Develop

Master

The Federal Election Commission (FEC) releases information to the public about money that's raised and spent in federal elections — that's elections for US President, Senate, and House of Representatives.

Are you interested in seeing how much money a candidate raised? Or spent? How much debt they took on? Who contributed to their campaign? The FEC is the authoritative source for that information.

The new FEC.gov is a collaboration between 18F and the FEC. It aims to make campaign finance information more accessible (and understandable) to all users.

All FEC repositories:

• FEC: a general discussion forum. We compile feedback from the FEC.gov feedback widget here, and this is the best place to submit general feedback.
• openFEC: the first RESTful API for the Federal Election Commission
• swagger-ui: forked repo that generates our interactive API documentation
• openFEC-web-app: the web app for exploring campaign finance data
• fec-style: shared styles and user interface components, including this project's glossary and feedback tools
• fec-cms: this project's content management system (CMS)
• fec-proxy: this is a lightweight app that coordinates the paths between the web app and CMS

We welcome you to explore, make suggestions, and contribute to our code.

• Read our contributing guidelines. Then, file an issue or submit a pull request.
• If you'd rather send us an email, we're thrilled to hear from you!
• Follow our Set up instructions to run the apps on your computer.
• Check out our StoriesonBoard FEC story map to get a sense of the user needs we'll be addressing in the future.

We are always trying to improve our documentation. If you have suggestions or run into problems please file an issue!

1. Ensure you have the following requirements installed:

   • Python 3.5.3 (which includes pip and and a built-in version of virtualenv called pyvenv)
   • The latest long term support (LTS) or stable release of Node.js (which includes npm)
   • PostgreSQL (the latest 9.6 release).
     • Read a Mac OSX tutorial
     • Read a Windows tutorial
     • Read a Linux tutorial (or follow your OS package manager)
   • Elastic Search 2.4 (instructions here

2. Set up your Node environment— learn how to do this with our Javascript Ecosystem Guide.

3. Set up your Python environment— learn how to do this with our Python Ecosystem Guide.

4. Clone this repository.

Use pip to install the Python dependencies:

pip install -r requirements.txt
pip install -r requirements-dev.txt

Use npm to install JavaScript dependencies:

npm install -g swagger-tools
npm install

This repo includes optional post-merge and post-checkout hooks to ensure that dependencies are up to date. If enabled, these hooks will update Python dependencies on checking out or merging changes to requirements.txt. To enable the hooks, run:

invoke add_hooks

To disable, run:

invoke remove_hooks

Before you can run this project locally, you'll need a development database and a test database.

To create these databases, run:

createdb cfdm_test
createdb cfdm_unit_test

Load our sample data into the development database (cfdm_test) by running:

pg_restore --dbname cfdm_test --no-acl --no-owner data/subset.dump
./manage.py update_all

Ignore user does not exist error messages. Everything will still work!

Next you'll need to load some documents into elastic search in order to create the search index. To load statutes into elasticsearch, run:

python manage.py index_statutes

Note: FEC and 18F members can set the SQL connection to one of the RDS boxes with:

export SQLA_CONN=<psql:address-to-box>

Note: An additional setting for connecting to and utilizing mirrors/replica boxes can also be set with:

export SQLA_FOLLOWERS=<psql:address-to-replica-box-1>[,<psql:address-to-replica-box-2>,...]

Note: This is a comma separated (with no spaces) string that contains one or more connection strings to any replicas/mirrors that are setup.

Reach out to a team member to get the actual addresses for all of these connections.

1. Run:

   export FEC_WEB_DEBUG=true
   

   This shows error details and more verbose request logging.

2. Run:

   export FEC_WEB_STYLE_URL=http://localhost:8080/css/styles.css
   export FEC_WEB_API_URL=http://localhost:5000
   export FEC_CMS_URL=http://localhost:8000
   

   These are the default URLs to the other local FEC applications. For complete set-up instructions, explore our documentation for fec-style, openFEC-webb-app, and fec-cms.

   Note: If you modify your local environment to run these applications at a different address, be sure to update these environment variables to match.

3. If you do not login to CloudFoundry with SSO (single sign-on), run:

   export FEC_WEB_USERNAME=<username_of_your_choosing>
   export FEC_WEB_PASSWORD=<password_of_your_choosing>
   

   Create these account credentials to gain full access to the application. You can set them to any username and password of your choosing.

   Note: 18F team members should not set these environment variables. 18F and FEC team members will also have additional environment variables to set up. Please reach out to a team member for detailed information.

4. If you are using database replicas/mirrors you can also restrict connections to them to be asynchronous tasks only by running:

   export SQLA_RESTRICT_FOLLOWER_TRAFFIC_TO_TASKS=enabled
   

Follow these steps every time you want to work on this project locally.

1. If you are using the legal search portion of the site, you will need Elastic Search running. Navigate to the installation folder (eg., elasticsearch-1.7.5) and run:

cd bin
./elasticsearch

2. Start the web server:

   ./manage.py runserver
   

3. View your local version of the site at http://localhost:5000.

We use Celery to schedule periodic tasks— for example, refreshing materialized views and updating incremental aggregates. We use Redis as the Celery message broker.

