From: Tom Lane <tgl@sss.pgh.pa.us>
To: Teodor Sigaev <teodor@sigaev.ru>
cc: Robert Haas <robertmhaas@gmail.com>,
        =?UTF-8?Q?Dagfinn_Ilmari_Manns=c3=a5ker?= <ilmari@ilmari.org>,
        Aleksander Alekseev <a.alekseev@postgrespro.ru>,
        David Steele <david@pgmasters.net>,
        Nikita Glukhov <n.gluhov@postgrespro.ru>,
        Darafei Praliaskouski <me@komzpa.net>,
        Anastasia Lubennikova <a.lubennikova@postgrespro.ru>,
        PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>
Subject: Re: Cast jsonb to numeric, int, float, bool
In-reply-to: <24822.1525789657@sss.pgh.pa.us>
References: 
 <151985418868.6915.17223208131889314092.pgcf@coridan.postgresql.org>
 <76b000ae-94d1-cd8f-a2e0-e536096469b0@postgrespro.ru>
 <d75ed4fe-574b-41c7-080f-ed662fa5d9ba@pgmasters.net>
 <20180327145337.GA20311@e733.localdomain>
 <ca0bdf99-98a5-3a44-3f49-13941e0634ca@sigaev.ru>
 <912e1a07-90a9-ad97-bc70-d71ec0f6f073@sigaev.ru>
 <be43822f-7fd2-c5c2-8435-6a896d1ed46c@sigaev.ru>
 <CA+Tgmob08StTV9yu04D0idRFNMh+UoyKax5Otvrix7rEZC8rMw@mail.gmail.com>
 <0250b45d-4a38-5453-cb5a-8fa9883da36b@sigaev.ru>
 <20180330142119.GA1411@e733.localdomain>
 <20180330142423.GB1411@e733.localdomain> <d8j1sg1mvwp.fsf@dalvik.ping.uio.no>
 <CA+TgmoaYgUkbrpzYgBPhH31+9BioCRs3RF2Bcz8E-+zttcCB4A@mail.gmail.com>
 <bab83c35-2a25-927b-d28d-f836402c48ba@sigaev.ru>
 <24822.1525789657@sss.pgh.pa.us>
Comments: In-reply-to Tom Lane <tgl@sss.pgh.pa.us>
	message dated "Tue, 08 May 2018 10:27:37 -0400"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <2085.1525792633.1@sss.pgh.pa.us>
Date: Tue, 08 May 2018 11:17:13 -0400
Message-ID: <2087.1525792633@sss.pgh.pa.us>
Precedence: bulk

I wrote:
> 1) Does this really pass muster from the translatability standpoint?
> I doubt it.

After further thought about that, it seems that what we typically don't
try to translate is SQL-standard type names, that is, error messages
along the line of "blah blah blah type %s" are considered fine.  So
the problem here is that you factorized the error reporting poorly.
I think you want the callers to look like

 	if (!JsonbExtractScalar(&in->root, &v) || v.type != jbvNumeric)
		cannotCastJsonbValue(v.type, "double precision");

where the subroutine contains the whole ereport() call, and its lookup
table entries are e.g.

	gettext_noop("cannot cast jsonb string to type %s")

			regards, tom lane