As my requirements for the analysis process is pretty unique, I have written multiple custom analyzers and registered them as a plugin in elasticsearch. This is working, but I don't really like the process of having to deploy the plugin for every bugfix/code addition/ES version upgrade, besides patches for my application.
I have read through the docs and there seems no way to deploy a custom analyzer just in my application, am I right? Just an interface between a lucene analyzer in an application and a elastic node.
To accomplish that, I thought about serializing the output of my analyzers to a token format like term:positionInc:posLength and just install an analyzer within elasticsearch which can read this format. This would reduce the frequency of updating the plugin in ES just for version upgrades.
Is that a bad idea? Does a solution like this already exist?