To work with Celery and Redis locally, install Redis and start a Redis server. By default, we connect to Redis at redis://localhost:6379; if Redis is running at a different URL, set the FEC_REDIS_URL environment variable.

Note: Both the API and Celery worker must have access to the relevant environment variables and services (PostgreSQL, S3).

Running Redis and Celery locally:

redis-server
celery worker --app webservices.tasks

This repo uses pytest.

If the test database server is not the default local Postgres instance, indicate it using:

export SQLA_TEST_CONN=<psql:address-to-box>

Running the tests:

py.test

If you add new tables to the data, you'll need to generate a new subset for testing. We use this nifty subsetting tool: rdbms-subsetter.

To build a new test subset, first delete and recreate the test database:

dropdb cfdm_test
createdb cfdm_test

Now use the build_test invoke task to populate the new database with the new subset:

invoke build_test <source> postgresql://:@/cfdm_test

where source is a valid PostgreSQL connection string. To update the version-controlled test subset after rebuilding, run:

invoke dump postgresql://:@/cfdm_test data/subset.dump

If you haven't used Cloud Foundry in other projects, you'll need to install the Cloud Foundry CLI and the autopilot plugin.

Before deploying, install the Cloud Foundry CLI and the autopilot plugin:

1. Read Cloud Foundry documentation to install Cloud Foundry CLI.

2. Install autopilot by running:

   cf install-plugin autopilot -r CF-Community
   

   Learn more about autopilot.

3. Set environment variables used by the deploy script:

   export FEC_CF_USERNAME=<your_cf_username>
   export FEC_CF_PASSWORD=<your_cf_password>
   

   If these variables aren't set, you'll be prompted for your Cloud Foundry credentials when you deploy the app.

To deploy to Cloud Foundry, run:

invoke deploy

Learn more about Invoke.

The deploy task will detect the appropriate Cloud Foundry space based the current branch. You can override this with the optional --space flag. For example:

invoke deploy --space dev

This command will explicitly target the dev space.

On Cloud Foundry, we use the redis28 service. The Redis service can be created as follows:

cf create-service redis28 standard fec-redis

cf target -s dev
cf cups fec-creds-dev -p '{"SQLA_CONN": "..."}'

To stand up a user-provided credential service that supports both the API and the webapp, ensure that the following keys are set:

• SQLA_CONN
• FEC_WEB_API_KEY
• FEC_WEB_API_KEY_PUBLIC
• FEC_GITHUB_TOKEN
• NEW_RELIC_LICENSE_KEY
• WRITE_AUTHORIZED_TOKENS

Deploys of a single app can be performed manually by targeting the env/space, and specifying the corresponding manifest, as well as the app you want, like so:

cf target -o [dev|stage|prod] && cf push -f manifest_<[dev|stage|prod]>.yml [api|web]

Note: Performing a deploy in this manner will result in a brief period of downtime.

There may be a time when you need to run a command remotely, e.g., a management command to update database schemas. Cloud Foundry currently doesn't support a way of connecting to an app that is running directly, so you need to deploy a one-off app specifically for running commands instead.

To accomplish this, follow these steps:

1. Make sure you're pointing to the correct target space:

   cf target -o fec -s <dev | stage | prod>
   

2. Create a new app manifest for this one-off app (choose whatever file name you wish, ending with the .yml extension):

   cf create-app-manifest api -p <one-off-app-manifest-filename.yml>
   

3. Open the newly created one-off app manifest file and add/modify the following YAML properties to it (be sure to maintain all of the other existing properties):

   name: one-off-app-name
   command: "<your command here, e.g., python manage.py update_all> && sleep infinity"
   no-route: true
   

   Note: the && sleep infinity part is needed as the end of the command you specify so that Cloud Foundry doesn't attempt to redeploy the app once the command finishes.

4. Using the same app name you just specified in the custom manifest file, push your application to Cloud Foundry:

   cf push <one-off-app-name> -f <one-off-app-manifest-filename.yml>
   

Once the app is pushed, you can also tail the logs to make see when the command finishes:

cf logs <one-off-app-name>

When the command you want to run finishes, be sure to stop and delete the app to free resources in Cloud Foundry:

cf stop <one-off-app-name>
cf delete <one-off-app-name>

One other thing you may want to consider doing is adding explicit log statements or an email notification to whatever command you are running so that you know for sure when the command finishes (or errors out). However, please do not check in these custom modifications.

Note: We hope to have a better way of accomplishing this in the future.

Likely only useful for 18F FEC team members

You can SSH directly into the running app container to help troubleshoot or inspect things with the instance(s). Run the following command:

cf ssh <app name>

Where is the name of the application instance you want to connect to. Once you are logged into the remote secure shell, you'll also want to run this command to setup the shell environment correctly:

. /home/vcap/app/bin/cf_env_setup.sh

More information about using SSH with cloud.dov can be found in the cloud.gov SSH documentation.

