Re: ICU locale validation / canonicalization
Jeff Davis <pgsql@j-davis.com>
From: Jeff Davis <pgsql@j-davis.com>
To: Noah Misch <noah@leadboat.com>
Cc: Peter Eisentraut <peter.eisentraut@enterprisedb.com>, Robert Haas <robertmhaas@gmail.com>, pgsql-hackers@postgresql.org
Date: 2023-07-07T16:13:35Z
Lists: pgsql-hackers
On Sat, 2023-07-01 at 10:31 -0700, Noah Misch wrote: > As of commit b8c3f6d, InstallCheck-C got daticulocale=en-US-u-va- > posix. Check > got daticulocale=NULL. With the same test setup, that locale takes about 8.6 seconds (opening it 10M times), about 2.5X slower than "en-US" and about 7X slower than "und". I think that explains it. The locale "en-US-u-va-posix" normally happens when passing a locale beginning with "C" to ICU. After 2535c74b1a we don't get ICU locales from the environment anywhere, so that should be rare (and probably indicates a user mistake). I don't think this is a practical problem any more. Regards, Jeff Davis
Commits
-
Add missing source file to nls.mk
- b8c3f6df85e7 16.0 cited
-
Fix MSVC warning introduced in ea1db8ae70.
- 36320cbc164a 16.0 landed
-
Canonicalize ICU locale names to language tags.
- ea1db8ae70e5 16.0 landed
-
Validate ICU locales.
- 1671f990dd66 16.0 landed
-
initdb: emit message when using default ICU locale.
- c1f1c1f87fd6 16.0 landed
-
initdb: replace check_icu_locale() with default_icu_locale().
- f8ca22295e99 16.0 landed
-
Fix error inconsistency in older ICU versions.
- 8b3eb0c584be 16.0 landed
-
Avoid potential UCollator leak for older ICU versions.
- a03b3b6b4a72 16.0 landed
-
pg_locale.c: change ereport() to elog().
- 9a242899158c 16.0 landed
-
Handle the "und" locale in ICU versions 54 and older.
- 3b50275b1295 16.0 landed
-
Wrap ICU ucol_open().
- a326aac8f171 16.0 landed
-
Support language tags in older ICU versions (53 and earlier).
- 869650fa86ad 16.0 landed
-
Improve support for UNICODE collation on older ICU
- d72900bdedc6 16.0 cited
-
Use ICU by default at initdb time.
- 27b62377b47f 16.0 cited