Repository navigation
Conversation
daguimu
added a commit
to daguimu/elasticsearch
that referenced
this pull request
May 8, 2026
Collaborator
|
Pinging @elastic/es-search-relevance (Team:Search Relevance) |
benwtrent
reviewed
May 12, 2026
Comment on lines
+239
to
+244
| Automaton automaton; | ||
| try { | ||
| automaton = WildcardQuery.toAutomaton(new Term(name(), matchPattern), Operations.DEFAULT_DETERMINIZE_WORK_LIMIT); | ||
| } catch (TooComplexToDeterminizeException e) { | ||
| throw new IllegalArgumentException("Pattern was too complex to determinize", e); | ||
| } |
Contributor
There was a problem hiding this comment.
let's do the same thing that was done in the KeywordFieldMapper. So something like:
Term term = new Term(name(), value);
if (context.getCircuitBreaker() != null) {
Automaton dfa = AutomatonQueries.toWildcardAutomaton(term, context.getCircuitBreaker());
return new AutomatonQuery(term, dfa, false, MultiTermQuery.DOC_VALUES_REWRITE);
}
return new WildcardQuery(term, Operations.DEFAULT_DETERMINIZE_WORK_LIMIT, MultiTermQuery.DOC_VALUES_REWRITE);
Contributor
Author
There was a problem hiding this comment.
Done — switched to AutomatonQueries.toWildcardAutomaton(term, circuitBreaker) so the NFA→DFA determinization heap is tracked alongside the other wildcard queries.
One deviation worth flagging: constant_keyword has no Lucene index (ConstantFieldType uses IndexType.NONE; parseCreateField doesn't write the value), so returning an AutomatonQuery / WildcardQuery against the field would scan an empty index and always return zero docs. I kept the in-memory CharacterRunAutomaton.run(value) → ALL_DOCS_INSTANCE / NO_DOCS_INSTANCE shape, matching the existing ConstantFieldType#automatonQuery pattern and the rest of the matches()-based queries on this type. Happy to switch approach if you'd prefer something else.
ConstantKeywordFieldType#matches uses Regex.simpleMatch, which only treats * as a wildcard. The wildcard query DSL also defines ? as matching exactly one character, so patterns like f?o failed against constant values that should match. Override wildcardQuery on the constant_keyword field type to use Lucene's WildcardQuery automaton, matching the semantics of other text field types. Also drop the final modifier from ConstantFieldType#wildcardQuery so constant-value field types can supply their own matcher when simpleMatch semantics are wrong for them. Closes elastic#141785
Compile via AutomatonQueries.toWildcardAutomaton when the SearchExecutionContext exposes a circuit breaker, so the NFA->DFA determinization heap is tracked alongside other wildcard queries. Falls back to work-limit-only WildcardQuery.toAutomaton when no breaker is available (e.g. coordinator rewrite paths).
daguimu
force-pushed
the
fix/constant-keyword-question-wildcard-141785
branch
from
May 15, 2026 05:54
02c9d65 to
43ecc0b
Compare
Contributor
|
buildkite test this please |
benwtrent
approved these changes
May 15, 2026
Contributor
|
Thank you for the contribution @daguimu! |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Problem
A
wildcardquery against aconstant_keywordfield treats?as a literal character instead of "match exactly one character". Per the wildcard query docs,?and*are both wildcard metacharacters.f*o*and other patterns containing only*work as expected — only?is broken.Root Cause
ConstantFieldType#wildcardQuery(final) delegates to the abstractmatches(String pattern, boolean caseInsensitive, QueryRewriteContext).ConstantKeywordFieldType#matchesis implemented withRegex.simpleMatch, which only handles*(per its own Javadoc — "xxx*", "xxx", "xxx", "xxxyyy"). When the pattern contains?,simpleMatchfalls through to literal equality and returnsfalse.The same
matches()is reused bytermQuery/termsQuery/prefixQuerywhere?should remain literal, so we cannot simply changematches()to honour?.Fix
ConstantFieldType#wildcardQuery(String, boolean, QueryRewriteContext): dropfinalso subclasses can plug in their own wildcard matcher whensimpleMatchsemantics are wrong for them. The default still callsmatches(), so behaviour for_index,_index_mode,_tier, etc. is unchanged.ConstantKeywordFieldType: overridewildcardQueryto compile the pattern withWildcardQuery.toAutomatonand run the resultingCharacterRunAutomatonagainst the constant value. This mirrors whatIndexFieldTypealready does for itswildcardLikeQuerypath.Tests Added
?matches exactly one character (f?o,???,f?o*)ConstantKeywordFieldTypeTests#testWildcardQueryWithQuestionMark?matches exactly one —f?does not matchfoonullconstant value short-circuits toNO_DOCSThe pre-existing
testWildcardQuerycontinues to cover*-only patterns and case-insensitive matching.Impact
wildcardqueries againstconstant_keywordfields now honour?as a single-character wildcard, matching the documented semantics.term,terms,prefix, orregexpqueries onconstant_keyword.Closes #141785