Re: sort and collation

[Pádraig Brady]

On 26/11/2024 04:31, Magnus Fromreide wrote:
> Hello!
>
> I tried to sort a file containing the lines
> [X].Y
> [X].W
> [XZ]
> [XX]
> and got confused when sort returned
> [X].W
> [XX]
> [X].Y
> [XZ]
> rather than
> [XX]
> [XZ]
> [X].W
> [X].Y
> where the [X] values where kept together.
>
> It seems that the problem is that my locale (en_US.utf8) specifies that
> punctuation, like [, ], and ., is insignificant.
>
> This is actually very close to what the -d flag to sort does but in my case
> it is just confusing.
>
> * Would it be possible to have a flag that tells sort to treat all input
>    characters as unique entities and just compare their values?
>
> * Would it be possible to combine that flag with the ignore case flag or
>    does input characters where case transformation adds multiple characters
>    kill that idea?
>
> * Would it be possible to have a flag that tells sort to just perform an
>    octet comparision, similar to 'LC_ALL=C sort' but slightly more ergonomic?

Well `LC_ALL=C sort -f` is the canonical way to do as you suggest,
which works with all sort versions and implementations.
I'm not sure it's worth adding a flag to do this.

thanks,
Pádraig.