Re: ustring::validate() costs?

[Matthias Kaeppler <matthias finitestate org>]

> The right way to do it is to set the G_FILENAME_ENCODING or G_BROKEN_FILENAMES 
> environmental variable in /etc/profile or an executable shell file 
> in /etc/profile.d, depending on your distribution (which is how system wide 
> environmental variable should be set), and then use Glib::filename_to_utf8().  
> This will enable your code to be usable (if G_FILENAME_ENCODING is 
> appropriately set by the user) with filenames in other than the UTF-8 or the 
> local system's locale codeset.

Okay, I see. I guess my system is really messed up then, since I 
obviously have exactly this problem: Some (or even most) of my files are 
neither encoded in UTF-8 nor in my locale's encoding (that's why 
g_filename_to_utf8() fails because I haven't set these variables properly).

It would be nice to figure out in my program for /each/ file I read, in 
which character set it is encoded. Is this possible? I only found 
functions so far which can either read the locale's character set or 
check if some filename is valid UTF-8 (or not), but no function which 
individually probes for a certain file in which character set its 
filename is encoded.

This would solve my problem, since I could then use Glib::convert() 
directly to convert from and to the correct encoding for each filename 
individually. Maybe this is how Nautilus does it (since it doesn't have 
any problems reading my files, I guess it has at least /some/ clever 
mechanism to detect for each file if a conversion is necessary and if 
so, from which character encoding it has to convert to UTF-8). Looks as 
if I have to peek at the Nautilus source again, but if you have any good 
ideas how to solve this without setting these environment variables I'm 
all ears :)

Thanks,
Matthias