Skip to content

Conversation

@missinglink
Copy link
Member

draft PR to test the effect of enabling docvalues for the source_id field.
this is motivated by the discussion in pelias/api#1608

it is hoped that this field can be used as an additional sorting criteria in order to make the ordering of results with the same _score value more deterministic, and therefore make testing more stable and predictable.

my concerns with this change are:

  • greatly increasing the index size (on disk)
  • increasing the memory consumption (at runtime once docvalues are loaded)

the source_id values are almost entirely unique and non-sequential, so I'd expect to see poor compression.

although.. these concerns are hopefully unwarranted, we might want to check those before merging this.

@missinglink
Copy link
Member Author

Snapshot size comparison:

Screenshot 2022-03-17 at 12 35 15

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

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants