Parliamentary Monitoring Group website

Parliamentary monitoring application for use by the Parliamentary Monitoring Group in Cape Town, South Africa. See: https://www.pmg.org.za.

What does this project do

Allow citizens and other interested parties to monitor what's going on in the South African parliament. With specific focus on tracking the progress of legislation as it moves through the various phases: from being introduced for the first time to finally being approved and signed into law.

The purpose of the project is to improve parliamentary oversight, make the parliamentary process more accessible and improve transparency surrounding the activities of parliament.

How it works

The project consists of the following major components:

• User-facing website, including free and paid-for content (built using Flask, Jinja2 templates, Bootstrap and jQuery)
  • https://pmg.org.za
• Database (PostgreSQL)
• Search engine (Elastic Search)
• Admin interface (Flask-Admin, integration with Mandrill for email notifications)
  • https://pmg.org.za/admin
• API (Flask)
  • https://api.pmg.org.za
  • When this web app connects to its own API, it always connects to 127.0.0.1:5000 and sends the Host header of the api. subdomain of SERVER_HOST to avoid routing "dogfooding" traffic to the outside internet.

Making use of the API

All of the data that is displayed through the frontend website, is served through an API at https://api.pmg.org.za which is freely accessible. However, please note that access to some content on the frontend website is restricted, and the same restrictions apply for the API.

• More details on the API and what it contains are in API.md

Contributing to the project

This project is open-source, and anyone is welcome to contribute. If you just want to make us aware of a bug / make a feature request, then please add a new GitHub Issue (if a similar one does not already exist).

NOTE: On 2015-07-05 we removed some very large files from the repo and its history, reducing the size of the repo from over 100MB to 30MB. This required re-writing the history of the repo. You must pull and rebase your changes.

If you want to contribute to the code, please fork the repository, make your changes, and create a pull request.

Local setup

Build the necessary services:

docker compose build

Setup the database:

docker compose run web python setup_dev_database.py
docker compose run web python app.py db stamp head
docker compose run web python bin/search.py --reindex all

Add the following lines to your .hosts file:

127.0.0.1 pmg.test
127.0.0.1 api.pmg.test

Start the server:

docker compose up

You should now see it running at http://pmg.test:5000/ and http://api.pmg.test:5000/.

You can login with:

user : admin
password : admin

Each time you pull in changes that might contain database changes:

docker compose run web python app.py db migrate
docker compose run web python app.py db upgrade

To delete the database for a completely fresh setup, run:

docker compose down --volumes

To start the task scheduler, run:

docker compose run web python app.py start_scheduler

Developing email features

Run a local mock SMTP server on port 2525

Set the SMTP environment variables

source env.localmail

Running tests

docker compose -f docker-compose.yml -f docker-compose-test.yml run --rm web nosetests tests --exe -v

Code formatting

We use Black to format our code. You can install it using pip install black and run it with:

black app.py bin config pmg tests

Deployment instructions

Deployment is to dokku, a Heroku-like environment. PMG now runs on two servers with an AWS Application Load Balancer.

git remote add dokku-prod-2 [email protected]:pmg-prod
git remote add dokku-prod-3 [email protected]:pmg-prod

To deploy push to one server, wait for it to complete, then push to the other server:

git push dokku-prod-2
git push dokku-prod-3

Reindexing for Search

To re-index all content for search, run:

ssh [email protected] run python bin/search.py --reindex all

This isn't normally necessary as the search index is updated as items are created, updated and deleted. It can be useful when the index has become out of date. Search functionality will fail while the indexing is in progress. Re-indexing takes about 10 minutes.

Task scheduler

Dokku won't automatically start the task scheduler process. To start it, run:

dokku ps:scale pmg worker=1

Database migration

We use Flask-Migrate and Alembic for applying changes to the data model. To setup a migration script:

python app.py db migrate -m "<revision description>"

Then to run the script on your local machine:

python app.py db upgrade

Updating parliamentary days

PMG needs to know the individual days in which Parliament sat, for each year. It uses this information to calculate the number of parliamentary days that it took for bills to be adopted. It reads these days from the file data/parliament-sitting-days.txt.

Updating this information is a two-step process:

1. Update the spreadsheet data/parliament-sitting-days.xlsx that lists the days parliament sits
2. Run python bin/load_parliamentary_days --pm-days data/parliament-sitting-days.xlsx to update data/parliament-sitting-days.txt
3. Run git diff to sanity check the changes
4. Commit the changes

Caching

Application-level caching is used for certain views, initially based on which views the server spends most time on as seen in NewRelic Transaction overview.

To add caching to a view, add the following decorator - it must be the decorator closest to the view method so that it caches the view result, and not the result from other decorators:

from pmg import cache, cache_key, should_skip_cache
...
@cache.memoize(make_name=lambda fname: cache_key(request),
               unless=lambda: should_skip_cache(request, current_user))

Arguments:

• unless must be true when the cache should not be used. Frontend (views.py) views must always use this because the view shows them as logged in, even on pages where the rest of the data is the same. API views that don't serve subscription data or have any user-specific data don't need it.
• make_name must be the cache key for the view. It's very important that query strings are taken into consideration for the cache key.