similarseri.es

Predicts what the user would like to watch next by analyzing previously watched shows.

Contributing

A fully-functional local version can be setup quite easily after cloning this repository.

Additional requirements

• Composer
• Bower
• RabbitMQ
• A TMDB-API-key
• Ruby

Typical installation

npm install
composer install
bower install
./app/console assetic:dump
./app/console doctrine:database:create
./app/console doctrine:schema:update --force
./app/console rabbitmq:setup-fabric # Create all needed RabbitMQ-qeues

(Excluding operations such as setting cache permissions and other Symfony generic tasks)

Deploying

Deploying the application is done using Capistrano. Edit the parameters in app/config/deploy.rb and then run:

cap deploy:cold # or cap:deploy if it isn't the first deploy
cap symfony:doctrine:database:create
cap symfony:doctrine:mongodb:schema:update

Understanding queues

Due to limitations in the TMDB-API, fetching fully fetching a show's similar shows require one request per similar show. This is extremely time consuming, which is why they instead are queued up to later on be processed by RabbitMQ consumers.

To consume shows from the queues, run:

./app/console rabbitmq:consumer -m 5 show_fetcher
./app/console rabbitmq:consumer -m 5 show_patcher

In a production enviroment, these should be run from e.g. crontab.

Making it run fast

As previosuly mentioned, fetching shows from the TMDB-API is slow. To make it run fast for users who adds uncached(*) shows, we can guess what shows will be like added. This is done by the similarseries:show:fetch <source> <count> command. E.g.:

./app/console similarseries:show:fetch popular 500

*An uncached show is a show which hasn't previosuly been cached after being downloaded by another user within the last 90 days.