Re: Order By weirdness?

Hi Carl,

This can be related to glibc2.38 update. I recommend you to check following
documents.

https://postgresql.verite.pro/blog/2018/08/27/glibc-upgrade.html
https://wiki.postgresql.org/wiki/Locale_data_changes

Best regards.
Samed YILDIRIM

On Sat, 7 Jan 2023 at 22:35, Carl Sopchak <carl(at)sopchak(dot)me> wrote:

> I'm seeing something (very) unexpected with ORDER BY. If I run this query:
>
> select txt
> from ( values('x12345'), ('xz1234'), ('x23456'), ('xz2345'), ('x34567'),
> ('xz3456') ) a(txt)
> order by txt;
>
> I get expected results with x<#> being sorted before xz. However, if I
> replace the z's with ~, giving
>
> select txt
> from ( values('x12345'), ('x~1234'), ('x23456'), ('x~2345'), ('x34567'),
> ('x~3456') ) a(txt)
> order by txt;
>
> I get this???
>
> txt
> --------
> x~1234
> x12345
> x~2345
> x23456
> x~3456
> x34567
>
> Which appears to mean that ~ is treated differently than z (basically ~ is
> ignored). Same if I use other special characters, such as @.
>
> Up until stumbling into this, I have never seen such behavior from a
> database. (Windows OS, yes, but I won't go there...) Character-based text
> always sorted in an alphabetic order (which puts special characters in
> different places in the ordering depending on encoding, but it's
> consistent).
>
> Two questions (which may be the same way of asking the same question):
>
> - How is this correct? I can see where this could be useful in limited
> scenarios, but IMHO it makes no sense as a default sort order.
>
> - What do I need to do to get a strictly character-based sort in ORDER BY?
>
> I am using postgres version 14.3 on Fedora 37.
>
> Thanks for the help.
>
> Carl
>
>
>