From: Peter Eisentraut <peter@eisentraut.org>
To: Laurenz Albe <laurenz.albe@cybertec.at>
To: patrick.van.dijk@gmail.com
To: pgsql-bugs@lists.postgresql.org
Subject: Re: BUG #18523: String compare not consistent
Date: Wed, 26 Jun 2024 20:30:56 +0200
Message-ID: <30c9e733-9573-4500-b191-5a8a67b3795f@eisentraut.org> (raw)
In-Reply-To: <3a173a9aba9650ce7ddf22af9c413a3d64debbf4.camel@cybertec.at>
References: <18523-d7b3cacb64d91cda@postgresql.org>
<3a173a9aba9650ce7ddf22af9c413a3d64debbf4.camel@cybertec.at>
On 26.06.24 17:59, Laurenz Albe wrote:
> Interestingly, glibc and ICU disagree about that:
>
> SELECT '|1' < '01' COLLATE "de_DE.utf8" AS glibc,
> '|1' < '01' COLLATE "de-AT-x-icu" AS icu;
>
> glibc │ icu
> ═══════╪═════
> f │ t
> (1 row)
Let's work this out.
The collation weights of the characters involved are:
0030 ; [.209F.0020.0002] # DIGIT ZERO
0031 ; [.20A0.0020.0002] # DIGIT ONE
007C ; [*06AF.0020.0002] # VERTICAL LINE
So under the ICU default "noignore" behavior, the sort keys are:
'|1' => 06AF 20A0 0000 0020 0020 0000 0002 0002
'01' => 209F 20A0 0000 0020 0020 0000 0002 0002
Therefore, '|1' < '01'.
Under "shifted", the effective collation weights are:
0030 ; [.209F.0020.0002.FFFF] # DIGIT ZERO
0031 ; [.20A0.0020.0002.FFFF] # DIGIT ONE
007C ; [.0000.0000.0000.06AF] # VERTICAL LINE
and the sort keys are:
'|1' => 20A0 0000 0020 0000 0002 0000 06AF FFFF
'01' => 209F 20A0 0000 0020 0020 0000 0002 0002 0000 FFFF FFFF
Therefore, '|1' > '01'. That is what you get in glibc. (Older glibc
use "shift-trimmed" behavior, which trims the trailing FFFF's, which
doesn't affect the result in this case.)
Under ICU (or rather CLDR), the '|' character is not actually a variable
collation element by default, so you actually get the same result if you
ask for "noignore" or "shifted" behavior. But you can make symbols
variable with 'und-u-ka-shifted-kv-symbol', and then you will also get
the same result as glibc.
Further reading:
-
https://peter.eisentraut.org/blog/2023/04/12/how-collation-of-punctuation-and-whitespace-works
-
https://peter.eisentraut.org/blog/2023/05/16/overview-of-icu-collation-settings
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Reply to all the recipients using the --to and --cc options:
reply via email
To: pgsql-bugs@postgresql.org
Cc: peter@eisentraut.org, laurenz.albe@cybertec.at, patrick.van.dijk@gmail.com, pgsql-bugs@lists.postgresql.org
Subject: Re: BUG #18523: String compare not consistent
In-Reply-To: <30c9e733-9573-4500-b191-5a8a67b3795f@eisentraut.org>
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
This inbox is served by DDX for PostgreSQL; see mirroring instructions
for how to clone and mirror all data and code used for this inbox