Re: sort -V and accents

[Jim Meyering]

Pierre-Jean wrote:

> Jim Meyering <address@hidden> wrote:
>
>> Pierre-Jean wrote:
>
>> > I'm trying to sort a file containing accents and numbers,
>> > but can't find a way to do this correctly:
>
>
>> The trick is to specify sorting with "-f" for the first column
>> and "-V" for the second.  Then it does what you seem to want:
>>
>>     echo "
>>     A 10
>>     A 9
>>     E 10
>>     E 9
>>     e 10
>>     e 9
>>     é 10
>>     é 9
>>     F 10
>>     f 9" | sort -k1,1f -k2,2V
>>
>>     A 9
>>     A 10
>>     e 9
>>     E 9
>>     e 10
>>     E 10
>>     é 9
>>     é 10
>>     f 9
>>     F 10
>
> This is better, but still not perfect: "é 9" should be
> before "e 10", like "E 9" is before "e 10", as in a
> dictionnary, where "éa" is before "eb". That means that
> e=E=é=è=É=È if something after makes a différence.
>
> Look at this example:
>
> echo "
> é 9
> e 10
> éa
> eb
> E 9" | sort -k1,1f -k2,2V
>
> E 9
> e 10
> é 9
> éa
> eb
>
> "E 9" is correctly moved on first place, but the placement of
> "é 9" doesn't follow the same law.
>
> I probably can do with that "unperfect" order, because such
> situation are not frequent in real life, but if there's a
> solution, I'd be happy to know it.

I left out an important option: you should use -t' ', to tell sort
to use a space as field separator:

      | sort -t' ' -k1,1f -k2,2V

If I'd used sort's --debug option, I would have had a hint about that problem.

Regarding accented and non-accented characters, I would not expect
sort's -f (--ignore-case) option to accomplish what you want.
Ignoring case does not ignore the presence of an accent.

However, there is another option, -d, --dictionary-order,
that you might think would be relevant.  Unfortunately not,
since it merely tells sort to consider only blanks
and alphanumeric characters (ignoring all else).

I don't know how sort can do that.