Thanks for your interest in contributing to OpenLayers 3.
Our preferred means of receiving contributions is through pull requests. Make sure that your pull request follows our pull request guidelines below before submitting it.
This page describes what you need to know to contribute code to ol3 as a developer.
Before accepting a contribution, we ask that you provide us a Contributor License Agreement. If you are making your contribution as part of work for your employer, please follow the guidelines on submitting a Corporate Contributor License Agreement. If you are making your contribution as an individual, you can submit a digital Individual Contributor License Agreement.
You will obviously start by forking the ol3 repository.
The Travis CI hook is enabled on the Github repository. This means every pull request is run through a full test suite to ensure it compiles and passes the tests. Failing pull requests will not be merged.
Although not mandatory, it is also recommended to set up Travis CI for your ol3 fork. For that go to your ol3 fork's Service Hooks page and set up the Travis hook. Then every time you push to your fork, the test suite will be run. This means errors can be caught before creating a pull request. For those making small or occasional contributions, this may be enough to check that your contributions are ok; in this case, you do not need to install the build tools on your local environment as described below.
The minimum requirements are:
- Git
- Node.js
- Python 2.6 or 2.7 with a couple of extra modules (see below)
- Java 7 (JRE and JDK)
The executables git
, java
, jar
, and python
should be in your PATH
.
You can check your configuration by running:
$ ./build.py checkdeps
To install the Node.js dependencies run
$ npm install
To install the extra Python modules, run:
$ sudo pip install -r requirements.txt
or
$ cat requirements.txt | sudo xargs easy_install
depending on your OS and Python installation.
As an ol3 developer you will need to use the build.py
Python script. This is
the script to use to run the linter, the compiler, the tests, etc. Windows users
can use build.cmd
which is a thin wrapper around build.py
.
The build.py
script is equivalent to a Makefile. It is actually based on
pake, which is a simple implementation of
make
in Python.
The usage of the script is:
$ ./build.py <target>
where <target>
is the name of the build target you want to execute. For
example:
$ ./build.py test
The main build targets are serve
, lint
, build
, test
, and check
. The
latter is a meta-target that basically runs lint
, build
, and test
.
The serve
target starts a node-based web server, which we will refer to as the dev server. You'll need to start that server for running the examples and the tests in a browser. More information on that further down.
Other targets include apidoc
and ci
. The latter is the target used on Travis CI. See ol3's Travis configuration file.
The check
target is to be run before pushing code to GitHub and opening pull
requests. Branches that don't pass check
won't pass the integration tests,
and have therefore no chance of being merged into master
.
To run the check
target:
$ ./build.py check
If you want to run the full suite of integration tests, see "Running the integration tests" below.
To run the examples you first need to start the dev server:
$ ./build.py serve
Then, just point your browser http://localhost:3000/examples in your browser. For example http://localhost:3000/examples/side-by-side.html.
Run examples against the ol.js
standalone build:
The examples can also be run against the ol.js
standalone lib, just like the examples
hosted on GitHub. Start by
executing the host-examples
build target:
$ ./build.py host-examples
After running host-examples
you can now open the examples index page in the browser, for example: http://localhost/~elemoine/ol3/build/hosted/master/examples/. (This assumes that the hosted
directory is a web directory, served by Apache for example.)
Append ?mode=raw
to make the example work in full debug mode. In raw mode the OpenLayers and Closure Library scripts are loaded individually by the Closure Library's base.js
script (which the example page loads and executes before any other script).
To run the tests in a browser start the dev server (./build.py serve
) and open http://localhost:3000/test/index.html in the browser.
To run the tests on the console (headless testing with PhantomJS) use the test
target:
$ ./build.py test
See also the test-specific README.
When you submit a pull request the Travis continuous integration server will run a full suite of tests, including building all versions of the library and checking that all of the examples work. You will receive an email with the results, and the status will be displayed in the pull request.
