Josh Berkus <josh@agliodbs.com> writes:
> I'm doing some analysis of PostgreSQL site traffic, and am being frequently
> hung up by the compile-time-fixed size of our regex cache (32 regexes, per
> MAX_CACHED_RES). Is there a reason why it would be hard to use work_mem
> or some other dynamically changeable limit for regex caching?
Hmmm ... Spencer's regex library makes a point of hiding its internal
representation of a compiled regex from the calling code. So measuring
the size of the regex cache in bytes would involve doing a lot of
violence to that API. We could certainly allow the size of the cache
measured in number-of-regexes to be controlled, though.
Having said that, I'm not sure it'd help your problem. If your query is
using more than 32 regexes concurrently, it likely is using $BIGNUM
regexes concurrently. How do we fix that?
regards, tom lane