kotman12 opened a new pull request, #4279:
URL: https://github.com/apache/solr/pull/4279
# Description
The upgrade core index endpoint doesn't currently support child docs and so
it guards against running on an index with such documents. The logic to check
for the existence of child docs contingent upon the `_root_` field existing
could be improved. If you just compare the doc count of `_root_` vs the unique
count (current behavior), it will falsely flag any segment with updated docs
(where the update happened within one segment) as containing child docs when in
actuality it may have no child docs. I actually ran into such a scenario
devising my own version of this check.
# Solution
Compare terms size of id field with the terms size of the `_root_` field to
ensure no documents share a root. We do this for each segment separately as was
done for the previous check.
# Tests
1. testNestedDocsDetection_nonNestedJustAdd — Plain docs, no updates.
Verifies no false positive.
2. testNestedDocsDetection_nestedJustAdd — Actual nested docs. Verifies
detection works.
3. testNestedDocsDetection_nonNestedWithWithinCommitUpdates — Non-nested
docs where some are re-added (same id) before commit. The deleted+re-added docs
share _root_ values within the segment, which previously caused false positives.
4. testNestedDocsDetection_nestedWithWithinCommitUpdates — Same
within-commit update pattern but with real nested docs present.
5. testNestedDocsDetection_nonNestedWithWithinCommitDeletesAndUpdates —
Non-nested docs with deletes AND updates before commit. Deleted entries in the
segment can also trigger false positives.
6. testNestedDocsDetection_nestedWithWithinCommitDeletesAndUpdates — Same
delete+update pattern with real nested docs.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
