Re: uninterruptable regexp_replace in 9.2 and 9.3

Sandro Santilli wrote, On 2014-02-28 12:28:
> Starting with commit 173e29aa5 the regexp_replace function
> became uninterruptable during operations, and takes a lot
> of time and RAM to process some patterns.
>
> Full story in this thread:
> http://www.postgresql.org/message-id/646.1393031856@sss.pgh.pa.us
>
> I'm hoping a mail to pgsql-bugs would assign this issue a ticket
> number for me to follow. Let me know if there's a better way.
> Thank you !

This may be relevant:
https://gist.github.com/johnbartholomew/8379265

I've added these lines:

printf 'Testing psql:\n'
time psql -c "SELECT regexp_matches('$pattern','$input');"

The results in my machine are (with PostgreSQL 9.1.9):

Pattern:
"a?a?a?a?a?a?a?a?a?a?a?a?a?a?a?a?a?a?a?a?a?a?a?a?a?a?a?aaaaaaaaaaaaaaaaaaaaaaaaaaa"
Input: "aaaaaaaaaaaaaaaaaaaaaaaaaaa"
Testing grep:
aaaaaaaaaaaaaaaaaaaaaaaaaaa

real 0m0.003s
user 0m0.000s
sys 0m0.000s
Testing perl:
match

real 0m6.103s
user 0m6.100s
sys 0m0.000s
Testing python:
<_sre.SRE_Match object at 0xb70e2250>

real 0m10.207s
user 0m10.141s
sys 0m0.060s
Testing psql:
regexp_matches
-------------------------------
{aaaaaaaaaaaaaaaaaaaaaaaaaaa}
(1 row)

real 0m0.039s
user 0m0.024s
sys 0m0.004s

Interestingly, as noted in the comments, PHP reports an error that the
backtracking limit was reached.

I'm adding this note in case it helps anyone get a bigger picture as to
what other implementations do about this problem.