Re: SRFI-14 and locale settings

[Neil Jerram]

address@hidden (Ludovic Courtès) writes:

> Hi,
>
> Neil Jerram <address@hidden> writes:
>
>> Yes.  So it seems to me, therefore, that we should not be using
>> isalpha() etc. to construct char-set:letter, but should instead hard
>> code it as the intersection of (char-set:letter as specified by SRFI
>> 14) with (the set of characters that Guile can represent).
>
> In practice, I can think of two ways to determine the set of _letters_
> available in the current encoding (which is what `char-set:letter'
> expects).
>
> 1. Since SRFI-14 lists all the characters that have to be added to the
>    ASCII `char-set:letter' to get the Latin-1 `char-set:letter', we
>    could somehow hard-code them.  But this is ugly.

I don't see why you think it's ugly.  If it's the right solution, it's
the right solution.

> 2. Or, we can use a predicate that uses the `is' functions which we
>    expect to be language-independent (i.e., those functions that only
>    depend on the locale's charset), such as:
>
>      (!isblank (c)) && (!ispunct (c)) && (!isdigit (c)) && (!iscntrl (c))

Now this is ugly, IMO!

>    This is certainly not perfect, but it should work for Latin-1, and
>    hopefully for other 8-bit charsets as well.
>
> As Kevin mentioned earlier, all the char sets could be re-computed in
> `scm_setlocale ()'.

This sounds even trickier, and wrong, given that the intention of SRFI
14 is for char-set:letter to be locale-independent.

Regards,
     Neil