From: Tom Lane <tgl@sss.pgh.pa.us>
To: Mark Steben <mark.steben@drivedominion.com>
cc: pgsql-admin <pgsql-admin@postgresql.org>
Subject: Re: using regular expression queries with regular expression indexes
In-reply-to: 
 <CADyzmyykegZ3ZSVm9hYYKirJYGUYYAx_cK1VQYPCz1q+EpQ89g@mail.gmail.com>
References: 
 <CADyzmyykegZ3ZSVm9hYYKirJYGUYYAx_cK1VQYPCz1q+EpQ89g@mail.gmail.com>
Comments: In-reply-to Mark Steben <mark.steben@drivedominion.com>
	message dated "Thu, 25 Jun 2020 12:45:55 -0400"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <2277794.1593104269.1@sss.pgh.pa.us>
Content-Transfer-Encoding: quoted-printable
Date: Thu, 25 Jun 2020 12:57:49 -0400
Message-ID: <2277795.1593104269@sss.pgh.pa.us>
Precedence: bulk

Mark Steben <mark.steben@drivedominion.com> writes:
> We are currently running postgres version 12.2. I have a query that uses
> regexp code to confirm if a text field is a date or not.  It is resource
> intensive and long running.  I have made a regexp index (actually severa=
l
> flavors of regexp indexes) to aid in this queries performance. However I
> cannot seem to 'convince' the planner to use the indexes I create.

I don't understand your complaint.  Your example shows that it *is*
using the index:

>    ->  Bitmap Index Scan on custattr_value  (cost=3D0.00..9787.64 rows=3D=
479041 width=3D0)

> Finally if I remove the 'ca.account_id is not null' expression
> [ it doesn't use the index ]

Well, of course not, because then the query doesn't match the index's
WHERE condition (i.e., there could be rows the query needs to return that
are not included in the index).  But I imagine that restriction isn't very
selective, so you could probably omit it from the index definition as
well; then the index would work for queries with or without that
restriction.

			regards, tom lane