Optimize StringMatcher for match-all patterns #77738
Conversation
Detect a StringMatcher of "*" and optimize it as a predicate of s -> true This is about 80% faster than using an automaton. This speed improvement is very minor on a single execution but can add up to milliseconds is used in a tight loop - e.g. testing against every indices in a large cluster
tvernum
added
>enhancement
:Security/Security
|
Pinging @elastic/es-security (Team:Security) |
ywangd
left a comment
ywangd
left a comment
There was a problem hiding this comment.
LGTM
It would be great if you could address two non-critical comments.
| if (nonExactMatch.contains("*")) { | ||
| return new StringMatcher(description, s -> true); | ||
| } |
There was a problem hiding this comment.
Since we are trying to squeeze every bit for performance, I wonder whether it could help a tiny bit if we perform this check in the include(String pattern) method. It already checks for wildcard and should be easy to tweak it to flag for seeing a match-all pattern.
There was a problem hiding this comment.
That was my original implementation, but it's really not necessary.
We're not looking to squeeze that level of performance in the Builder, just in the StringMatcher itself.
Testing every pattern that is included to see whether it's equal to * and then setting a flag isn't really going to make a difference, and it means carrying more state in the builder.
...gin/core/src/test/java/org/elasticsearch/xpack/core/security/support/StringMatcherTests.java
Show resolved
Hide resolved
|
@elasticmachine update branch |
|
@elasticmachine run elasticsearch-ci/part-2 please Test failed due to: #78035 |
Detect a StringMatcher of "*" and optimize it as a predicate of
s -> true
This is about 80% faster than using an automaton.
This speed improvement is very minor on a single execution but can
add up to milliseconds is used in a tight loop - e.g. testing
against every indices in a large cluster
💚 Backport successful
|
* Optimize StringMatcher for match-all patterns (#77738) Detect a StringMatcher of "*" and optimize it as a predicate of s -> true This is about 80% faster than using an automaton. This speed improvement is very minor on a single execution but can add up to milliseconds is used in a tight loop - e.g. testing against every indices in a large cluster * Fix List references for backport Co-authored-by: Elastic Machine <[email protected]>
5 participants
Detect a StringMatcher of "*" and optimize it as a predicate of
This is about 80% faster than using an automaton.
The speed improvement is very minor on a single execution but can
add up to milliseconds if used in a tight loop - e.g. testing
against every index in a large cluster