If you're preparing a release to production, you should also create a changelog. The preferred way to do this is using the changelog generator.

Once installed, run:

github_changelog_generator --since-tag <last public-relase> --t <your-gh-token>

When this finishes, commit the log to the release.

We use git-flow for naming and versioning conventions. Both the API and web app are continuously deployed through Travis CI accordingly.

• Developer creates a feature branch and pushes to origin:

  git flow feature start my-feature
  git push origin feature/my-feature
  

• Reviewer merges feature branch into develop via GitHub

• [auto] develop is deployed to dev

• Developer creates a hotfix branch:

  git flow hotfix start my-hotfix
  

• Reviewer merges hotfix branch into develop and master and pushes to origin:

  git flow hotfix finish my-hotfix
  git checkout master
  git push origin master --follow-tags
  git checkout develop
  git push origin develop
  

• develop is deployed to dev

• master is deployed to prod

• Developer creates a release branch and pushes to origin:

  git flow release start my-release
  git push origin release/my-release
  

• [auto] release/my-release is deployed to stage

• Review of staging

• Issue a pull request to master

• Check if there are any SQL files changed. Depending on where the changes are, you may need to run migrations. Ask the person who made the change what, if anything, you need to run.

• Developer merges release branch into master (and backmerges into develop) and pushes to origin:

  git config --global push.followTags true
  git flow release finish my-release
  

  You'll need to save several merge messages, and add a tag message which is named the name of the release (eg., public-beta-20170118).

  git checkout develop
  git push origin develop
  

  Watch the develop build on travis and make sure it passes. Now you are ready to push to prod (:tada:).

  git checkout master
  git push origin master --follow-tags
  

  Watch travis to make sure it passes, then test the production site manually to make sure everything looks ok.

• master is deployed to prod

• develop is deployed to dev

This section covers a few topics we think might help developers after setup.

The staging and production environments use the API Umbrella for rate limiting, authentication, caching, and HTTPS termination and redirection. Both environments use the FEC_API_WHITELIST_IPS flag to reject requests that are not routed through the API Umbrella.

All API responses are set to expire after one hour (Cache-Control: public, max-age=3600). In production, the API Umbrella will check this response header and cache responses for the specified interval, such that repeated requests to a given endpoint will only reach the Flask application once. This means that responses may be stale for up to an hour following the nightly refresh of the materialized views.

The production and staging environments use relational database service (RDS) instances that receive streaming updates from the FEC database. The development environment uses a separate RDS instance created from a snapshot of the production instance.

Incrementally-updated aggregates and materialized views are updated nightly; see webservices/tasks/refresh.py for details. When the nightly update finishes, logs and error reports are emailed to the development team--specifically, to email addresses specified in FEC_EMAIL_RECIPIENTS.

There are individual management commands for loading individual legal documents. More information is available by invoking each of these commands with a --help option. These commands can be run as tasks on cloud.gov, e.g.,

cf run-task api  "python manage.py reinitialize_all_legal_docs" -m 2G --name reinit-legal

The progress of these tasks can be monitored using, e.g.,

cf logs api | grep reinit-legal

python manage.py index_statutes

python manage.py index_regulations

This command requires that the environment variable FEC_EREGS_API is set to the API endpoint of a valid eregs instance.

python manage.py load_advisory_opinions [-f FROM_AO_NO]

python manage.py load_current_murs [-f FROM_MUR_NO]

python manage.py reinitialize_all_legal_docs

python manage.py refresh_legal_docs_zero_downtime

This command is typically used when there is a schema change. A staging index is built and populated in the background. When ready, the staging index is moved to the production index with no downtime.

The OpenFEC API is a Flask application deployed using the gunicorn WSGI server behind an nginx reverse proxy. Static files are compressed and served directly through nginx; dynamic content is routed to the Flask application via proxy_pass. The entire application is served through the API Umbrella, which handles API keys, caching, and rate limiting.

Sorting fields include a compound index on on the filed to sort and a unique field. Because in cases where there were large amounts of data that had the same value that was being evaluated for sort, the was not a stable sort view for results and the results users received were inconsistent, some records given more than once, others given multiple times.

Database mirrors/replicas are supported by the API if the SQLA_FOLLOWERS is set to one or more valid connection strings. By default, setting this environment variable will shift all read operations to any mirrors/replicas that are available (and randomly choose one to target per request if there are more than one).

You can optionally choose to restrict traffic that goes to the mirrors/replicas to be the asynchronous tasks only by setting the SQLA_RESTRICT_FOLLOWER_TRAFFIC_TO_TASKS environment variable to something that will evaluate to True in Python (simply using True as the value is fine). If you do this, you can also restrict which tasks are supported on the mirrors/replicas. Supported tasks are configured by adding their fully qualified names to the app.config['SQLALCHEMY_FOLLOWER_TASKS'] list in order to whitelist them. By default, only the download task is enabled.