Home

Awesome

<img src="https://i.imgur.com/kAvg4w9.png" align="right" />

Banditore

CI Coveralls Status PHPStan level max

Banditore retrieves new releases from your GitHub starred repositories and put them in a RSS feed, just for you.

Requirements

Installation

  1. Clone the project

    git clone https://github.com/j0k3r/banditore.git

  2. Register a new OAuth GitHub application and get the Client ID & Client Secret for the next step (for the Authorization callback URL put http://127.0.0.1:8000/callback)

  3. Install dependencies using Composer and define your parameter during the installation

    APP_ENV=prod composer install -o --no-dev

    If you want to use:

    • Sentry to retrieve all errors, register here and get your dsn (in Project Settings > DSN).

  4. Setup the database

    php bin/console doctrine:database:create -e prod
php bin/console doctrine:schema:create -e prod

  5. Install assets

    nvm install
yarn install

  6. You can now launch the website:

    php bin/console server:run -e prod

    And access it at this address: http://127.0.0.1:8000

Running the instance

Once the website is up, you now have to setup few things to retrieve new releases. You have two choices:

Without RabbitMQ

You just need to define these 2 cronjobs (replace all /path/to/banditore with real value):

# retrieve new release of each repo every 10 minutes
*/10  *   *   *   *   php /path/to/banditore/bin/console -e prod banditore:sync:versions >> /path/to/banditore/var/logs/command-sync-versions.log 2>&1
# sync starred repos of each user every 5 minutes
*/5   *   *   *   *   php /path/to/banditore/bin/console -e prod banditore:sync:starred-repos >> /path/banditore/to/var/logs/command-sync-repos.log 2>&1

With RabbitMQ

  1. You'll need to declare exchanges and queues. Replace guest by the user of your RabbitMQ instance (guest is the default one):
php bin/console messenger:setup-transports -vvv sync_starred_repos
php bin/console messenger:setup-transports -vvv sync_versions
  1. You now have two queues and two exchanges defined:
  1. Enable these 2 cronjobs which will periodically push messages in queues (replace all /path/to/banditore with real value):
# retrieve new release of each repo every 10 minutes
*/10  *   *   *   *   php /path/to/banditore/bin/console -e prod banditore:sync:versions --use_queue >> /path/to/banditore/var/logs/command-sync-versions.log 2>&1
# sync starred repos of each user every 5 minutes
*/5   *   *   *   *   php /path/to/banditore/bin/console -e prod banditore:sync:starred-repos --use_queue >> /path/banditore/to/var/logs/command-sync-repos.log 2>&1
  1. Setup Supervisor using the sample file from the repo. You can copy/paste it into /etc/supervisor/conf.d/ and adjust path. The default file will launch:

Once you've put the file in the supervisor conf repo, run supervisorctl update && supervisorctl start all (update will read your conf, start all will start all workers)

Monitoring

There is a status page available at /status, it returns a json with some information about the freshness of fetched versions:

{
    "latest": {
        "date": "2019-09-17 19:50:50.000000",
        "timezone_type": 3,
        "timezone": "Europe\/Berlin"
    },
    "diff": 1736,
    "is_fresh": true
}

For example, I've setup a check on updown.io to check that status page and if the page contains "is_fresh":true. So I receive an alert when is_fresh is false: which means there is a potential issue on the server.

Running the test suite

If you plan to contribute (you're awesome, I know that :v:), you'll need to install the project in a different way (for example, to retrieve dev packages):

git clone https://github.com/j0k3r/banditore.git
composer install -o
php bin/console doctrine:database:create -e=test
php bin/console doctrine:schema:create -e=test
php bin/console doctrine:fixtures:load --env=test -n
php bin/simple-phpunit -v

By default the test connexion login is root without password. You can change it in app/config/config_test.yml.

How it works

Ok, if you goes that deeper in the readme, it means you're a bit more than interested, I like that.

Retrieving new release / tag

This is the complex part of the app. Here is a simplified solution to achieve it.

New release

It's not as easy as using the /repos/:owner/:repo/releases API endpoint to retrieve latest release for a given repo. Because not all repo owner use that feature (which is a shame in my case).

All information for a release are available on that endpoint:

Check a new release of that repo as example: https://api.github.com/repos/j0k3r/banditore/releases/5770680

New tag

Some owners also use tag which is a bit more complex to retrieve all information because a tag only contains information about the SHA-1 of the commit which was used to make the tag. We only have these information:

Check tag list of swarrot/SwarrotBundle as example: https://api.github.com/repos/swarrot/SwarrotBundle/tags

After retrieving the tag, we need to retrieve the commit to get these information:

Check a commit from the previous tag list as example: https://api.github.com/repos/swarrot/SwarrotBundle/commits/84c7c57622e4666ae5706f33cd71842639b78755

GitHub Client Discovery

This is the most important piece of the app. One thing that I ran though is hitting the rate limit on GitHub. The rate limit for a given authenticated client is 5.000 calls per hour. This limit is never reached when looking for new release (thanks to the conditional requests of the GitHub API) on a daily basis.

But when new user sign in, we need to sync all its starred repositories and also all their releases / tags. And here come the gourmand part:

Let's say the repo:

And keep in mind that some repos got also 1.000+ tags (!!).

To avoid hitting the limit in such case and wait 1 hour to be able to make requests again I created the GitHub Client Discovery class. It aims to find the best client with enough rate limit remain (defined as 50).

Which means, if you have 5 users on the app, you'll be able to make (1 + 5) x 5.000 = 30.000 calls per hour