scm_{to,from}_locale_string

scm_{to,from}_locale_string

[Andy Wingo]

Hi Mike,

I'm looking at changing to use the helper "locale_charset()" function
from libunistring in the scm_to_locale_string and scm_from_locale_string
functions.  It seems like that's more correct than snarfing through the
current input/output ports.

Likewise I'll just use the scm_i_get_conversion_strategy for the
handler.

Please let me know if this is the wrong thing to do.

Thanks,

Andy
-- 
http://wingolog.org/

Re: scm_{to,from}_locale_string

[Mike Gran]

> From:Andy Wingo <wingo@pobox.com>
> 
> Hi Mike,
> 
> I'm looking at changing to use the helper "locale_charset()" 
> function
> from libunistring in the scm_to_locale_string and scm_from_locale_string
> functions.  It seems like that's more correct than snarfing through the
> current input/output ports.
> 
> Likewise I'll just use the scm_i_get_conversion_strategy for the
> handler.
> 
> Please let me know if this is the wrong thing to do.

I'm trying to imagine a case where this is wrong.

So, if have a CGI script where the stdout could have one
a couple of different encodings based on a web client's language
preference settings, but, where the CGI program is running in a "C"
or "en_US.utf8" locale, this might count.

But presumably, any snarfing would have been done before
the CGI program reset its stdout locale before displaying its
output web content.

Would that violate your presuppositions?

-Mike

Re: scm_{to,from}_locale_string

[Andy Wingo]

Evening,

On Thu 17 Mar 2011 19:38, Mike Gran <spk121@yahoo.com> writes:

> So, if have a CGI script where the stdout could have one
> a couple of different encodings based on a web client's language
> preference settings, but, where the CGI program is running in a "C"
> or "en_US.utf8" locale, this might count.

The thing is that we need to be giving the C library strings encoded in
the current locale, not the locale of any particular port, because
that's how library functions will decode the strings internally -- take
the regexec case, for example, recently fixed.

I think that if the CGI script wants to read or write in a particular
locale, either you set the LC_ALL/LC_CTYPE vars, do a setlocale in the
Guile script, or set the encoding on a port.  But that's port encodings,
not something related to scm_{to/from}_locale_string, I don't think...

Of course, in the CGI context, the encoding of any _environment
variables_ is another thorny question, but not one that should depend on
port encodings.

Regards,

Andy
-- 
http://wingolog.org/

Re: scm_{to,from}_locale_string

[Ludovic Courtès]

Andy Wingo <wingo@pobox.com> writes:

> The thing is that we need to be giving the C library strings encoded in
> the current locale, not the locale of any particular port, because
> that's how library functions will decode the strings internally -- take
> the regexec case, for example, recently fixed.

Agreed.

Ludo’.