dev/null +++ b/src/topic.cc @@ -0,0 +1,163 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +#include +#include + +#include "src/common.h" +#include "src/connection.h" +#include "src/topic.h" + +namespace NodeKafka { + +/** + * @brief Producer v8 wrapped object. + * + * Wraps the RdKafka::Producer object with compositional inheritence and + * provides methods for interacting with it exposed to node. + * + * The base wrappable RdKafka::Handle deals with most of the wrapping but + * we still need to declare its prototype. + * + * @sa RdKafka::Producer + * @sa NodeKafka::Connection + */ + +Topic::Topic(std::string topic_name, RdKafka::Conf* config, Connection * handle) { // NOLINT + m_topic = + RdKafka::Topic::create(handle->GetClient(), topic_name, config, errstr); +} + +Topic::~Topic() { + delete m_topic; +} + +Nan::Persistent Topic::constructor; + +void Topic::Init(v8::Local exports) { + Nan::HandleScope scope; + + v8::Local tpl = Nan::New(New); + tpl->SetClassName(Nan::New("Topic").ToLocalChecked()); + tpl->InstanceTemplate()->SetInternalFieldCount(1); + + Nan::SetPrototypeMethod(tpl, "name", NodeGetName); + + // connect. disconnect. resume. pause. get meta data + constructor.Reset(tpl->GetFunction()); + + exports->Set(Nan::New("Topic").ToLocalChecked(), tpl->GetFunction()); +} + +void Topic::New(const Nan::FunctionCallbackInfo& info) { + if (!info.IsConstructCall()) { + return Nan::ThrowError("non-constructor invocation not supported"); + } + + if (info.Length() < 3) { + return Nan::ThrowError("Handle, topic name, configuration required"); + } + + if (!info[0]->IsString()) { + return Nan::ThrowError("Topic name must be a string"); + } + + Nan::Utf8String parameterValue(info[0]->ToString()); + std::string topic_name(*parameterValue); + + if (!info[1]->IsObject()) { + return Nan::ThrowError("Configuration data must be specified"); + } + + std::string errstr; + + RdKafka::Conf* config = + Config::Create(RdKafka::Conf::CONF_TOPIC, info[1]->ToObject(), errstr); + + if (!config) { + return Nan::ThrowError(errstr.c_str()); + } + + if (!info[2]->IsObject()) { + return Nan::ThrowError("Client is not of valid type."); + } + + Connection* connection = ObjectWrap::Unwrap(info[2]->ToObject()); + + Topic* topic = new Topic(topic_name, config, connection); + // Wrap it + topic->Wrap(info.This()); + + // Then there is some weird initialization that happens + // basically it sets the configuration data + // we don't need to do that because we lazy load it + + info.GetReturnValue().Set(info.This()); +} + +// handle + +v8::Local Topic::NewInstance(v8::Local arg) { + Nan::EscapableHandleScope scope; + + const unsigned argc = 1; + + v8::Local argv[argc] = { arg }; + v8::Local cons = Nan::New(constructor); + v8::Local instance = cons->NewInstance(argc, argv); + + return scope.Escape(instance); +} + +std::string Topic::name() { + return m_topic->name(); +} + +RdKafka::Topic * Topic::toRDKafkaTopic() { + return m_topic; +} + +/* + +bool partition_available(int32_t partition) { + return topic_->partition_available(partition); +} + +Baton offset_store (int32_t partition, int64_t offset) { + RdKafka::ErrorCode err = topic_->offset_store(partition, offset); + + switch (err) { + case RdKafka::ERR_NO_ERROR: + + break; + default: + + break; + } +} + +*/ + +NAN_METHOD(Topic::NodeGetName) { + Nan::HandleScope scope; + + Topic* topic = ObjectWrap::Unwrap(info.This()); + + info.GetReturnValue().Set(Nan::New(topic->name()).ToLocalChecked()); +} + +NAN_METHOD(Topic::NodePartitionAvailable) { + // @TODO(sparente) +} + +NAN_METHOD(Topic::NodeOffsetStore) { + // @TODO(sparente) +} + +} // namespace NodeKafka diff --git a/src/topic.h b/src/topic.h new file mode 100644 index 00000000..8c2b85cb --- /dev/null +++ b/src/topic.h @@ -0,0 +1,52 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +#ifndef SRC_TOPIC_H_ +#define SRC_TOPIC_H_ + +#include +#include + +#include "deps/librdkafka/src-cpp/rdkafkacpp.h" + +#include "src/config.h" + +namespace NodeKafka { + +class Topic : public Nan::ObjectWrap { + public: + static void Init(v8::Local); + static v8::Local NewInstance(v8::Local arg); + + RdKafka::Topic * toRDKafkaTopic(); + + protected: + static Nan::Persistent constructor; + static void New(const Nan::FunctionCallbackInfo& info); + + static NAN_METHOD(NodeGetMetadata); + + RdKafka::Topic * m_topic; + // TopicConfig * config_; + + std::string errstr; + std::string name(); + + private: + Topic(std::string, RdKafka::Conf *, Connection *); + ~Topic(); + + static NAN_METHOD(NodeGetName); + static NAN_METHOD(NodePartitionAvailable); + static NAN_METHOD(NodeOffsetStore); +}; + +} // namespace NodeKafka + +#endif // SRC_TOPIC_H_ diff --git a/src/workers.cc b/src/workers.cc new file mode 100644 index 00000000..dea53487 --- /dev/null +++ b/src/workers.cc @@ -0,0 +1,743 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +#include +#include + +#include "src/workers.h" + +typedef std::vector BrokerMetadataList; +typedef std::vector PartitionMetadataList; +typedef std::vector TopicMetadataList; + +using NodeKafka::Producer; +using NodeKafka::Connection; +using NodeKafka::Message; + +namespace NodeKafka { +namespace Workers { + +ConnectionMetadata::ConnectionMetadata( + Nan::Callback *callback, Connection* connection, + std::string topic, int