[SPARK-55430][SQL] Cache ICU StringSearch for collation string predicates with constant patterns #54241
+388
−6
Conversation
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
…tsWith/EndsWith Add StringSearch object caching for UNICODE and UNICODE_CI collation string predicates when the pattern is a compile-time constant (foldable). Instead of creating a new StringSearch on every row evaluation, a single StringSearch is created once and reused by calling setTarget() for each new input string. This applies to both the interpreted path (via @transient private lazy val) and the codegen path (via addMutableState). Changes: - CollationFactory: add getStringSearchForPattern() factory method - CollationSupport: add cached execICU() overloads for Contains, StartsWith, EndsWith that accept a pre-built StringSearch - stringExpressions: wire caching into Contains, StartsWith, EndsWith expression evaluation and code generation - CollationBenchmark: add fixed-pattern benchmarks measuring caching benefit (3-3.4X improvement for UNICODE collation)
nateab
force-pushed
the
cache-icu-stringsearch-collation
branch
from
d73acc5 to
f9d4f58
Compare
nateab
changed the title
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.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.
What changes were proposed in this pull request?
Add StringSearch object caching for
Contains,StartsWith, andEndsWithexpressions when used with ICU-based collations (UNICODE, UNICODE_CI) and a compile-time constant (foldable) pattern.Currently, every row evaluation creates a new
com.ibm.icu.text.StringSearchobject. When the pattern is constant, this repeated construction is unnecessary. With this change, a singleStringSearchis created once and reused viasetTarget()for each new input string — both in interpreted (@transient private lazy val) and codegen (ctx.addMutableState) paths.Changes:
CollationFactory: addgetStringSearchForPattern()factory methodCollationSupport: add cachedexecICU()overloads for Contains, StartsWith, EndsWithstringExpressions.scala: wire caching into expression eval and codegen when pattern is foldable and collation is ICU-basedCollationBenchmark: add fixed-pattern benchmarksWhy are the changes needed?
ICU StringSearch construction is expensive. For queries scanning large tables with constant string predicates under ICU collations, this overhead is incurred on every row. Caching yields 3-3.4X improvement.
Does this PR introduce any user-facing change?
No. Performance optimization only.
How was this patch tested?
All 192 existing collation tests pass across 7 test suites. New fixed-pattern benchmarks added:
Was this patch authored or co-authored using generative AI tooling?
Yes, Claude Code was used as an AI coding assistant.