OAK-11603: lucene 4.x fuzzy queries don't work in Elastic

[fabriziofortino]

No description provided.

[alagodasii]

Tests in this repo fail very frequently. Here again. Should it be a warning for us or we can ignore it?

[nfsantos]

Could we log a warning if we find a fuzzy query with the old format? Maybe the log could be throttled, printed only once per run or only once every x times. This could help in migrating to the new fuzzy format.

I think the PR is missing some tests (or do we already have these tests?)

• The ~ character appears in the query but not as part of a fuzzy match.
• The query contains several fuzzy match expressions.
• The query contains a mix of old and new style fuzzy matches.

[fabriziofortino]

@alagodasii CI in this repo fails mostly because it's running with low resources. If you have failures while building the repo locally, please report them.

[fabriziofortino]

@nfsantos

Could we log a warning if we find a fuzzy query with the old format? Maybe the log could be throttled, printed only once per run or only once every x times. This could help in migrating to the new fuzzy format.

I added a log.trace call when this conversion happens. I don't think it would be correct to have a warning here because, so far, there is no plan of deprecating/migrating lucene in favour of elasticsearch or a more recent version of lucene. Queries executed against an Elastic index have to be backward compatible with Lucene queries based on version 4.x but not the other way around. A query with roam~2 would work on Elastic only. We cannot change Lucene to be forward compatible with more recent versions.

I think the PR is missing some tests (or do we already have these tests?)

• The ~ character appears in the query but not as part of a fuzzy match.
• The query contains several fuzzy match expressions.
• The query contains a mix of old and new style fuzzy matches.

I added all these cases.