DEVELOPER.md

Building and Testing Protractor

This document describes building, testing, releasing Protractor and provides an overview of the repository layout.

Prerequisite software

The prerequisite software (Node.js, npm, git, jdk) are the same as for angular. See https://github.com/angular/angular/blob/master/docs/DEVELOPER.md#prerequisite-software

Getting the sources

Fork Protractor from github, then clone your fork with:

git clone git@github.com:<github username>/protractor.git

# Go to the Protractor directory:
cd protractor/

# Add the main protractor repository as an upstream remote to your repository:
git remote add upstream https://github.com/angular/protractor.git

Installing and Building

All Protractor dependencies come from npm. Install with:

npm install

This will also trigger our build step. The build step runs the TypeScript compiler and copies necessary files into the output built directory. To run the build step independently, run:

npm run prepublish

You can see the other available npm scripts in package.json. Note that most of these scripts just call our gulp commands, which can be seen in gulpfile.js.

Formatting

Protractor uses clang-format to format the source code. If the source code is not properly formatted, the CI will fail and the PR can not be merged.

You can automatically format your code by running:

npm run format

You can check that you will pass lint tests with:

gulp lint

# or if you don't have gulp installed globally:
./node_modules/.bin/gulp lint

Code layout

docs/ contains markdown documentation files. lib/ contains the actual Protractor code. scripts/ contains scripts used for CI setup and running tests. spec/ contains e2e and unit tests and configuration files for tests. testapp/ contains the code for the Angular applications that e2e tests run against. website/ contains code for generating Protractor API documentation and the website at protractortest.org.

Most of the code is written in TypeScript, with the exception of a few js files.

lib/driverProviders controls how WebDriver instances are created. lib/frameworks contains adapters for test frameworks such as Jasmine and Mocha. lib/selenium-webdriver and lib/webdriver-js-extender are used ONLY for API documentation generation.

Lightning Code Walkthrough

TBD.

Testing

Run npm test to run the full test suite. This assumes that you have the testapp and a selenium server running. Start these as separate processes with:

webdriver-manager update
webdriver-manager start

and

npm start

This suite is described in scripts/test.js. It uses some small helper functions to run commands as child processes and capture the results, so that we can run protractor commands which should result in failures and verify that we get the expected number and type of failures.

The suite contains unit tests, end to end tests using the built binary, and interactive tests. Interactive tests are for testing browser.pause and element explorer.

End to end tests all have configuration files which live in spec/. Many tests do not need an actual Selenium server backing them and use the mockSelenium configuration, which saves time by not connecting to a real selenium server.

Important dependencies

Protractor has very close dependencies with several other projects under the Angular umbrella:

jasminewd2 is an extension of the Jasmine test framework that adds utilities for working with selenium-webdriver. jasminewd

blocking-proxy is a separate binary, which handles traffic between a test script and webdriver. It can be turned on via a protractor configuration file, and in the future all logic to wait for Angular will be handled through the blocking proxy. blocking-proxy

webdriver-manager is a separate binary which manages installing and starting up the various binaries necessary for running webdriver tests. These binaries include specific drivers for various browsers (e.g. chromedriver) and the selenium standalone server. webdriver-manager

webdriver-js-extender extends selenium-webdriver to add Appium commands. webdriver-js-extender

Continuous Integration

PRs or changes submitted to master will automatically trigger continuous integration on two different services - Travis, and Circle CI. We use Travis for tests run with SauceLabs because we have more vm time on Travis and their integration with SauceLabs is smooth. CircleCI gives us greater control over the vm, which allows us to run tests against local browsers and get better logs.

Travis runs e2e tests via SauceLabs against a variety of browsers. The essential browsers run a more complete test suite, specified by spec/ciFullConf.js. We also run a set of smoke tests against a larger set of browsers, which is allowed to fail - this is configured in spec/ciSmokeConf.js. This is due to flakiness in IE, Safari and older browser versions. We also run a small set of tests using BrowserStack to verify that our integration with their Selenium farm works.

Circle CI runs a slightly modified version of npm test in a single VM. It installs the browsers it needs locally. Circle CI runs unit tests and a set of e2e tests against Chrome.

Releasing

See release.md for full instructions.