we deferred the particular handling to Drupal’s queue program. That nicely prevented competition ailments around accessing nodes during node salvage and stored the consumer program fast and receptive.
There is another requirement: Because arriving facts is usually unfinished we must furthermore transfer information from RottenTomatoes.com. Regarding we developed a two covering program: One is a simple PHP bundle using the Guzzle collection that indicated Rotten Tomatoes information as PHP objects, even though the additional next bridges that program to create Drupal nodes filled from Rotten Tomatoes facts. We subsequently harmonized Rotten Tomatoes motion pictures and ratings using the customer’s resource data and enabled editors to decide to utilize information from Rotten Tomatoes and only unique where appropriate. That data was actually merged in throughout indexing procedure too, so as soon as information is in Elasticsearch it doesn’t matter where they originated in. We in addition exposed Critic product reviews to Elasticsearch too so as that clients software could read feedback of motion pictures and individual ratings before buying.
Incoming demands from client solutions never hit Drupal. They only previously strike the Silex software host.
The Silex software doesn’t have even doing much. For your cable format we picked the Hypertext software code, or HAL. HAL is a simple JSON-based hypermedia style used by Drupal 8, Zend Appagility, among others, and it is an IETF draft specification. What’s more, it possess a really robust PHP library offered that we were able to use. Since Elasticsearch currently shop and returns JSON it was trivial to map things from Elasticsearch into HAL. The hard work was actually just in drawing and connecting the appropriate hypermedia backlinks and embedded prices. Keyword as well as other lookup queries were just passed right through to Elasticsearch in addition to effects regularly weight the appropriate files.
Finally, we wrapped the HAL item up in Symfony’s reaction item, set the HTTP caching parameters and ETags, and delivered the message returning.
A big benefit of the split-architecture would be that spinning up an innovative new Silex instance is unimportant. There isn’t any meaningful setting beyond identifying the Elasticsearch host to use, and most rule is yanked down via Composer. Meaning rotating up several instances of the API machine for redundancy, high-availability, or abilities is virtually no efforts. We did not need to stress, however; the API was read-only, very with appropriate usage of HTTP headers and a simple Varnish machine in front of they the API was remarkably snappy.
The Upshot
A huge element of Drupal’s readiness as a CMS is actually realizing that it isn’t the be-all end-all response to all troubles.
For Ooyala and its particular consumers, Drupal got great for dealing with contents, not for helping an internet API. The good thing is, Palantir’s comprehension of the future Drupal 8 launch and its reliance about Symfony pipeline let us combine Drupal with Silex – and that is perfect for providing a Web API yet not what hot for controlling and curating content material. In the long run, Palantir find the right instrument for the task, together with project gained using this greatly.
Ooyala presently has a sturdy and dependable API that is in a position to serve customer software we never ever actually touched our selves; Ooyala’s customers have what they want; clients bring an easy and receptive Web service running her mass media applications. In addition, Palantir encountered the opportunity to see our very own fingers dirty with another member of the Symfony families – a financial investment that can pay long-term with Drupal 8 and also the raising rise in popularity of Symfony within PHP environment.
An excellent option for Ooyala; great for Palantir; an excellent option for the community.
Picture by Todd Lappin “Above Suburbia” under CC BY-NC 2.0, changed with greeen overlay and the inclusion of orange county gay escort arrows.