SOLR-8088: Grouping on a multiValued non-numeric field is rejected up front with a 400 error

[nick-boss-tech]

🤖 AI text below 🤖 (posted on behalf of Nick Shanin)

https://issues.apache.org/jira/browse/SOLR-8088

What happens today

Distributed grouping on a multiValued non-numeric field is accepted, then fails later inside Lucene.

A distributed grouped query can name a multiValued field, such as cat in the test schema. The search-phase command builder checks only that its required fields are set (SearchGroupsFieldCommand.java lines 68-74 at the merge-base). Nothing checks the field type there. The grouping code then reads the field as single valued docValues, and the request fails inside Lucene with an IllegalStateException about docValues.

What this change does

Grouping on a multiValued non-numeric field is rejected up front with a 400 error.

• Both command builders run the same check. The search-phase builder does (SearchGroupsFieldCommand.java lines 69-91 at the head), and so does the second-phase builder (TopGroupsFieldCommand.java line 109).
• For a field named cat, the 400 message reads: Distributed grouping requires a single valued field, but 'cat' is multiValued; group on a single valued (e.g. copyField) field instead.
• A distributed test sends a grouped query naming the multiValued field cat and expects the 400 error (TestDistributedGrouping.java lines 1731-1745 at the head).
• Single valued fields are unchanged. The check skips numeric fields (line 84).

Proof

A new test fails on the base code and passes with this change.

• SearchGroupsFieldCommandTest.testMultiValuedFieldIsRejected (test lines 33-49) builds the command for a multiValued field and expects a 400 error. On the base code the builder returns normally, so the expected error never comes.
• Verified 2026-10-14 at head c36cfc0: SearchGroupsFieldCommandTest 2 of 2. On the base code, with this test in place, the class runs 2 tests and one fails: this test.

Limits

Numeric multiValued fields are not checked.

• Grouping on a multiValued field that works end to end is a larger change, and it is not made here. I can open a follow-up ticket and PR for it on request.

Changelog: changelog/unreleased/SOLR-8088-grouping-multivalued.yml

AI assistance

AI agents assisted with research, implementation, review, and drafting. Nick Shanin directed the work and takes responsibility for this contribution.