Recently, I discovered a phenomenon where there is a large amount of CacheEvaluationContext instances in the jmap histo from my Spring Boot 2.7.15 application. When I examined the creation points in the source code, I found that the creation of this instance is triggered by three methods in CacheOperationContext: canPutToCache, generateKey, and isConditionPassing. The triggering conditions for these methods are StringUtils.hasText(unless), StringUtils.hasText(this.metadata.operation.getKey()), and StringUtils.hasText(this.metadata.operation.getCondition()) respectively. How should I optimize to reduce the number of CacheEvaluationContext instances?
Comment From: snicoll
@liudaac you have described a situation but didn't explain how it was a problem. CacheEvaluationContext is a support for SpEL expression handling and shouldn't be costly. We don't keep references to those so they will be garbage collected when the JVM decides it should. Without any description of a concrete problem this issue is not actionable and looks more like a question.
If you meat to ask a question, please use StackOverlfow, as mentioned in the guidelines for contributing, we prefer to use the issue tracker only for bugs and enhancements.
Comment From: snicoll
In the meantime, we've found something trying to fix an unrelated issue that goes in the direction of the report, see #31617 for more details.
Comment From: liudaac
Thank you very much for your attention. We hope the bug can be fixed as soon as possible.
Comment From: snicoll
@liudaac if you felt this was a bug, feel free to add a comment. Me closing the issue as not actionable was because I felt this was a question. But that doesn't mean you can't provide more context to change our mind. In this case, we don't feel this is a bug, but it's true that it's putting a bit of unnecessary pressure on GC and that's what the linked issue intends to fix. If you want to make sure whatever issue you've discovered is fixed, please watch the related issue and don't hesitate to comment there if the fix doesn't meet your expectations.