timeout_ms) : + ErrorAwareWorker(callback), + connection_(connection), + topic_(topic), + timeout_ms_(timeout_ms) {} + +ConnectionMetadata::~ConnectionMetadata() {} + +void ConnectionMetadata::Execute() { + if (!connection_->IsConnected()) { + SetErrorMessage("You are not connected"); + return; + } + + Baton b = connection_->GetMetadata(topic_, timeout_ms_); + + if (b.err() == RdKafka::ERR_NO_ERROR) { + // No good way to do this except some stupid string delimiting. + // maybe we'll delimit it by a | or something and just split + // the string to create the object + metadata_ = b.data(); + } else { + SetErrorCode(b.err()); + } +} + +void ConnectionMetadata::HandleOKCallback() { + Nan::HandleScope scope; + + const unsigned int argc = 2; + + // This is a big one! + + v8::Local obj = Nan::New(); + + v8::Local broker_data = Nan::New(); + v8::Local topic_data = Nan::New(); + + const BrokerMetadataList* brokers = metadata_->brokers(); // NOLINT + + unsigned int broker_i = 0; + + for (BrokerMetadataList::const_iterator it = brokers->begin(); + it != brokers->end(); ++it, broker_i++) { + // Start iterating over brokers and set the object up + + const RdKafka::BrokerMetadata* x = *it; + + v8::Local current_broker = Nan::New(); + + Nan::Set(current_broker, Nan::New("id").ToLocalChecked(), + Nan::New(x->id())); + Nan::Set(current_broker, Nan::New("host").ToLocalChecked(), + Nan::New(x->host().c_str()).ToLocalChecked()); + Nan::Set(current_broker, Nan::New("port").ToLocalChecked(), + Nan::New(x->port())); + + broker_data->Set(broker_i, current_broker); + } + + unsigned int topic_i = 0; + + const TopicMetadataList* topics = metadata_->topics(); + + for (TopicMetadataList::const_iterator it = topics->begin(); + it != topics->end(); ++it, topic_i++) { + // Start iterating over topics + + const RdKafka::TopicMetadata* x = *it; + + v8::Local current_topic = Nan::New(); + + Nan::Set(current_topic, Nan::New("name").ToLocalChecked(), + Nan::New(x->topic().c_str()).ToLocalChecked()); + + v8::Local current_topic_partitions = Nan::New(); + + const PartitionMetadataList* current_partition_data = x->partitions(); + + unsigned int partition_i = 0; + PartitionMetadataList::const_iterator itt; + + for (itt = current_partition_data->begin(); + itt != current_partition_data->end(); ++itt, partition_i++) { + // partition iterate + const RdKafka::PartitionMetadata* xx = *itt; + + v8::Local current_partition = Nan::New(); + + Nan::Set(current_partition, Nan::New("id").ToLocalChecked(), + Nan::New(xx->id())); + Nan::Set(current_partition, Nan::New("leader").ToLocalChecked(), + Nan::New(xx->leader())); + + const std::vector * replicas = xx->replicas(); + const std::vector * isrs = xx->isrs(); + + std::vector::const_iterator r_it; + std::vector::const_iterator i_it; + + unsigned int r_i = 0; + unsigned int i_i = 0; + + v8::Local current_replicas = Nan::New(); + + for (r_it = replicas->begin(); r_it != replicas->end(); ++r_it, r_i++) { + current_replicas->Set(r_i, Nan::New(*r_it)); + } + + v8::Local current_isrs = Nan::New(); + + for (i_it = isrs->begin(); i_it != isrs->end(); ++i_it, i_i++) { + current_isrs->Set(r_i, Nan::New(*i_it)); + } + + Nan::Set(current_partition, Nan::New("replicas").ToLocalChecked(), + current_replicas); + Nan::Set(current_partition, Nan::New("isrs").ToLocalChecked(), + current_isrs); + + current_topic_partitions->Set(partition_i, current_partition); + } // iterate over partitions + + Nan::Set(current_topic, Nan::New("partitions").ToLocalChecked(), + current_topic_partitions); + + topic_data->Set(topic_i, current_topic); + } // End iterating over topics + + Nan::Set(obj, Nan::New("orig_broker_id").ToLocalChecked(), + Nan::New(metadata_->orig_broker_id())); + + Nan::Set(obj, Nan::New("orig_broker_name").ToLocalChecked(), + Nan::New(metadata_->orig_broker_name()).ToLocalChecked()); + + Nan::Set(obj, Nan::New("topics").ToLocalChecked(), topic_data); + Nan::Set(obj, Nan::New("brokers").ToLocalChecked(), broker_data); + + v8::Local argv[argc] = { Nan::Null(), obj}; + + callback->Call(argc, argv); + + delete metadata_; +} + +void ConnectionMetadata::HandleErrorCallback() { + Nan::HandleScope scope; + + const unsigned int argc = 1; + v8::Local argv[argc] = { GetErrorObject() }; + + callback->Call(argc, argv); +} + +/** + * @brief Producer connect worker. + * + * Easy Nan::AsyncWorker for setting up client connections + * + * @sa RdKafka::Producer::connect + * @sa NodeKafka::Producer::Connect + */ + +ProducerConnect::ProducerConnect(Nan::Callback *callback, Producer* producer): + ErrorAwareWorker(callback), + producer(producer) {} + +ProducerConnect::~ProducerConnect() {} + +void ProducerConnect::Execute() { + Baton b = producer->Connect(); + + if (b.err() != RdKafka::ERR_NO_ERROR) { + SetErrorCode(b.err()); + } +} + +void ProducerConnect::HandleOKCallback() { + Nan::HandleScope scope; + + const unsigned int argc = 2; + + v8::Local obj = Nan::New(); + Nan::Set(obj, Nan::New("name").ToLocalChecked(), + Nan::New(producer->Name()).ToLocalChecked()); + + v8::Local argv[argc] = { Nan::Null(), obj}; + + callback->Call(argc, argv); +} + +void ProducerConnect::HandleErrorCallback() { + Nan::HandleScope scope; + + const unsigned int argc = 1; + v8::Local argv[argc] = { GetErrorObject() }; + + callback->Call(argc, argv); +} + +ProducerDisconnect::ProducerDisconnect(Nan::Callback *callback, + Producer* producer): + ErrorAwareWorker(callback), + producer(producer) {} + +ProducerDisconnect::~ProducerDisconnect() {} + +void ProducerDisconnect::Execute() { + producer->Disconnect(); +} + +void ProducerDisconnect::HandleOKCallback() { + Nan::HandleScope scope; + + const unsigned int argc = 2; + v8::Local argv[argc] = { Nan::Null(), Nan::True()}; + + callback->Call(argc, argv); +} + +void ProducerDisconnect::HandleErrorCallback() { + Nan::HandleScope scope; + + const unsigned int argc = 1; + v8::Local argv[argc] = { Nan::Error(ErrorMessage()) }; + + callback->Call(argc, argv); +} + +ProducerProduce::ProducerProduce( + Nan::Callback *callback, + Producer *producer, + ProducerMessage *message): + ErrorAwareWorker(callback), + producer(producer), + message(message) {} + +ProducerProduce::~ProducerProduce() { + delete message; +} + +void ProducerProduce::Execute() { + Baton b = producer->Produce(message); + + if (b.err() != RdKafka::ERR_NO_ERROR) { + SetErrorCode(b.err()); + } +} + +void ProducerProduce::HandleOKCallback() { + Nan::HandleScope scope; + + const unsigned int argc = 1; + v8::Local argv[argc] = { Nan::Null() }; + + callback->Call(argc, argv); +} + +void ProducerProduce::HandleErrorCallback() { + Nan::HandleScope scope; + + const unsigned int argc = 1; + v8::Local argv[argc] = { GetErrorObject() }; + + callback->Call(argc, argv); +} + +/** + * @brief Consumer connect worker. + * + * Easy Nan::AsyncWorker for setting up client connections + * + * @sa RdKafka::KafkaConsumer::connect + * @sa NodeKafka::Consumer::Connect + */ + +ConsumerConnect::ConsumerConnect(Nan::Callback *callback, Consumer* consumer): + ErrorAwareWorker(callback), + consumer(consumer) {} + +ConsumerConnect::~ConsumerConnect() {} + +void ConsumerConnect::Execute() { + Baton b = consumer->Connect(); + // consumer->Wait(); + + if (b.err() != RdKafka::ERR_NO_ERROR) { + SetErrorCode(b.err()); + } +} + +void ConsumerConnect::HandleOKCallback() { + Nan::HandleScope scope; + + const unsigned int argc = 2; + + // Create the object + v8::Local obj = Nan::New(); + Nan::Set(obj, Nan::New("name").ToLocalChecked(), + Nan::New(consumer->Name()).ToLocalChecked()); + + v8::Local argv[argc] = { Nan::Null(), obj }; + + callback->Call(argc, argv); +} + +void ConsumerConnect::HandleErrorCallback() { + Nan::HandleScope scope; + + const unsigned int argc = 1; + v8::Local argv[argc] = { Nan::Error(ErrorMessage()) }; + + callback->Call(argc, argv); +} + +/** + * @brief Consumer disconnect worker. + * + * Easy Nan::AsyncWorker for disconnecting and cleaning up librdkafka artifacts + * + * @sa RdKafka::KafkaConsumer::disconnect + * @sa NodeKafka::Consumer::Disconnect + */ + +ConsumerDisconnect::ConsumerDisconnect(Nan::Callback *callback, + Consumer* consumer): + ErrorAwareWorker(callback), + consumer(consumer) {} + +ConsumerDisconnect::~ConsumerDisconnect() {} + +void ConsumerDisconnect::Execute() { + Baton b = consumer->Disconnect(); + + if (b.err() != RdKafka::ERR_NO_ERROR) { + SetErrorCode(b.err()); + } +} + +void ConsumerDisconnect::HandleOKCallback() { + Nan::HandleScope scope; + + const unsigned int argc = 2; + v8::Local argv[argc] = { Nan::Null(), Nan::True() }; + + callback->Call(argc, argv); +} + +void ConsumerDisconnect::HandleErrorCallback() { + Nan::HandleScope scope; + + const unsigned int argc = 1; + v8::Local argv[argc] = { GetErrorObject() }; + + callback->Call(argc, argv); +} + +/** + * @brief Consumer subscribe worker. + * + * Easy Nan::AsyncWorker for subscribing to a list of topics + * + * @sa RdKafka::KafkaConsumer::Subscribe + * @sa NodeKafka::Consumer::Subscribe + */ + +ConsumerSubscribe::ConsumerSubscribe(Nan::Callback *callback, + Consumer* consumer, + std::vector topics) : + ErrorAwareWorker(callback), + consumer(consumer), + topics(topics) {} + +ConsumerSubscribe::~ConsumerSubscribe() {} + +void ConsumerSubscribe::Execute() { + Baton b = consumer->Subscribe(topics); + + if (b.err() != RdKafka::ERR_NO_ERROR) { + SetErrorCode(b.err()); + } +} + +void ConsumerSubscribe::HandleOKCallback() { + Nan::HandleScope scope; + + const unsigned int argc = 1; + v8::Local argv[argc] = { Nan::Null() }; + + callback->Call(argc, argv); +} + +void ConsumerSubscribe::HandleErrorCallback() { + Nan::HandleScope scope; + + const unsigned int argc = 1; + v8::Local argv[argc] = { GetErrorObject() }; + + callback->Call(argc, argv); +} + +/** + * @brief Consumer unsubscribe worker. + * + * Easy Nan::AsyncWorker for unsubscribing from the currently subscribed topics + * + * @sa RdKafka::KafkaConsumer::Unsubscribe + * @sa NodeKafka::Consumer::Unsubscribe + */ + +ConsumerUnsubscribe::ConsumerUnsubscribe(Nan::Callback *callback, + Consumer* consumer) : + ErrorAwareWorker(callback), + consumer(consumer) {} + +ConsumerUnsubscribe::~ConsumerUnsubscribe() {} + +void ConsumerUnsubscribe::Execute() { + Baton b = consumer->Unsubscribe(); + + if (b.err() != RdKafka::ERR_NO_ERROR) { + SetErrorCode(b.err()); + } +} + +void ConsumerUnsubscribe::HandleOKCallback() { + Nan::HandleScope scope; + + const unsigned int argc = 1; + v8::Local argv[argc] = { Nan::Null() }; + + callback->Call(argc, argv); +} + +void ConsumerUnsubscribe::HandleErrorCallback() { + Nan::HandleScope scope; + + const unsigned int argc = 1; + v8::Local argv[argc] = { Nan::Error(ErrorMessage()) }; + + callback->Call(argc, argv); +} + +/** + * @brief Consumer