Ferret

What is it?

ferret is a web scraping system aiming to simplify data extraction from the web for such things like UI testing, machine learning and analytics.
Having its own declarative language, ferret abstracts away technical details and complexity of the underlying technologies, helping to focus on the data itself.
It's extremely portable, extensible and fast.

Show me some code

The following example demonstrates the use of dynamic pages.
First of all, we load the main Google Search page, type search criteria into an input box and then click a search button.
The click action triggers a redirect, so we wait till its end.
Once the page gets loaded, we iterate over all elements in search results and assign the output to a variable.
The final for loop filters out empty elements that might be because of inaccurate use of selectors.

LET google = DOCUMENT("https://www.google.com/", true)

INPUT(google, 'input[name="q"]', "ferret", 25)
CLICK(google, 'input[name="btnK"]')

WAIT_NAVIGATION(google)

FOR result IN ELEMENTS(google, '.g')
    // filter out extra elements like videos and 'People also ask'
    FILTER TRIM(result.attributes.class) == 'g'
    RETURN {
        title: INNER_TEXT(result, 'h3'),
        description: INNER_TEXT(result, '.st'),
        url: INNER_TEXT(result, 'cite')
    }

More examples you can find here

Features

Declarative language
Support of both static and dynamic web pages
Embeddable
Extensible

Motivation

Nowadays data is everything and who owns data - owns the world.
I have worked on multiple data-driven projects where data was an essential part of a system and I realized how cumbersome writing tons of scrapers is.
After some time looking for a tool that would let me to not write a code, but just express what data I need, decided to come up with my own solution.
ferret project is an ambitious initiative trying to bring the universal platform for writing scrapers without any hassle.

Inspiration

FQL (Ferret Query Language) is heavily inspired by AQL (ArangoDB Query Language).
But due to the domain specifics, there are some differences in how things work.

WIP

Be aware, that the project is under heavy development. There is no documentation and some things may change in the final release.
For query syntax, you may go to ArangoDB web site and use AQL docs as docs for FQL - since they are identical.

Installation

Binary

You can download latest binaries from here.

Source code

Production

Go >=1.10
Chrome or Docker

Development

GoDep
GNU Make
ANTLR4 >=4.7.1

go get github.com/MontFerret/ferret

Environment

In order to use all Ferret features, you will need to have Chrome either installed locally or running in Docker. For ease of use we recommend to run Chrome inside a Docker container:

docker pull alpeware/chrome-headless-trunk
docker run -d -p=0.0.0.0:9222:9222 --name=chrome-headless -v /tmp/chromedata/:/data alpeware/chrome-headless-trunk

But if you want to see what's happening during query execution, just start your Chrome with remote debugging port:

chrome.exe --remote-debugging-port=9222

Quick start

Browserless mode

If you want to play with fql and check its syntax, you can run CLI with the following commands:

ferret

ferret will run in REPL mode.

Welcome to Ferret REPL
Please use `Ctrl-D` to exit this program.
>%
>LET doc = DOCUMENT('https://news.ycombinator.com/')
>FOR post IN ELEMENTS(doc, '.storylink')
>RETURN post.attributes.href
>%

Note: symbol % is used to start and end multi-line queries. You also can use the heredoc format.

If you want to execute a query stored in a file, just pass a file name:

ferret ./docs/examples/static-page.fql

cat ./docs/examples/static-page.fql | ferret

ferret < ./docs/examples/static-page.fql

Browser mode

By default, ferret loads HTML pages via HTTP protocol, because it's faster.
But nowadays, there are more and more websites rendered with JavaScript, and therefore, this 'old school' approach does not really work.
For such cases, you may fetch documents using Chrome or Chromium via Chrome DevTools protocol (aka CDP).
First, you need to make sure that you launched Chrome with remote-debugging-port=9222 flag.
Second, you need to pass the address to ferret CLI.

ferret --cdp http://127.0.0.1:9222

NOTE: By default, ferret will try to use this local address as a default one, so it makes sense to explicitly pass the parameter only in case of either different port number or remote address.

Alternatively, you can tell CLI to launch Chrome for you.

ferret --cdp-launch

NOTE: Launch command is currently broken on MacOS.

Once ferret knows how to communicate with Chrome, you can use a function DOCUMENT(url, isDynamic) with true boolean value for dynamic pages:

Welcome to Ferret REPL
Please use `exit` or `Ctrl-D` to exit this program.
>%
>LET doc = DOCUMENT('https://soundcloud.com/charts/top', true)
>WAIT_ELEMENT(doc, '.chartTrack__details', 5000)
>LET tracks = ELEMENTS(doc, '.chartTrack__details')
>FOR track IN tracks
>    LET username = ELEMENT(track, '.chartTrack__username')
>    LET title = ELEMENT(track, '.chartTrack__title')
>    RETURN {
>       artist: username.innerText,
>        track: title.innerText
>    }
>%

Welcome to Ferret REPL
Please use `exit` or `Ctrl-D` to exit this program.
>%
>LET doc = DOCUMENT("https://github.com/", true)
>LET btn = ELEMENT(doc, ".HeaderMenu a")

>CLICK(btn)
>WAIT_NAVIGATION(doc)
>WAIT_ELEMENT(doc, '.IconNav')

>FOR el IN ELEMENTS(doc, '.IconNav a')
>    RETURN TRIM(el.innerText)
>%

Embedded mode

ferret is a very modular system and therefore, can be easily be embedded into your Go application.

package main

import (
	"context"
	"encoding/json"
	"fmt"
	"os"

