Wabisabi: A Scala HTTP Client for ElasticSearch
ElasticSearch is pretty great. I found it after spending a lot of time with Solr in production use. ElasticSearch was so great compared to the state of the art at the time that I leapt at the chance to use it and have since used it in a few production deployments.
I’ve also used it in many of my personal projects because it enables such fanastic discovery of information. I occasionally flirt with using it as a database but that’s going to require some more work.
The Client Situation
My initial uses of ElasticSearch (henceforth called ES to save my wrists) leveraged the ability to use an in-VM instance of ES. This was pretty awesome, since I could start ES and use it withouth requiring my users to set up a separate instance. But in production this is a pretty dumb idea, since you’d like to use shards and replicas.
Eventually I decided to leverage an external ES and I used the extensive native Java API. Eventually I wanted a slimmer, asynchronous and Scala-native version. Wabisabi was born!
Wabi-sabi is about the acceptance of imperfection. This name appealed to me for two reasons. The first was that I knew the client would be iterative and not contain all features in it’s first version. The second was that I had no plans to handle any JSON parsing to avoid pushing a specific library on the user.
Deal with it!
How Wabisabi Is Different
Wabisabi is written in Scala. It uses the HTTP API for ElasticSearch. It uses Dispatch to be asynchronous using Futures. No JSON libraries are required and your narrow ass will have to pass it Strings.
Each of these was novel at the time. There are more Scala clients now, so you may want to look at them. I haven’t. ;)
The documentation covers this but it’s nice to reiterate here in a more verbose way. Step one is adding the dependency and resolver to your
Next, you’ll want to set up your connection to the cluster:
I just realized that you might want to connect to one of many instances and providing a single one sort of sucks. Oh well, I’ll have to add that. :)
Next we can perform a search:
Response above is Response from AsyncHttpClient which Dispatch is a wrapper around. Notable methods are
getStatusText. This is an area that could use improvement, as exposing the actual underlying implementation is a shortcoming in my opinion. What if I wanna change?
That’s Not All
Importantly, Wabisabi also implements quite a few of the features that ES exposes. As an example here’s a sample that creates an index, verifies it, applies a mapping and creates an alias! I used this to do upgrades of indices without interrupting reads and writes via aliasing.
Wabisabi is one of my most popular OSS projects and while I’m not currently working on any signficant projects that use ES, I do actively maintain and improve Wabisabi as users send me PRs or issues. If there’s something you’d like to see let me know!