get messages worker. + * + * A more complex Nan::AsyncProgressWorker. I made a custom superclass to deal + * with more real-time progress points. Instead of using ProgressWorker, which + * is not time sensitive, this custom worker will poll using libuv and send + * data back to v8 as it comes available without missing any + * + * The actual event runs through a continuous while loop. It stops when the + * consumer is flagged as disconnected or as unsubscribed. + * + * @todo thread-safe isConnected checking + * @note Chances are, when the connection is broken with the way librdkafka works, + * we are shutting down. But we want it to shut down properly so we probably + * need the consumer to have a thread lock that can be used when + * we are dealing with manipulating the `client` + * + * @sa RdKafka::KafkaConsumer::Consume + * @sa NodeKafka::Consumer::GetMessage + */ + +ConsumerConsumeLoop::ConsumerConsumeLoop(Nan::Callback *callback, + Consumer* consumer) : + MessageWorker(callback), + consumer(consumer) {} + +ConsumerConsumeLoop::~ConsumerConsumeLoop() {} + +void ConsumerConsumeLoop::Execute(const ExecutionMessageBus& bus) { + // Do one check here before we move forward + while (consumer->IsConnected() && consumer->IsSubscribed()) { + NodeKafka::Message* message = consumer->Consume(); + if (message->errcode() == RdKafka::ERR__PARTITION_EOF) { + delete message; + usleep(1*1000); + } else if (message->errcode() == RdKafka::ERR__TIMED_OUT) { + // If it is timed out this could just mean there were no + // new messages fetched quickly enough. This isn't really + // an error that should kill us. + // + // But... this error is given when we are disconnecting so + // we need to check that + delete message; + usleep(1000*1000); + } else { + bus.Send(message); + if (message->IsError() || message->ConsumerShouldStop()) { + break; + } + } + } +} + +void ConsumerConsumeLoop::HandleMessageCallback(NodeKafka::Message* msg) { + Nan::HandleScope scope; + + const unsigned int argc = 2; + v8::Local argv[argc]; + + if (msg->IsError()) { + argv[0] = msg->GetErrorObject(); + argv[1] = Nan::Null(); + // Delete message here. If it is not passed to a buffer, we need to get rid + // of it + delete msg; + } else { + argv[0] = Nan::Null(); + argv[1] = msg->Pack(); + } + + callback->Call(argc, argv); +} + +void ConsumerConsumeLoop::HandleOKCallback() { + Nan::HandleScope scope; +} + +void ConsumerConsumeLoop::HandleErrorCallback() { + Nan::HandleScope scope; + + + const unsigned int argc = 1; + v8::Local argv[argc] = { Nan::Error(ErrorMessage()) }; + + callback->Call(argc, argv); +} + +/** + * @brief Consumer get messages worker. + * + * This callback will get a number of message. Can be of use in streams or + * places where you don't want an infinite loop managed in C++land and would + * rather manage it in Node. + * + * @see RdKafka::KafkaConsumer::Consume + * @see NodeKafka::Consumer::GetMessage + */ + +ConsumerConsumeNum::ConsumerConsumeNum(Nan::Callback *callback, + Consumer* consumer, + const uint32_t & num_messages) : + ErrorAwareWorker(callback), + m_consumer(consumer), + m_num_messages(num_messages) {} + +ConsumerConsumeNum::~ConsumerConsumeNum() {} + +void ConsumerConsumeNum::Execute() { + const int max = static_cast(m_num_messages); + for (int i = 0; i < max; i++) { + // Get a message + NodeKafka::Message* message = m_consumer->Consume(); + if (message->IsError()) { + if (message->errcode() != RdKafka::ERR__TIMED_OUT && + message->errcode() != RdKafka::ERR__PARTITION_EOF) { + SetErrorCode(message->errcode()); + usleep(1000); + } + break; + } + + m_messages.push_back(message); + } +} + +void ConsumerConsumeNum::HandleOKCallback() { + Nan::HandleScope scope; + const unsigned int argc = 2; + v8::Local argv[argc]; + argv[0] = Nan::Null(); + + v8::Local returnArray = Nan::New(); + + if (m_messages.size() > 0) { + int i = -1; + for (std::vector::iterator it = m_messages.begin(); + it != m_messages.end(); ++it) { + i++; + NodeKafka::Message* message = *it; + returnArray->Set(i, message->Pack()); + } + } + + argv[1] = returnArray; + + callback->Call(argc, argv); +} + +void ConsumerConsumeNum::HandleErrorCallback() { + Nan::HandleScope scope; + + if (m_messages.size() > 0) { + for (std::vector::iterator it = m_messages.begin(); + it != m_messages.end(); ++it) { + NodeKafka::Message* message = *it; + delete message; + } + } + + const unsigned int argc = 1; + v8::Local argv[argc] = { GetErrorObject() }; + + callback->Call(argc, argv); +} + +/** + * @brief Consumer get message worker. + * + * This callback will get a single message. Can be of use in streams or places + * where you don't want an infinite loop managed in C++land and would rather + * manage it in Node. + * + * @see RdKafka::KafkaConsumer::Consume + * @see NodeKafka::Consumer::GetMessage + */ + +ConsumerConsume::ConsumerConsume(Nan::Callback *callback, + Consumer* consumer) : + ErrorAwareWorker(callback), + consumer(consumer) {} + +ConsumerConsume::~ConsumerConsume() {} + +void ConsumerConsume::Execute() { + _message = consumer->Consume(); + if (_message->IsError()) { + if (_message->errcode() != RdKafka::ERR__TIMED_OUT || + _message->errcode() != RdKafka::ERR__PARTITION_EOF) { + SetErrorMessage(RdKafka::err2str(_message->errcode()).c_str()); + } + } +} + +void ConsumerConsume::HandleOKCallback() { + Nan::HandleScope