Siddhant Deshmukh

Right, looks like the time complexity will be pretty much similar in both the approaches (min and max heap approach and best effort approach). The storage impact will be more...

Inspired by the algorithm proposed [here](https://github.com/opensearch-project/OpenSearch/issues/13357#issuecomment-2157563513), here is the algorithm used in the [PR](https://github.com/opensearch-project/query-insights/pull/66): ### Data Structures ``` /** * Map storing groupingId to Tuple containing Aggregate search query record...

Resolved in the following : https://github.com/opensearch-project/query-insights/pull/66

resolved in https://github.com/opensearch-project/query-insights/pull/66

> My thinking is that as soon as you apply approach 2 or 3, it should override any identity passed in the the request. Agreed with this approach. Approach 1...

> Is this actually faster? Got any micro benchmarks? Did not do any benchmarking and this is based on https://github.com/opensearch-project/OpenSearch/pull/12131 that I am helping merge.

Ran some benchmarks by creating a cluster with my local changes and using opensearch benchmarks: ``` ------------------------------------------------------ _______ __ _____ / ____(_)___ ____ _/ / / ___/_________ ________ / /_...

Ran a manual benchmark on my VM to add 100000 documents to an index and run 1000 filter_path queries on the index. I then calculated the total time taken for...

> @deshsidd - Thank you for sharing these results. Can you also share the query, just curious about the `query shape (wink)`? ``` filter_path=$(printf 'h.%0.s' {1..20000}) curl -s "$OPENSEARCH_URL/$INDEX_NAME/_search?pretty&filter_path=$filter_path" |...

Closing in favor of : https://github.com/opensearch-project/OpenSearch/pull/14528/files