zhongjin
/
cartodb
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
Branches Tags
${ item.name }
Create tag ${ searchTerm }
Create branch ${ searchTerm }
from '3e95835308'
${ noResults }
cartodb/CONTRIBUTING.md

12 KiB
Raw Blame History Escape

The development tracker for CartoDB is on GitHub: http://github.com/cartodb/cartodb

Bug fixes are best reported as pull requests over there. Features are best discussed on the mailing list: https://groups.google.com/d/forum/cartodb

  1. General
  2. Documentation
  3. Frontend
  4. Submitting contributions

General

Every new feature (as well as bugfixes) should come with a test case. Depending on context different guidelines might apply, see following sections.

Unless you plan to develop frontend code you can serve assets from our CDN instead, make sure the following is set in the config/app_config.yml:

app_assets:

  asset_host: '//cartodb-libs.global.ssl.fastly.net/cartodbui'

Don't forget to restart Rails after you have modified config/app_config.yml.

Documentation

Documentation that don't fit well inline (e.g. high-level stuff) should be placed in the /doc directory.

Frontend

The frontend is really standalone code, but is integrated with/served by the Rails application.

CSS

We use SASS, with .scss format, which are located at app/assets/stylesheets. Grunt is used to compile the files into .css files, instead of the default Sprockets pipeline that Rails provide.

See doc/frontend/README.md for more in-depth documentation.

Also CartoDB makes use of a linter machine for checking possible errors in those stylesheets. Rules are specified in the scss-style.yml file. Once a new Pull Request is started, Hound application will check those SCSS changes for warnings.

JS

CartoDB is built on top of CartoDB.js, which in turns depends on some common libraries, in particular worth mentioning:

Source code is located at lib/assets/javascripts, dependencies at vendor/assets/javascripts.

See doc/frontend/README.md for more in-depth documentation.

We apply semistandard for syntax consistency of all new code at least, it's checked as part of test run. It's recommended to use a linter in your IDE of choice.

Until our guidelines are publically available follow the existing file/directory and style structure.

Update CartoDB.js

Follow these steps to update to get latest changes:

  • go to lib/assets/javascripts/cdb/
  • git checkout develop && git pull
  • go back to root and run grunt cdb
  • commit both the new revision of the submodule and the generated file vendor/assets/javascripts/cartodb.uncompressed.js

Writing & running tests

We use

There are two test suites: one for the old Editor and one for Builder.

The old Editor specs reside in lib/assets/test/spec/cartodb.

The core Builder specs reside in lib/assets/core/test/spec/cartodb3.

Running all tests

To check all the application specs, please run the next command through command line.

grunt test

This grunt task pass both Editor and Builder suites and the linting process.

Run this task to be sure that everything is OK.

Old Editor specs

In order to develop tests for the codebase outside Builder (that is, old Editor and dashboard pages) we advise to run

grunt && grunt dev

After the building process finish, a webpage will show up with a link to the Jasmine page with all the specs. The URL of this page is http://localhost:8089/_SpecRunner.html

Since this is the command for normal developing, as you modify any file you'll see the changes immediately both in the application and in the specs page.

Builder specs

The development of Builder specs is separated from regular development. This means that you can develop new specs or modify the existing ones without having the whole application running. This speeds up the development task.

Another feature of Builder specs is that we only generate the affected ones by default. That means that we check the current branch changes against master branch and only build those specs that are affected by those changes. This way, we pass only the needed subset of specs.

To start specs development type the next command:

grunt affected_specs

After building the whole suite for the first time, a webpage will show up with a link to the Jasmine page with all the specs. This suite is at http://localhost:8088/_SpecRunner-affected.html

Then, the process will watch changes in the codebase and will regenerate the specs as needed. Just refresh the Jasmine page to pass again the tests.

⚠️ Node v6+ is required to run grunt affected_specs

If you prefer to generate all specs anyway, you can pass a flag to the grunt task:

grunt affected_specs --specs=all

This will generate the whole Builder suite, not only the specs affected by the current branch.

Run specs and regular codebase simultaneously

If you want to run simultaneously the application and the specs generation follow these steps:

  1. Open a terminal with Node v0.10.48 (use nvm) and run grunt && grunt dev. This will build the application assets and will watch for changes.

  2. Open a second terminal and use nvm to set the Node version to 6+.

  3. In this second terminal run grunt affected_specs.

  4. You will see in the first terminal that a lot of changes build the bundle again. That's normal. The first step of the point 3 is to copy all needed files, so the watch of grunt dev triggers. Don't worry about it.

