RediSearch 1.0 Release Notes
This is a bug-fix release with some stability fixe, a few processing bug fixes, and a few small additions:
Fixed false positives in NOT iterator (5fe948e)
Prevent crashing with uninitialized sorting vectors (96f2473)
Fixed overflow bug in triemap, and limited tag length to 4KB (09e682f)
Fixed false positives in intersects and negative intersects (4998772)
Fixed bug in union of id lists (b8e74ef3)
- Added support for tag prefix completions (137b346)
This is a maintenance release from the 1.0 branch, that does not yet contain the aggregation engine that's already in master. It includes a few small fixes and API additions that were backported from master.
1. Hamming Distance Scoring
RediSearch can now sort the search result by the inverse Hamming Distance between document payloads and the query payloads (provided they are both the same length). This can be used as a nearest neighbor search ranking when a feature vector is encoded as a bitmap and the distance metric is Hamming distance.
2. Wildcard queries
It's now possible to search for just
*, scanning all the documents in the index. This is useful in conjunction with the Hamming Distance scorer.
3. Optional deletion of documents in FT.DEL
Up until now, FT.DEL did not delete the Hash key containing the actual document. As of this version, you can call it with an optional DD (Delete Document) argument, and it will also delete the document.
FT.DEL myIndex myDoc DD.
4. Optionally keep document hashes in FT.DROP
Up until now, FT.DROP always deleted the document Hash keys. As of this version, you can specify the
KEEPDOCS argument to FT.DROP and it will not touch them.
5. Delete geo-sets when dropping an index
Until now we did not delete the geo-set keys where documents were geographically indexed. This version fixes this (#295)