Fetching contributors…
Cannot retrieve contributors at this time
189 lines (137 sloc) 5 KB

Cookbook of stuff to do while developing on coursier

General note: always explicitly set the scala version at the sbt prompt, like

> ++2.12.4
> ++2.11.11
> ++2.10.6

Some modules of coursier are only built in specific scala versions (sbt plugins in 2.10 and 2.12, cli and web modules in 2.11, …). coursier doesn't use sbt-doge to handle that for now (but any help to make it work would be welcome).

The sources of coursier rely on some git submodules. Clone the sources of coursier via

$ git clone --recursive

or run

$ git submodule update --init --recursive

from the coursier sources to initialize them.

The latter command also needs to be run whenever these submodules are updated.

Compile and run the CLI

$ sbt ++2.11.11 "project cli" pack
$ cli/target/pack/bin/coursier --help

Note: sbt ++2.11.11 cli/pack used to work fine, but doesn't anymore, see

Automatically re-compile the CLI

Doesn't work anymore :/ sbt ++2.11.11 ~cli/pack used to work, but doesn't anymore for now (see above). sbt ++2.11.11 "project cli" ~pack only watches the sources of the cli module, not those of the modules it depends on (core, cache, …).

Run a scripted test of sbt-coursier or sbt-shading

$ sbt
> ++2.12.4
> sbt-plugins/publishLocal
> sbt-coursier/scripted sbt-coursier/simple
> sbt-shading/scripted sbt-shading/shading

++2.12.4 sets the scala version, which automatically builds the plugins for sbt 1.0. For sbt 0.13, do ++2.10.6.

sbt-plugins/publishLocal publishes locally the plugins and their dependencies, which scripted seems not to do automatically.

Run all the scripted tests of sbt-coursier or sbt-shading

$ sbt
> ++2.12.4
> sbt-plugins/publishLocal
> sbt-coursier/scripted
> sbt-shading/scripted

Use ++2.10.6 for sbt 0.13. See discussion above too.

Run unit tests (JVM)

$ sbt
> ++2.12.4
> testsJVM/testOnly coursier.util.TreeTests
> testsJVM/test

testOnly runs the tests that match the expression it is passed. test runs all the tests.

To run the tests each time the sources change, prefix the test commands with ~, like

$ sbt
> ++2.12.4
> ~testsJVM/testOnly coursier.util.TreeTests
> ~testsJVM/test

Run unit tests (JS)

The JS tests require node to be installed. They automatically run npm install from the root of the coursier sources if needed.

JS tests can be run like JVM tests, like

$ sbt
> ++2.12.4
> testsJS/testOnly coursier.util.TreeTests
> testsJS/test

Like for the JVM tests, prefix test commands with ~ to watch sources (see above).

Run integration tests

Main tests

Run the small web repositories with:

$ scripts/ --port 8080 --list-pages
$ scripts/ --port 8081

Both of these commands spawn a web server in the background.

Run the main ITs with

$ sbt ++2.12.4 testsJVM/it:test

Nexus proxy tests

Start the test Nexus servers with

$ scripts/

This spawns two docker-based Nexus servers in the background (a Nexus 2 and a Nexus 3).

Then run the proxy ITs with

$ sbt ++2.12.4 proxy-tests/it:test

Build with Pants

Pants build tool is also added to an experimental path to build the software

Currently only the CLI command can be built via Pants with Scala 2.12.4.

To iterate on code changes:

./pants run cli/src/main/scala-2.12:coursier-cli -- fetch --help

To build a distributable binary

./pants binary cli/src/main/scala-2.12:coursier-cli

# Artifact will be placed under dist/
java -jar dist/coursier-cli.jar fetch --help

Merging PRs on GitHub

Use either "Create merge commit" or "Squash and merge".

Use "Create merge commit" if the commit list is clean enough (each commit has a clear message, and doesn't break simple compilation and test tasks).

Use "Squash and merge" in the other cases.

General Versioning Guideline

  • Major Version 1.x.x : Increment this field when there is a major change.
  • Minor Version x.1.x : Increment this field when there is a minor change that breaks backward compatibility for an method.
  • Patch version x.x.1 : Increment this field when a minor format change that just adds information that an application can safely ignore.

Deprecation Strategy

When deprecating a method/field, we want to know

  1. Since which version this field/method is being deprecated
  2. Migration path, i.e. what to use instead
  3. At which point the deprecation will be removed

Due to scala's builtin deprecation works like

class deprecated(message: String = {}, since: String = {})

we need to put 2) and 3) into message:

@deprecated(message = "<migration path>. <version to be removed>", since: "deprecation start version")

Typically there needs to be at least 2 minor versions between since-version and to-be-removed-version to help migration.

For example, if since version is 1.1.0, then deprecation can be removed in 1.3.0