	"github.com/MontFerret/ferret/pkg/compiler"
	"github.com/MontFerret/ferret/pkg/drivers"
	"github.com/MontFerret/ferret/pkg/drivers/cdp"
	"github.com/MontFerret/ferret/pkg/drivers/http"
)

type Topic struct {
	Name        string `json:"name"`
	Description string `json:"description"`
	URL         string `json:"url"`
}

func main() {
	topics, err := getTopTenTrendingTopics()

	if err != nil {
		fmt.Println(err)
		os.Exit(1)
	}

	for _, topic := range topics {
		fmt.Println(fmt.Sprintf("%s: %s %s", topic.Name, topic.Description, topic.URL))
	}
}

func getTopTenTrendingTopics() ([]*Topic, error) {
	query := `
		LET doc = DOCUMENT("https://github.com/topics")

		FOR el IN ELEMENTS(doc, ".py-4.border-bottom")
			LIMIT 10
			LET url = ELEMENT(el, "a")
			LET name = ELEMENT(el, ".f3")
			LET desc = ELEMENT(el, ".f5")

			RETURN {
				name: TRIM(name.innerText),
				description: TRIM(desc.innerText),
				url: "https://github.com" + url.attributes.href
			}
	`

	comp := compiler.New()

	program, err := comp.Compile(query)

	if err != nil {
		return nil, err
	}

	// create a root context
	ctx := context.Background()

	// enable HTML drivers
	// by default, Ferret Runtime does not know about any HTML drivers
	// all HTML manipulations are done via functions from standard library
	// that assume that at least one driver is available
	ctx = drivers.WithDynamic(ctx, cdp.NewDriver())
	ctx = drivers.WithStatic(ctx, http.NewDriver())

	out, err := program.Run(ctx)

	if err != nil {
		return nil, err
	}

	res := make([]*Topic, 0, 10)

	err = json.Unmarshal(out, &res)

	if err != nil {
		return nil, err
	}

	return res, nil
}

Extensibility

That said, ferret is a very modular system which also allows not only embed it, but extend its standard library.

package main

import (
	"context"
	"encoding/json"
	"fmt"
	"os"
	"strings"

	"github.com/MontFerret/ferret/pkg/compiler"
	"github.com/MontFerret/ferret/pkg/runtime/core"
	"github.com/MontFerret/ferret/pkg/runtime/values"
)

func main() {
	strs, err := getStrings()

	if err != nil {
		fmt.Println(err)
		os.Exit(1)
	}

	for _, str := range strs {
		fmt.Println(str)
	}
}

func getStrings() ([]string, error) {
	// function implements is a type of a function that ferret supports as a runtime function
	transform := func(ctx context.Context, args ...core.Value) (core.Value, error) {
		// it's just a helper function which helps to validate a number of passed args
		err := core.ValidateArgs(args, 1, 1)

		if err != nil {
			// it's recommended to return built-in None type, instead of nil
			return values.None, err
		}

		// this is another helper functions allowing to do type validation
		err = core.ValidateType(args[0], core.StringType)

		if err != nil {
			return values.None, err
		}

		// cast to built-in string type
		str := args[0].(values.String)

		return values.NewString(strings.ToUpper(str.String() + "_ferret")), nil
	}

	query := `
		FOR el IN ["foo", "bar", "qaz"]
			// conventionally all functions are registered in upper case
			RETURN TRANSFORM(el)
	`

	comp := compiler.New()

	if err := comp.RegisterFunction("transform", transform); err != nil {
		return nil, err
	}

	program, err := comp.Compile(query)

	if err != nil {
		return nil, err
	}

	out, err := program.Run(context.Background())

	if err != nil {
		return nil, err
	}

	res := make([]string, 0, 3)

	err = json.Unmarshal(out, &res)

	if err != nil {
		return nil, err
	}

	return res, nil
}

On top of that, you can completely turn off the standard library, bypassing the following option:

comp := compiler.New(compiler.WithoutStdlib())

And after that, you can easily provide your own implementation of functions from standard library.

If you don't need a particular set of functions from standard library, you can turn off the entire stdlib and register separate packages from that:

package main

import (
    "github.com/MontFerret/ferret/pkg/compiler"
    "github.com/MontFerret/ferret/pkg/stdlib/strings"
)

func main() {
    comp := compiler.New(compiler.WithoutStdlib())

    comp.RegisterFunctions(strings.NewLib())
}

Proxy

By default, Ferret does not use any proxies. Partially, due to inability to force Chrome/Chromium (or any other Chrome Devtools Protocol compatible browser) to use a prticular proxy. It should be done during a browser launch.

But you can pass an address of a proxy server you want to use for static pages.

CLI

ferret --proxy=http://localhost:8888 my-query.fql

Code

package main

import (
    "context"
    "encoding/json"
    "fmt"
    "os"
	
    "github.com/MontFerret/ferret/pkg/compiler"
    "github.com/MontFerret/ferret/pkg/drivers"
    "github.com/MontFerret/ferret/pkg/drivers/http"
)

func run(q string) ([]byte, error) {
    proxy := "http://localhost:8888"
    comp := compiler.New()
	program := comp.MustCompile(q)

	// create a root context
	ctx := context.Background()
	// we inform the driver what proxy to use
	ctx = html.WithStatic(
	    ctx,
	    http.NewDriver(http.WithProxy(proxy)),
	)

	return program.Run(ctx)
}

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

README.md

README.md

Ferret

What is it?

Show me some code

Features

Motivation

Inspiration

WIP

Installation

Binary

Source code

Production

Development

Environment

Quick start

Browserless mode

Browser mode

Embedded mode

Extensibility

Proxy

CLI

Code

Files

README.md

Latest commit

History

README.md

File metadata and controls

Ferret

What is it?

Show me some code

Features

Motivation

Inspiration

WIP

Installation

Binary

Source code

Production

Development

Environment

Quick start

Browserless mode

Browser mode

Embedded mode

Extensibility

Proxy

CLI

Code