Details
-
New Feature
-
Status: Resolved
-
Minor
-
Resolution: Fixed
-
None
-
None
-
None
-
Patch Available
Description
As part of the new TermVectorMapper approach to TermVectors, the ensuing patch loads term vectors and stores the term info by position. This should let people directly index into a term vector given a position. Actually, it does it through Maps, b/c the array based bookkeeping is a pain given the way positions are stored.
The map looks like:
Map<String, Map<Integer, TVPositionInfo>>
where the String is the field name, the integer is the position, and TVPositionInfo is a storage mechanism for the terms and offsets that occur at a position. It should handle multiple terms per position (which is always my downfall! )
I have not tested performance of this approach.