Support for Linker with Query type

Describe the solution you’d like I’d like to use Lucence query builder to generate my queries instead of the query String. The toString method does not guarantee compatibility and it fails for some cases.

For example :

val addressTerms = Seq("harvard","denver")
 val booleanQueryBuilder = new BooleanQuery.Builder()

addressTerms.foreach(t=>{
          booleanQueryBuilder.add(new FuzzyQuery(new Term("address", t)), BooleanClause.Occur.SHOULD)
        })

val booleanQuery: BooleanQuery = booleanQueryBuilder.setMinimumNumberShouldMatch(addressTerms.size/2).build()
val query = booleanQuery.toString()

Produces the following exception :

(address:harvard~2 address:denver~2 )~3': Encountered " <FUZZY_SLOP> "~3 "" at line 1, column 95.

Doing a search and replace for the token “~” is not really an option because the query won’t behave as expected

[zouzias]

I think you have a valid point here and put some effort on this.

I have a prototype (see below), where you can specify a linker as a function Row => Query. Will this work for you? I tested it a bit and do not get a serialization error!

https://github.com/zouzias/spark-lucenerdd/pull/154/files#diff-4a4a384e7770d218cc77733cd7d485a9R547

Thank you a lot for the feedback / suggestion, I think such an improvement will improve a lot the API of LuceneRDD.

If you agree on the feature, I can quickly make a snapshot release to test.

[yeikel]

I added a little bit more of support in this commit https://github.com/yeikel/spark-lucenerdd/commit/3e408c7fbd884aaeaae14a9add0c4151dd904017.

Could you please review it?