scope; + + const unsigned int argc = 2; + v8::Local argv[argc]; + argv[0] = Nan::Null(); + if (_message->IsError()) { + argv[1] = Nan::False(); + delete _message; + } else { + argv[1] = _message->Pack(); + } + callback->Call(argc, argv); +} + +void ConsumerConsume::HandleErrorCallback() { + Nan::HandleScope scope; + + const unsigned int argc = 1; + v8::Local argv[argc] = { _message->GetErrorObject() }; + + callback->Call(argc, argv); + + delete _message; +} + +// Commit + +ConsumerCommit::ConsumerCommit(Nan::Callback *callback, + Consumer* consumer, + consumer_commit_t config) : + ErrorAwareWorker(callback), + consumer(consumer), + _conf(config) { + committing_message = true; + } + +ConsumerCommit::ConsumerCommit(Nan::Callback *callback, + Consumer* consumer) : + ErrorAwareWorker(callback), + consumer(consumer) { + committing_message = false; + } + +ConsumerCommit::~ConsumerCommit() {} + +void ConsumerCommit::Execute() { + Baton b(NULL); + + if (committing_message) { + b = consumer->Commit(); + } else { + b = consumer->Commit(_conf._topic_name, _conf._partition, _conf._offset); + } + + if (b.err() != RdKafka::ERR_NO_ERROR) { + SetErrorCode(b.err()); + } +} + +void ConsumerCommit::HandleOKCallback() { + Nan::HandleScope scope; + + const unsigned int argc = 1; + v8::Local argv[argc] = { Nan::Null() }; + + callback->Call(argc, argv); +} + +void ConsumerCommit::HandleErrorCallback() { + Nan::HandleScope scope; + + const unsigned int argc = 1; + v8::Local argv[argc] = { Nan::Error(ErrorMessage()) }; + + callback->Call(argc, argv); +} + +} // namespace Workers +} // namespace NodeKafka diff --git a/src/workers.h b/src/workers.h new file mode 100644 index 00000000..ebd214f7 --- /dev/null +++ b/src/workers.h @@ -0,0 +1,318 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +#ifndef SRC_WORKERS_H_ +#define SRC_WORKERS_H_ + +#include +#include +#include +#include +#include + +#include "src/common.h" +#include "src/producer.h" +#include "src/consumer.h" + +namespace NodeKafka { +namespace Workers { + +class ErrorAwareWorker : public Nan::AsyncWorker { + public: + explicit ErrorAwareWorker(Nan::Callback* callback_) + : Nan::AsyncWorker(callback_) {} + virtual ~ErrorAwareWorker() {} + + virtual void Execute() = 0; + virtual void HandleOKCallback() = 0; + virtual void HandleErrorCallback() = 0; + + protected: + void SetErrorCode(const int & code) { + RdKafka::ErrorCode rd_err = static_cast(code); + SetErrorMessage(RdKafka::err2str(rd_err).c_str()); + m_error_code = code; + } + void SetErrorCode(const RdKafka::ErrorCode & err) { + SetErrorCode(static_cast(err)); + } + + int GetErrorCode() { + return m_error_code; + } + v8::Local GetErrorObject() { + int code = GetErrorCode(); + Baton b = Baton(static_cast(code)); + return b.ToObject(); + } + + int m_error_code; +}; + +class MessageWorker : public Nan::AsyncWorker { + public: + explicit MessageWorker(Nan::Callback* callback_) + : Nan::AsyncWorker(callback_), m_asyncdata() { + m_async = new uv_async_t; + uv_async_init( + uv_default_loop(), + m_async, + m_async_message); + m_async->data = this; + + uv_mutex_init(&m_async_lock); + } + + virtual ~MessageWorker() { + uv_mutex_destroy(&m_async_lock); + } + + void WorkMessage() { + if (!callback) { + return; + } + + std::vector message_queue; + + { + scoped_mutex_lock lock(m_async_lock); + // Copy the vector and empty it + m_asyncdata.swap(message_queue); + } + + for (unsigned int i = 0; i < message_queue.size(); i++) { + HandleMessageCallback(message_queue[i]); + + // we are done with it. it is about to go out of scope + // for the last time so let's just free it up here. can't rely + // on the destructor + } + } + + class ExecutionMessageBus { + friend class MessageWorker; + public: + void Send(NodeKafka::Message* m) const { + that_->Produce_(m); + } + private: + explicit ExecutionMessageBus(MessageWorker* that) : that_(that) {} + MessageWorker* const that_; + }; + + virtual void Execute(const ExecutionMessageBus&) = 0; + virtual void HandleMessageCallback(NodeKafka::Message*) = 0; + + virtual void Destroy() { + uv_close(reinterpret_cast(m_async), AsyncClose_); + } + + private: + void Execute() { + ExecutionMessageBus message_bus(this); + Execute(message_bus); + } + + void Produce_(NodeKafka::Message* m) { + scoped_mutex_lock lock(m_async_lock); + m_asyncdata.push_back(m); + uv_async_send(m_async); + } + NAN_INLINE static NAUV_WORK_CB(m_async_message) { + MessageWorker *worker = static_cast(async->data); + worker->WorkMessage(); + } + + NAN_INLINE static void AsyncClose_(uv_handle_t* handle) { + MessageWorker *worker = static_cast(handle->data); + delete reinterpret_cast(handle); + delete worker; + } + + uv_async_t *m_async; + uv_mutex_t m_async_lock; + std::vector m_asyncdata; +}; + +class ConnectionMetadata : public ErrorAwareWorker { + public: + ConnectionMetadata(Nan::Callback*, NodeKafka::Connection*, std::string, int); + ~ConnectionMetadata(); + + void Execute(); + void HandleOKCallback(); + void HandleErrorCallback(); + + private: + NodeKafka::Connection * connection_; + + std::string topic_; + int timeout_ms_; + + RdKafka::Metadata* metadata_; + + // Now this is the data that will get translated in the OK callback +}; + +class ProducerConnect : public ErrorAwareWorker { + public: + ProducerConnect(Nan::Callback*, NodeKafka::Producer*); + ~ProducerConnect(); + + void