Skip to content

Store pre-aggregated sum and value count in DocValuesSkipper#15737

Open
jainankitk wants to merge 8 commits intoapache:mainfrom
jainankitk:skip-sum
Open

Store pre-aggregated sum and value count in DocValuesSkipper#15737
jainankitk wants to merge 8 commits intoapache:mainfrom
jainankitk:skip-sum

Conversation

@jainankitk
Copy link
Contributor

Description

This PR extends the DocValues skip index to store pre-aggregated sum and valueCount alongside the existing min/max per interval. This enables collectors and faceting operations to leverage pre-computed aggregation data at search time, avoiding the need to iterate through individual doc values.

Issue

Resolves #13870

Signed-off-by: Ankit Jain <jainankitk@apache.org>
Signed-off-by: Ankit Jain <jainankitk@apache.org>
Signed-off-by: Ankit Jain <jainankitk@apache.org>
Signed-off-by: Ankit Jain <jainankitk@apache.org>
Signed-off-by: Ankit Jain <jainankitk@apache.org>
Signed-off-by: Ankit Jain <jainankitk@apache.org>
@github-actions github-actions bot added this to the 10.5.0 milestone Feb 20, 2026
@navneet1v
Copy link
Contributor

@jainankitk will adding these new things in the docvaluesSkipper add an extra burden on the indexing side?

@sgup432
Copy link
Contributor

sgup432 commented Feb 25, 2026

add an extra burden on the indexing side?

I am not sure if you went through the original issue, the entire point is to store pre-aggregated data in the doc values skipper so we can take advantage of it at search time. So that it's intentional.

@navneet1v
Copy link
Contributor

add an extra burden on the indexing side?

I am not sure if you went through the original issue, the entire point is to store pre-aggregated data in the doc values skipper so we can take advantage of it at search time. So that it's intentional.

Thank you. Will look into the issue.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Store pre-aggregated data in sparse indexes

3 participants