From: Andrew Gierth <andrew@tao11.riddles.org.uk>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: Peter Eisentraut <peter.eisentraut@enterprisedb.com>,  Sandro Santilli
 <strk@kbt.io>,  Regina Obe <lr@pcorp.us>,
  pgsql-hackers@lists.postgresql.org,  Jeff Davis <pgsql@j-davis.com>
Subject: Re: Order changes in PG16 since ICU introduction
In-Reply-To: <3369718.1682100417@sss.pgh.pa.us> (Tom Lane's message of "Fri,
	21 Apr 2023 14:06:57 -0400")
Message-ID: <874jp9f5jo.fsf@news-spur.riddles.org.uk>
References: <000b01d97465$c34bbd60$49e33820$@pcorp.us>
	<3353947.1682092131@sss.pgh.pa.us> <ZELDOktA75yYG4Z2@hst.kbt.io>
	<533d93c5-f604-8e5f-4a48-975c08c53d59@enterprisedb.com>
	<878relf7cb.fsf@news-spur.riddles.org.uk>
	<3369718.1682100417@sss.pgh.pa.us>
User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/27.1 (berkeley-unix)
Date: Fri, 21 Apr 2023 20:14:20 +0100
MIME-Version: 1.0
Content-Type: text/plain
Archived-At: <https://www.postgresql.org/message-id/874jp9f5jo.fsf%40news-spur.riddles.org.uk>
Precedence: bulk

>>>>> "Tom" == Tom Lane <tgl@sss.pgh.pa.us> writes:

 >> Also, somewhere along the line someone broke initdb --no-locale,
 >> which should result in C locale being the default everywhere, but
 >> when I just tested it it picked 'en' for an ICU locale, which is not
 >> the right thing.

 Tom> Confirmed:

 Tom> $ LANG=en_US.utf8 initdb --no-locale
 Tom> The files belonging to this database system will be owned by user "postgres".
 Tom> This user must also own the server process.

 Tom> Using default ICU locale "en_US".
 Tom> Using language tag "en-US" for ICU locale "en_US".
 Tom> The database cluster will be initialized with this locale configuration:
 Tom>   provider:    icu
 Tom>   ICU locale:  en-US
 Tom>   LC_COLLATE:  C
 Tom>   LC_CTYPE:    C
 Tom>   ...

 Tom> That needs to be fixed: --no-locale should prevent any
 Tom> consideration of initdb's LANG/LC_foo environment.

Would it also not make sense to also take into account any --locale and
--lc-* options before choosing an ICU default locale? Right now if you
do, say, initdb --locale=fr_FR you get an ICU locale based on the
environment but lc_* settings based on the option, which seems maximally
confusing.

Also, what happens now to lc_collate_is_c() when the provider is ICU? Am
I missing something, or is it never true now, even if you specified C /
POSIX / en-US-u-va-posix as the ICU locale? This seems like it could be
an important pessimization.

Also also, we now have the problem that it is much harder to create a
'C' collation database within an existing cluster (e.g. for testing)
without knowing whether the default provider is ICU. In the past one
would have done:

CREATE DATABASE test TEMPLATE=template0 ENCODING = 'UTF8' LOCALE = 'C';

but now that creates a database that uses the same ICU locale as
template0 by default. If instead one tries:

CREATE DATABASE test TEMPLATE=template0 ENCODING = 'UTF8' LOCALE = 'C' ICU_LOCALE='C';

then one gets an error if the default locale provider is _not_ ICU. The
only option now seems to be:

CREATE DATABASE test TEMPLATE=template0 ENCODING = 'UTF8' LOCALE = 'C' LOCALE_PROVIDER = 'libc';

which of course doesn't work in older pg versions.

-- 
Andrew.