Execute(); + void HandleOKCallback(); + void HandleErrorCallback(); + + private: + NodeKafka::Producer * producer; +}; + +class ProducerDisconnect : public ErrorAwareWorker { + public: + ProducerDisconnect(Nan::Callback*, NodeKafka::Producer*); + ~ProducerDisconnect(); + + void Execute(); + void HandleOKCallback(); + void HandleErrorCallback(); + + private: + NodeKafka::Producer * producer; +}; + +class ProducerProduce : public ErrorAwareWorker { + public: + ProducerProduce( + Nan::Callback*, NodeKafka::Producer*, NodeKafka::ProducerMessage*); + ~ProducerProduce(); + + void Execute(); + void HandleOKCallback(); + void HandleErrorCallback(); + + private: + NodeKafka::Producer * producer; + NodeKafka::ProducerMessage * message; +}; + +class ConsumerConnect : public ErrorAwareWorker { + public: + ConsumerConnect(Nan::Callback*, NodeKafka::Consumer*); + ~ConsumerConnect(); + + void Execute(); + void HandleOKCallback(); + void HandleErrorCallback(); + + private: + NodeKafka::Consumer * consumer; +}; + +class ConsumerDisconnect : public ErrorAwareWorker { + public: + ConsumerDisconnect(Nan::Callback*, NodeKafka::Consumer*); + ~ConsumerDisconnect(); + + void Execute(); + void HandleOKCallback(); + void HandleErrorCallback(); + + private: + NodeKafka::Consumer * consumer; +}; + +class ConsumerSubscribe : public ErrorAwareWorker { + public: + ConsumerSubscribe( + Nan::Callback*, NodeKafka::Consumer*, std::vector); + ~ConsumerSubscribe(); + + void Execute(); + void HandleOKCallback(); + void HandleErrorCallback(); + private: + NodeKafka::Consumer * consumer; + std::vector topics; +}; + +class ConsumerUnsubscribe : public ErrorAwareWorker { + public: + ConsumerUnsubscribe( + Nan::Callback*, NodeKafka::Consumer*); + ~ConsumerUnsubscribe(); + + void Execute(); + void HandleOKCallback(); + void HandleErrorCallback(); + private: + NodeKafka::Consumer * consumer; +}; + +class ConsumerConsumeLoop : public MessageWorker { + public: + ConsumerConsumeLoop(Nan::Callback*, NodeKafka::Consumer*); + ~ConsumerConsumeLoop(); + + void Execute(const ExecutionMessageBus&); + void HandleOKCallback(); + void HandleErrorCallback(); + void HandleMessageCallback(NodeKafka::Message*); + private: + NodeKafka::Consumer * consumer; +}; + +class ConsumerConsume : public ErrorAwareWorker { + public: + ConsumerConsume(Nan::Callback*, NodeKafka::Consumer*); + ~ConsumerConsume(); + + void Execute(); + void HandleOKCallback(); + void HandleErrorCallback(); + private: + NodeKafka::Consumer * consumer; + NodeKafka::Message* _message; +}; + +class ConsumerConsumeNum : public ErrorAwareWorker { + public: + ConsumerConsumeNum(Nan::Callback*, NodeKafka::Consumer*, const uint32_t &); + ~ConsumerConsumeNum(); + + void Execute(); + void HandleOKCallback(); + void HandleErrorCallback(); + private: + NodeKafka::Consumer * m_consumer; + const uint32_t m_num_messages; + std::vector m_messages; +}; + +class ConsumerCommit : public ErrorAwareWorker { + public: + ConsumerCommit( + Nan::Callback*, NodeKafka::Consumer*); + ConsumerCommit( + Nan::Callback*, NodeKafka::Consumer*, consumer_commit_t); + ~ConsumerCommit(); + + void Execute(); + void HandleOKCallback(); + void HandleErrorCallback(); + private: + NodeKafka::Consumer * consumer; + consumer_commit_t _conf; + bool committing_message; +}; + +} // namespace Workers + +} // namespace NodeKafka + +#endif // SRC_WORKERS_H_ diff --git a/test/binding.spec.js b/test/binding.spec.js new file mode 100644 index 00000000..7e17f2d3 --- /dev/null +++ b/test/binding.spec.js @@ -0,0 +1,36 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +var addon = require('bindings')('node-librdkafka'); +var t = require('assert'); + +var consumerConfig = { + 'group.id': 'awesome' +}; + +module.exports = { + 'native addon': { + 'exports something': function() { + t.equal(typeof(addon), 'object'); + }, + 'exports valid producer': function() { + t.equal(typeof(addon.Producer), 'function'); + t.throws(addon.Producer); // Requires constructor + t.equal(typeof(new addon.Producer({}, {})), 'object'); + }, + 'exports valid consumer': function() { + t.equal(typeof(addon.KafkaConsumer), 'function'); + t.throws(addon.KafkaConsumer); // Requires constructor + t.equal(typeof(new addon.KafkaConsumer(consumerConfig, {})), 'object'); + }, + 'exports version': function() { + t.ok(addon.librdkafkaVersion); + } + }, +}; diff --git a/test/consumer.spec.js b/test/consumer.spec.js new file mode 100644 index 00000000..ccbd63e8 --- /dev/null +++ b/test/consumer.spec.js @@ -0,0 +1,86 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +var addon = require('bindings')('node-librdkafka'); +var t = require('assert'); + +var client; +var defaultConfig = { + 'client.id': 'kafka-mocha', + 'group.id': 'kafka-mocha-grp', + 'metadata.broker.list': 'localhost:9092' +}; + +module.exports = { + 'Consumer': { + 'afterEach': function() { + client = null; + }, + 'cannot be set without a topic config': function() { + t.throws(function() { + client = new addon.KafkaConsumer(defaultConfig); + }); + }, + 'can be given a topic config': function() { + client = new addon.KafkaConsumer(defaultConfig, {}); + }, + 'throws us an error if we provide an invalid configuration value': function() { + t.throws(function() { + client = new addon.KafkaConsumer({ + 'foo': 'bar' + }); + }, 'should throw