From: Tom Lane <tgl@sss.pgh.pa.us>
To: Erwin Brandstetter <brsaweda@gmail.com>
cc: "David G. Johnston" <david.g.johnston@gmail.com>,
        Jeff Janes <jeff.janes@gmail.com>,
        PostgreSQL Bug List <pgsql-bugs@lists.postgresql.org>
Subject: Re: Variable substitution in jsonb functions fails for jsonpath
 operator like_regex
In-reply-to: 
 <CAGHENJ7YdNYzZruGM2+fRDJZOn_7p6u3f686mzodbti_AQ8xbg@mail.gmail.com>
References: 
 <CAGHENJ4A8awD2uXHkf50eV+9wyiDm3eP+G10J09+0VRmXBQAdw@mail.gmail.com>
 <3157573.1697663770@sss.pgh.pa.us>
 <CAMkU=1x1bkHBczE4XZ1w=pyE75EGbkxGMUOYLwHMkp4t74vUwg@mail.gmail.com>
 <3241895.1697679989@sss.pgh.pa.us>
 <CAKFQuwaBB+PPZkn0GtTzt5z6Vtypt2CQ-5+nK7XzW6zm9UqaDQ@mail.gmail.com>
 <3378578.1697742079@sss.pgh.pa.us>
 <CAGHENJ7YdNYzZruGM2+fRDJZOn_7p6u3f686mzodbti_AQ8xbg@mail.gmail.com>
Comments: In-reply-to Erwin Brandstetter <brsaweda@gmail.com>
	message dated "Tue, 24 Oct 2023 00:59:26 +0200"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <460387.1698108792.1@sss.pgh.pa.us>
Date: Mon, 23 Oct 2023 20:53:12 -0400
Message-ID: <460388.1698108792@sss.pgh.pa.us>
Archived-At: 
 <https://www.postgresql.org/message-id/460388.1698108792%40sss.pgh.pa.us>
Precedence: bulk

Erwin Brandstetter <brsaweda@gmail.com> writes:
> On Thu, 19 Oct 2023 at 21:01, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> Hmm ... looks like we *are* following the standard:
>>   ...
>> The text mentions that "the second operand is permitted to be an SQL/JSON
>> sequence and to support existential semantics", whereas they evidently
>> don't want that for a regex pattern.

> So input from "vars" cannot be substituted into the jsonpath expression
> after "like_regex" (as opposed to all other jsonpath operators). Seems
> pretty random from a user's perspective.

I agree it looks pretty random if you haven't drilled down into the
spec's fine print.  Personally I wouldn't be opposed to extending
the spec here (not that I'm volunteering to write the patch).

Nosing around in jsonpath_gram.y, I see datetime_template as the
only other place where there's a random-seeming choice to allow
STRING_P but not VARIABLE_P.  Should we tackle that too?

			regards, tom lane