To run the full suite of integration tests use the ci
target:
$ ./build.py ci
Running the full suite of integration tests currently takes 5-10 minutes.
This makes sure that your commit won't break the build. It also runs JSDoc3 to make sure that there are no invalid API doc directives.
Adding functionality often implies adding one or several examples. This section provides explanations related to adding examples.
The examples are located in the examples
directory. Adding a new example
implies creating two files in this directory, an .html
file and a .js
file.
See examples/simple.html
and examples/simple.js
for instance.
The .html
file needs to include a script tag with
loader.js?id=<example_name>
as its src
. For example, if the two files for
the example are myexample.js
and myexample.html
then the script tag's src
should be set to myexample
.
You can use simple.js
and simple.html
as templates for new examples.
Short story: the ol3 examples should not use the goog
namespace, except
for goog.require
.
Longer story: we want that the ol3 examples work in multiple modes, with the standalone lib (which has implications of the symbols and properties we export), and compiled together with the ol3 library.
Compiling the examples together with the library makes it mandatory to declare dependencies with goog.require
statements.
Your pull request must:
-
Follow OpenLayers 3's coding style.
-
Pass the integration tests run automatically by the Travis Continuous Integration system.
-
Address a single issue or add a single item of functionality.
-
Contain a clean history of small, incremental, logically separate commits, with no merge commits.
-
Use clear commit messages.
-
Be possible to merge automatically.
It is strongly recommended that you run
$ ./build.py check
before every commit. This will catch many problems quickly, and it is much faster than waiting for the Travis CI integration tests to run.
The check
build target runs a number of quick tests on your code. These
include:
- Lint
- Compile
- Tests
OpenLayers 3 follows Google's JavaScript Style
Guide.
This is checked using the Closure
Linter in strict mode. You
can run the linter locally on your machine before committing using the lint
target to build.py
:
$ ./build.py lint
In addition to fixing problems identified by the linter, please also follow the style of the existing OpenLayers 3 code, which includes:
-
Always wrap the body of
for
,if
, andwhile
statements in braces. -
Class methods should be in alphabetical order.
-
var
declarations should not span multiple lines. If you cannot fit all the declarations in a single line, then start a newvar
declaration on a new line. Within a single line, variables should be declared in alphabetical order. -
Do not use assignments inside expressions.
The integration tests contain a number of automated checks to ensure that the
code follows the OpenLayers 3 style and does not break tests or examples. You
can run the integration tests locally using the ci
target:
$ ./build.py ci
Please submit separate pull requests for separate issues. This allows each to be reviewed on its own merits.
The commit history explains to the reviewer the series of modifications to the code that you have made and breaks the overall contribution into a series of easily-understandable chunks. Any individual commit should not add more than one new class or one new function. Do not submit commits that change thousands of lines or that contain more than one distinct logical change. Trivial commits, e.g. to fix lint errors, should be merged into the commit that introduced the error. See the Atomic Commit Convention on Wikipedia for more detail.
git apply --patch
and git rebase
can help you create a clean commit
history.
Reviewboard.org
and Pro GIT have
explain how to use them.
Commit messages should be short, begin with a verb in the imperative, and contain no trailing punctuation. We follow http://tbaggery.com/2008/04/19/a-note-about-git-commit-messages.html for the formatting of commit messages.
Git commit message should look like:
Header line: explaining the commit in one line
Body of commit message is a few lines of text, explaining things
in more detail, possibly giving some background about the issue
being fixed, etc etc.
The body of the commit message can be several paragraphs, and
please do proper word-wrap and keep columns shorter than about
74 characters or so. That way "git log" will show things
nicely even when it's indented.
Further paragraphs come after blank lines.
Please keep the header line short, no more than 50 characters.
Occasionally other changes to master
might mean that your pull request cannot
be merged automatically. In this case you may need to rebase your branch on a
more recent master
, resolve any conflicts, and git push --force
to update
your branch so that it can be merged automatically.