because the key is invalid1'); + }, + 'throws us an error if topic config is given something invalid': function() { + t.throws(function() { + client = new addon.KafkaConsumer(defaultConfig, { 'foo': 'bar' }); + }); + }, + 'ignores function arguments for global configuration': function() { + client = new addon.KafkaConsumer({ + 'event_cb': function() {}, + 'group.id': 'mocha-test' + }, {}); + t.ok(client); + }, + 'ignores function arguments for topic configuration': function() { + client = new addon.KafkaConsumer(defaultConfig, { + 'partitioner_cb': function() {} + }); + } + }, + 'KafkaConsumer client': { + 'beforeEach': function() { + client = new addon.KafkaConsumer(defaultConfig, {}); + }, + 'afterEach': function() { + client = null; + }, + 'is an object': function() { + t.equal(typeof(client), 'object'); + }, + 'requires configuration': function() { + t.throws(function() { + return new addon.KafkaConsumer(); + }); + }, + 'has necessary methods from superclass': function() { + var methods = ['connect', 'disconnect', 'onEvent', 'getMetadata']; + methods.forEach(function(m) { + t.equal(typeof(client[m]), 'function', 'Client is missing ' + m + ' method'); + }); + }, + 'has necessary bindings for librdkafka 1:1 binding': function() { + var methods = ['assign', 'unassign', 'subscribe']; + methods.forEach(function(m) { + t.equal(typeof(client[m]), 'function', 'Client is missing ' + m + ' method'); + }); + } + }, +}; diff --git a/test/error.spec.js b/test/error.spec.js new file mode 100644 index 00000000..13587fc8 --- /dev/null +++ b/test/error.spec.js @@ -0,0 +1,8 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ diff --git a/test/index.spec.js b/test/index.spec.js new file mode 100644 index 00000000..13587fc8 --- /dev/null +++ b/test/index.spec.js @@ -0,0 +1,8 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ diff --git a/test/mocha.opts b/test/mocha.opts new file mode 100644 index 00000000..e3ca00a6 --- /dev/null +++ b/test/mocha.opts @@ -0,0 +1 @@ +--ui exports diff --git a/test/mock.js b/test/mock.js new file mode 100644 index 00000000..4a06bc82 --- /dev/null +++ b/test/mock.js @@ -0,0 +1,55 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +'use strict'; + +var net = require('net'); +var util = require('util'); +var Emitter = require('events'); + +function KafkaServer(config) { + if (!(this instanceof KafkaServer)) { + return new KafkaServer(config); + } + + if (config === undefined) { + config = {}; + } else if (typeof config !== 'object') { + throw new TypeError('"config" must be an object'); + } + + Emitter.call(this); + + var self = this; + + this.socket = net.createServer(function(socket) { + socket.end(); + }); //.unref(); + + this.socket.on('error', function(err) { + console.error(err); + }); + + this.socket.listen({ + port: 9092, + host: 'localhost' + }, function() { + self.address = self.socket.address(); + self.emit('ready'); + }); + +} + +util.inherits(KafkaServer, Emitter); + +KafkaServer.prototype.close = function(cb) { + this.socket.close(cb); +}; + +module.exports = KafkaServer; diff --git a/test/producer.spec.js b/test/producer.spec.js new file mode 100644 index 00000000..2147957e --- /dev/null +++ b/test/producer.spec.js @@ -0,0 +1,65 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +var addon = require('bindings')('node-librdkafka'); +var t = require('assert'); +// var Mock = require('./mock'); + +var client; +var defaultConfig = { + 'client.id': 'kafka-mocha', + 'metadata.broker.list': 'localhost:9092', + 'socket.timeout.ms': 250 +}; + +var server; + +module.exports = { + 'Producer client': { + 'beforeEach': function() { + client = new addon.Producer(defaultConfig, {}); + }, + 'afterEach': function() { + client = null; + }, + 'is an object': function() { + t.equal(typeof(client), 'object'); + }, + 'requires configuration': function() { + t.throws(function() { + return new addon.Producer(); + }); + }, + 'has necessary methods from superclass': function() { + var methods = ['connect', 'disconnect', 'onEvent', 'getMetadata']; + methods.forEach(function(m) { + t.equal(typeof(client[m]), 'function', 'Client is missing ' + m + ' method'); + }); + }, + /* + 'with mock server': { + 'before': function(cb) { + server = new Mock(); + server.on('ready', cb); + }, + + 'after': function(cb) { + server.close(cb); + }, + + 'can connect': function(cb) { + var producer = new addon.Producer(defaultConfig, {}); + producer.connect(function(err, info) { + t.ifError(err); + cb(); + }); + } + }*/ + }, +}; diff --git a/test/util/topicReadable.spec.js b/test/util/topicReadable.spec.js new file mode 100644 index 00000000..868133e7 --- /dev/null +++ b/test/util/topicReadable.spec.js @@ -0,0 +1,123 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +var TopicReadable = require('../../lib/util/topicReadable'); +var t = require('assert'); +var Writable = require('stream').Writable; +var Emitter = require('events'); + +var fakeClient; + +module.exports = { + 'TopicReadable stream': { + 'beforeEach': function() { + fakeClient = new Emitter(); + fakeClient.isConnected = function() { + return true; + }; + fakeClient.consume = function(size, cb) { + if (!size) { + cb = size; + } + + t.equal(typeof cb, 'function', + 'Provided callback should always be a function'); + setImmediate(function() { + cb(null, { + message: new Buffer('test'), + offset: 