Add check to directly use ANN Search when filters match all docs. #2320
+106
−4
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
heemin32
reviewed
Dec 11, 2024
heemin32
reviewed
Dec 11, 2024
Signed-off-by: Wei Wang <[email protected]>
weiwang118
force-pushed
the
mainline-wei
branch
from
e05c94f to
fbc6cf9
Compare
weiwang118
requested review from
navneet1v,
VijayanB,
vamshin,
jmazanec15,
naveentatikonda,
junqiu-lei,
martin-gaievski,
ryanbogan,
luyuncheng and
shatejas
as code owners
weiwang118
force-pushed
the
mainline-wei
branch
from
5cebe5c to
d39c6e8
Compare
VijayanB
reviewed
Dec 18, 2024
Signed-off-by: Wei Wang <[email protected]>
weiwang118
force-pushed
the
mainline-wei
branch
from
d39c6e8 to
105c39a
Compare
| * If filters match all docs in this segment, then there is no need to do any extra step | ||
| * and should directly do ANN Search*/ | ||
| if (filterWeight != null && cardinality == maxDoc) { | ||
| return new PerLeafResult(new FixedBitSet(0), doANNSearch(context, new FixedBitSet(0), 0, k)); |
There was a problem hiding this comment.
I see lots of new FixedBitSet(0) in KNNWeight.
Could you factor it out with a meaningful name as a static variable then pass it down to the downstream?
private static final FixedBitSet ... = new FixedBitSet(0);
| @@ -145,6 +146,12 @@ public PerLeafResult searchLeaf(LeafReaderContext context, int k) throws IOExcep | |||
| Map<Integer, Float> result = doExactSearch(context, new BitSetIterator(filterBitSet, cardinality), cardinality, k); | |||
| return new PerLeafResult(filterWeight == null ? null : filterBitSet, result); | |||
| } | |||
| /* | |||
| * If filters match all docs in this segment, then there is no need to do any extra step | |||
There was a problem hiding this comment.
Could you give more context to the comment?? Would like to see why there's no need for extra steps for all doc matches case, and what passing new FixedBitSet(0) means. Basically, we want to save a bitset look up cost when it's not required as possible.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Description
This change aims to resolve this github issue: #1387
by adding a simple check in searchLeaf method so that if the filters match all doc IDs in the segment (Max Doc == size of FilterBitSet), then we directly do ANN search and avoid other unnecessary steps. This will help improve the search performance.
Related Issues
Resolves #1387
Check List
--signoff.By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.
For more information on following Developer Certificate of Origin and signing off your commits, please check here.