Date: Thu, 12 Jun 2025 14:10:00 +0200
From: hubert depesz lubaczewski <depesz@depesz.com>
Sender: depesz@depesz.com
To: anudeepvattikonda0404@gmail.com, pgsql-bugs@lists.postgresql.org
Subject: Re: BUG #18956: Observing an issue in regexp_count()
Message-ID: <aErDmEWoyiNrhY1M@depesz.com>
Reply-To: depesz@depesz.com
References: <18956-9169151921910054@postgresql.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: <18956-9169151921910054@postgresql.org>
Archived-At: 
 <https://www.postgresql.org/message-id/aErDmEWoyiNrhY1M%40depesz.com>
Precedence: bulk

On Thu, Jun 12, 2025 at 08:03:25AM +0000, PG Bug reporting form wrote:
> The following bug has been logged on the website:
> 
> Bug reference:      18956
> Logged by:          Anudeep Vattikonda
> Email address:      anudeepvattikonda0404@gmail.com
> PostgreSQL version: 17.5
> Operating system:   Mac
> Description:        
> 
> Hi
> I am trying to run the below query
> select REGEXP_COUNT('cat at the flat', '\Bat\b') ;
> I was expecting it to return 2 but I see Postgres is returning 0. I see that
> there are two matches, cat and flat. All it should do is to look for the
> word at whose left side shoudn't be a word boundary while the right side
> should be a word boundary

What makes you think that \B/\b has anything to do with word boundary?

Docs
(https://www.postgresql.org/docs/17/functions-matching.html#FUNCTIONS-POSIX-REGEXP)
show:

\b - backspace, as in C
\B - synonym for backslash (\) to help reduce the need for backslash doubling

As far as I can tell pg regexps have nothing related to word boundaries.

You could get 2 by changing the regexp to something that actually works:

$ select REGEXP_COUNT('cat at the flat', '[a-z]at(?![a-z])');
 regexp_count
──────────────
            2
(1 row)

Best regards,

depesz