1 + }); + }); + }; + fakeClient.subscribe = function(topics) { + t.equal(Array.isArray(topics), true); + return this; + }; + }, + + 'exports a stream class': function() { + t.equal(typeof(TopicReadable), 'function'); + }, + + 'can be instantiated': function() { + t.equal(typeof new TopicReadable(fakeClient, 'topic', {}), 'object'); + }, + + 'properly reads off the fake client': function(cb) { + var stream = new TopicReadable(fakeClient, 'topic', {}); + stream.on('error', function(err) { + t.fail(err); + }); + stream.once('readable', function() { + var message = stream.read(); + t.notEqual(message, null); + t.ok(Buffer.isBuffer(message.message)); + t.equal('test', message.message.toString()); + t.equal(typeof message.offset, 'number'); + stream.pause(); + cb(); + }); + }, + + 'properly reads correct number of messages but does not stop': function(next) { + var numMessages = 10; + var numReceived = 0; + var numSent = 0; + + fakeClient.consume = function(size, cb) { + if (numSent < numMessages) { + numSent++; + setImmediate(function() { + cb(null, { + message: new Buffer('test'), + offset: 1 + }); + }); + } else { + } + }; + var stream = new TopicReadable(fakeClient, 'topic', {}); + stream.on('error', function(err) { + // Ignore + }); + stream.on('readable', function() { + var message = stream.read(); + numReceived++; + t.notEqual(message, null); + t.ok(Buffer.isBuffer(message.message)); + t.equal(typeof message.offset, 'number'); + if (numReceived === numMessages) { + // give it a second to get an error + next(); + } + }); + }, + + 'can be piped around': function(cb) { + var stream = new TopicReadable(fakeClient, 'topic', {}); + var writable = new Writable({ + write: function(message, encoding, next) { + t.notEqual(message, null); + t.ok(Buffer.isBuffer(message.message)); + t.equal(typeof message.offset, 'number'); + this.cork(); + cb(); + }, + objectMode: true + }); + + stream.pipe(writable); + stream.on('error', function(err) { + t.fail(err); + }); + + } + + }, +}; diff --git a/test/util/topicWritable.spec.js b/test/util/topicWritable.spec.js new file mode 100644 index 00000000..83ff2f0d --- /dev/null +++ b/test/util/topicWritable.spec.js @@ -0,0 +1,135 @@ +/* + * node-rdkafka - Node.js wrapper for RdKafka C/C++ library + * + * Copyright (c) 2016 Blizzard Entertainment + * + * This software may be modified and distributed under the terms + * of the MIT license. See the LICENSE.txt file for details. + */ + +var TopicWritable = require('../../lib/util/topicWritable'); +var t = require('assert'); +var Readable = require('stream').Readable; +var Emitter = require('events'); + +var fakeClient; + +module.exports = { + 'TopicWritable stream': { + 'beforeEach': function() { + fakeClient = new Emitter(); + fakeClient.isConnected = function() { + return true; + }; + fakeClient.Topic = function(name) { + return { + topicName: name + }; + }; + fakeClient.connect = function(opts, cb) { + setImmediate(function() { + this.emit('ready'); + }.bind(this)); + return this; + }; + }, + + 'exports a stream class': function() { + t.equal(typeof(TopicWritable), 'function'); + }, + + 'can be instantiated': function() { + t.equal(typeof new TopicWritable(fakeClient, 'topic', {}), 'object'); + }, + + 'properly reads off the fake client': function(done) { + + var message; + + fakeClient.produce = function(message, cb) { + t.equal(typeof cb, 'function', + 'Provided callback should always be a function'); + t.deepEqual({ topicName: 'topic' }, message.topic); + t.equal(message.message.toString(), 'Awesome'); + t.equal(Buffer.isBuffer(message.message), true); + done(); + }; + + var stream = new TopicWritable(fakeClient, 'topic', {}); + stream.on('error', function(err) { + t.fail(err); + }); + + stream.write(new Buffer('Awesome')); + }, + + 'properly reads more than one message in order': function(done) { + + var message; + var currentMessage = 0; + + fakeClient.produce = function(message, cb) { + currentMessage++; + t.equal(typeof cb, 'function', + 'Provided callback should always be a function'); + t.deepEqual({ topicName: 'topic' }, message.topic); + t.equal(message.message.toString(), 'Awesome' + currentMessage); + t.equal(Buffer.isBuffer(message.message), true); + if (currentMessage === 2) { + done(); + } else { + cb(); + } + }; + + var stream = new TopicWritable(fakeClient, 'topic', {}); + stream.on('error', function(err) { + t.fail(err); + }); + + stream.write(new Buffer('Awesome1')); + stream.write(new Buffer('Awesome2')); + }, + + 'can be piped into a readable': function(done) { + + var message; + var currentMessage = 0; + var iteration = 0; + + var readable = new Readable({ + read: function(size) { + iteration++; + if (iteration > 1) { + + } else { + this.push('Awesome1'); + this.push('Awesome2'); + } + } + }); + + fakeClient.produce = function(message, cb) { + currentMessage++; + t.equal(typeof cb, 'function', + 'Provided callback should always be a function'); + t.deepEqual({ topicName: 'topic' }, message.topic); + t.equal(message.message.toString(), 'Awesome' + currentMessage); + t.equal(Buffer.isBuffer(message.message), true); + if (currentMessage === 2) { + done(); + } else { + cb(); + } + }; + + var stream = new TopicWritable(fakeClient, 'topic', {}); + stream.on('error', function(err) { + t.fail(err); + }); + + readable.pipe(stream); + }, + + }, +};