From: Tom Lane <tgl@sss.pgh.pa.us>
To: Simon Ellmann <simon.ellmann@tum.de>
cc: "pgsql-bugs@lists.postgresql.org" <pgsql-bugs@lists.postgresql.org>
Subject: Re: regexp_replace not respecting greediness
In-reply-to: <1B268DC1-F058-4243-B6F8-71451FB68F08@tum.de>
References: <1B268DC1-F058-4243-B6F8-71451FB68F08@tum.de>
Comments: In-reply-to Simon Ellmann <simon.ellmann@tum.de>
	message dated "Fri, 19 Sep 2025 12:36:20 -0000"
MIME-Version: 1.0
Content-Type: text/plain; charset="UTF-8"
Content-ID: <2134892.1758297712.1@sss.pgh.pa.us>
Content-Transfer-Encoding: quoted-printable
Date: Fri, 19 Sep 2025 12:01:53 -0400
Message-ID: <2134893.1758297713@sss.pgh.pa.us>
Archived-At: 
 <https://www.postgresql.org/message-id/2134893.1758297713%40sss.pgh.pa.us>
Precedence: bulk

Simon Ellmann <simon.ellmann@tum.de> writes:
> With the following regular expression, the second .* seems to match non-=
greedily although (if I am correct) it should match greedily:
> postgres=3D# SELECT REGEXP_REPLACE('jane.smith@example.com', '.*?@.*', '=
ab');

This is correct according to the rules given at

https://www.postgresql.org/docs/current/functions-matching.html#POSIX-MATC=
HING-RULES

specifically that "A branch =E2=80=94 that is, an RE that has no top-level=
 |
operator =E2=80=94 has the same greediness as the first quantified atom in=
 it
that has a greediness attribute."  Because of that, the RE as a whole
is non-greedy and will match the shortest not longest amount of text
overall.  The discussion in that manual section shows what to do
when you don't like the results.

> Other database systems (e.g., DuckDB, Umbra) match the whole input:

If your complaint is "but it's not like Perl!", I suggest using
a plperl function to do your regexp work.

			regards, tom lane