  5. That's it. When you change any Builder Javascript file grunt dev will build the application bundle and grunt affected_specs will build the specs.

Running a particular spec

If you only want to run a subset of tests the easiest and fastest way is to use focused specs, but you can also append ?spec=str-matching-a-describe to test URL, or use --filter flag if running tests in a terminal.

Grunt

We use Grunt to automate build tasks related to both CSS and JS.

We use v0.10.x of node (we recommend to use NVM).

Install dependencies using a normal npm install as such:

npm install
npm install -g grunt-cli

Run grunt availabletasks to see available tasks.

First time starting to work you need to run grunt, to build all static assets (will be written to public/assets/:version).

After that, for typical frontend work, it's recommended to run once:

grunt

This will generate all necessary frontend assets, and then:

grunt dev

That enables CSS and JS watchers for rebuilding bundles automatically upon changes.

Note! Make sure config/app_config.yml don't contain the app_assets configuration, i.e.:

# Make sure the following lines are removed, or commented like this:
#app_assets:

#  asset_host: '//cartodb-libs.global.ssl.fastly.net/cartodbui'

Don't forget to restart Rails after you have modified config/app_config.yml.

Backend

Backend is a Rails 3 application, there's no specific workflow you must follow to run it.

Every PR should be covered by tests. If you create a new file please add it to Makefiles. Useful commands:

  • make check: prepare the test database and run the full suite (takes a while).
  • make prepare-test-db: prepare the test database.
  • bundle exec rspec <spec file>: run a spec.
  • bundle exec rspec <spec file>:<line number>: run an specific test.

Once a new Pull Request is started, Hound application will check code style and you should fix them as much as possible (with common sense, no need to honor every rule but now most of them are actually useful to make code more readable).

Submitting contributions

Before opening a pull request (or submitting a contribution) you will need to sign a Contributor License Agreement (CLA) before making a submission, learn more here.

After that, there are several rules you should follow when a new pull request is created:

  • Title has to be descriptive. If you are fixing a bug don't use the ticket title or number.
  • Explain what you have achieved in the description and choose a reviewer (it has to be a CartoDB team member) of your code. If you have doubts, just ask for one.
  • If you change something related with the UI of the application:
    • Add an image or an animation (LiceCap is your friend) about the feature you have just implemented. Or show the change against what it is already done.
    • Change UI assets version, present in package.json file. Minor if it is a bugfixing or a small feature, major when it is a big change.
  • Our linter machine, Hound, should not trigger any warnings about your changes.
  • All tests should pass, both for JS and Ruby.

Development environment accessories

The development environment can be quite slow to start up, but we have some workarounds that help speed it up:

  • Using Zeus to avoid the load times of the Rails environment by keeping it into memory. It provides a very fast execution of rails commands.
  • Using Stellar (database snapshotting tool) in order to quickly reload the test database. This is also useful while testing code, in order to quickly rollback to a previous DB state.

Zeus

  1. Install zeus globally with gem install zeus. This is recommended but not needed. You can also use bundle exec zeus, which is a bit slower.
  2. Start the zeus server. zeus start. This will start preloading the environments. and display a colorful status.
  3. In a different console, run your rails commands prefixed by zeus. For example: zeus c for console, zeus rspec xxx for testing. Run zeus commands for a full list (or check zeus.json).

Notes:

  • If you want to pass ENV variables, pass them to the zeus start process (master), not to the slaves.
  • When testing, you can run TURBO=1 zeus start to enable some extra optimizations for the testing environment (less database cleaning).
  • If your console breaks after running zeus, add something like zeus() { /usr/bin/zeus "$@"; stty sane; } to .bashrc.
  • If using Vagrant and getting errors, check out zeus docs. Basically, you have to run with ZEUSSOCK=/tmp/zeus.sock as an environment variable.

Stellar

  1. Install stellar. Check your distribution packages, or install with pip: pip install stellar
  2. Create a configuration file by running stellar init and following the steps. The connection string is: postgresql://postgres@localhost/carto_db_test. The project name doesn't matter.
  3. Create a clean testing database make prepare-test-db
  4. Create a snapshot: stellar snapshot

Then to use it for testing, you can pass STELLAR=stellar (you can pass the executable path) as an ENV variable and the testing environment will use it to clean up the database (instead of manually truncating tables).

For development

Stellar can also be useful for development, to quickly restore the database to its original configuration. Just create a different configuration (by going to a different directory, stellar always reads stellar.yaml in the current path) for the development environment.

Then, you can use stellar snapshot and stellar restore to take and restore snapshot quickly. Also check stellar list to list current db snapshots and stellar gc to remove old ones.