By skimming through the Jest documentation is looks like that to perform many useful operation it requires to pull in the ES client as well. Specifically, the ES query builder is very polished and people have been asking for all the capabilities it provides; We might want to wrap that in a custom API but that would force us to constantly keep up with new features being added to ES (not to mention users needing to learn two APIs).

I haven't decided if we should expose those APIs, I was hoping that we would have a better picture after having a POC to play with.. but to get such a prototype quickly I'd avoid coding all of the client side when you can expose the ES query builders to get a good starting point.

ES / master is using Apache Lucene 5.2.1.Final: you could simply build a snapshot and avoid the dependency problem.

I am going to work on this with the goal of providing ElasticSearch integration. Some thoughts/questions:

1. Which client to use?

• ES default

  • Pro: Offers all functionality provided by latest ES

  • Con: Ties to a specific ES and Lucene version which are pulled into the application classpath; The latter is 4.x atm. Makes it a non-viable option really

• Jest

  • Pro: Works with several ES versions; Does not pull in Lucene libraries into client

  • Con: Quite dependency-heavy, specifically Guava is pulled in

• Raw REST/JSON client, e.g. RESTEasy

  • Pro: Probably less dependencies

  • Con: Quite low-level API, not specific to ES domain

For my first steps I will be using Jest.