-
Type: Improvement
-
Resolution: Duplicate
-
Priority: Major - P3
-
None
-
Affects Version/s: None
-
Component/s: Performance
-
None
Right now distinct will pull data out of indexes, but will scan the entire thing.
Can skip whole regions where the key is the same
Correct way to implement is probably adding a method to a btree bucket like
getDisinctKeys( int keyOffset )
- depends on
-
SERVER-1673 Allow distinct to use indexes
- Closed
-
SERVER-11939 Implement advanceToNextUniqueKey access method on btree
- Closed
- duplicates
-
SERVER-2094 distinct cheat with indexes
- Closed
- is related to
-
SERVER-13271 remove surplus projection from distinct
- Closed
- related to
-
SERVER-12297 Distinct - nscanned for $in and $or greater than nscannedObjects
- Closed
-
SERVER-9507 Optimize $sort+$group+$first pipeline to avoid full index scan
- Closed