Re: ICU for global collation

Peter Eisentraut <peter.eisentraut@enterprisedb.com>

From: Peter Eisentraut <peter.eisentraut@enterprisedb.com>
To: Marina Polyakova <m.polyakova@postgrespro.ru>, pgsql-hackers@postgresql.org
Cc: Michael Paquier <michael@paquier.xyz>, Julien Rouhaud <rjuju123@gmail.com>, Daniel Verite <daniel@manitou-mail.org>, pryzby@telsasoft.com
Date: 2022-08-22T14:10:59Z
Lists: pgsql-hackers
On 15.08.22 14:06, Marina Polyakova wrote:
> 1.1) It looks like there's a bug in the function get_db_infos 
> (src/bin/pg_upgrade/info.c), where the version of the old cluster is 
> always checked:
> 
> if (GET_MAJOR_VERSION(old_cluster.major_version) < 1500)
>      snprintf(query + strlen(query), sizeof(query) - strlen(query),
>               "'c' AS datlocprovider, NULL AS daticulocale, ");
> else
>      snprintf(query + strlen(query), sizeof(query) - strlen(query),
>               "datlocprovider, daticulocale, ");
> 
> With the simple patch
> 
> diff --git a/src/bin/pg_upgrade/info.c b/src/bin/pg_upgrade/info.c
> index 
> df374ce4b362b4c6c87fc1fd0e476e5d6d353d9e..53ea348e211d3ac38334292bc16cb814bc13bb87 
> 100644
> --- a/src/bin/pg_upgrade/info.c
> +++ b/src/bin/pg_upgrade/info.c
> @@ -319,7 +319,7 @@ get_db_infos(ClusterInfo *cluster)
> 
>       snprintf(query, sizeof(query),
>                "SELECT d.oid, d.datname, d.encoding, d.datcollate, 
> d.datctype, ");
> -    if (GET_MAJOR_VERSION(old_cluster.major_version) < 1500)
> +    if (GET_MAJOR_VERSION(cluster->major_version) < 1500)
>           snprintf(query + strlen(query), sizeof(query) - strlen(query),
>                    "'c' AS datlocprovider, NULL AS daticulocale, ");
>       else

fixed

> 1.2) It looks like the mentioned asserion in dbcommands.c conflicts with 
> the following lines earlier:
> 
> if (dbiculocale == NULL)
>      dbiculocale = src_iculocale;

fixed

> I'm wondering if this is not a fully-supported feature (because createdb 
> creates an SQL command with LC_COLLATE and LC_CTYPE options instead of 
> LOCALE option) or is it a bug in CREATE DATABASE?.. From 
> src/backend/commands/dbcommands.c:
> 
> if (dblocprovider == COLLPROVIDER_ICU && !dbiculocale)
> {
>      if (dlocale && dlocale->arg)
>          dbiculocale = defGetString(dlocale);
> }

I think this piece of code was left over from some earlier attempts to 
specify the libc locale and the icu locale with one option, which never 
really worked well.  The CREATE DATABASE man page does not mention that 
LOCALE provides the default for ICU_LOCALE.  Hence, I think we should 
delete this.



Commits

Same data as JSON: GET /api/v1/messages/:b64id/commits the thread's linked commits as JSON, with link sources. API reference →
  1. Add CHECK_FOR_INTERRUPTS while restoring changes during decoding.

  2. Improve ICU option handling in CREATE DATABASE

  3. Don't allow creation of database with ICU locale with unsupported encoding

  4. Make locale option behavior more consistent

  5. pg_dump: Dump colliculocale

  6. Remove further unwanted linker flags from perl_embed_ldflags

  7. Doc: document possible need to raise kernel's somaxconn limit.

  8. Reduce warnings with -Wshadow=compatible-local builds

  9. Remove redundant spaces in _outA_Expr() output

  10. Fix outdated --help message for postgres -f

  11. pg_upgrade: Fix version comparison for global ICU support

  12. Silence -Wmaybe-uninitialized compiler warning in dbcommands.c.

  13. Add option to use ICU as global locale provider

  14. DefineCollation() code cleanup

  15. Change collate and ctype fields to type text

  16. Call pg_newlocale_from_collation() also with default collation

  17. Use libc version as a collation version on glibc systems.

  18. Silence -Wmaybe-uninitialized compiler warnings in dbcommands.c.

  19. Make LC_COLLATE and LC_CTYPE database-level settings. Collation and