Currently, facet encoding is very lucene-specific and limited to either string or integer. But with other indexing services (such as Elasticsearch), other types may be needed (such as "date").
We should change the way facet mapping works, with those objectives in mind:
The whole solution should integrate well with Field Bridge 2.0 ( )
The main part of facet mapping should be independent of the indexing service
There should be a way for specific indexing service implementations to automatically derive encoding from the "generic" metadata
There should be a way for users to force using a different encoding? For instance with Lucene, we might want to use string encoding on integers from time to time. Maybe using a speicifc "NumberAsString" field bridge would do the trick: further investigation is required.