Hi,

Kevin Ryde <user42@zip.com.au> writes:

> By way of concrete suggestion, something like
>
> 	(locale-weekday N)
> 	(locale-weekday-short N)
> 	(locale-month N)
> 	(locale-month-short N)
>
> or similar names would be much nicer than messing about with
>
> 	(... (+ ABDAY_1 N))
> 	(... (+ MON_1 N -1))    ;; if N is 1 to 12

The patch below (against HEAD) addresses some of the concerns that were
raised earlier, notably:

  1. The awful, C-ish `nl_langinfo'-based API is now replaced by a shiny
     Schemey API, following Kevin's suggestion.  `nl-langinfo' itself is
     only used internally and not exported.  As a result, most of the
     `nl-langinfo'-based procedures are written in Scheme.

  2. Since `nl-langinfo' is hidden from Scheme programmers, all the
     `locale-' procedures are guaranteed to work, no matter whether
     `nl_langinfo ()' is actually available (see the feature-conditional
     macros in `i18n.scm').

     As a consequence, SRFI-19 internationalization is now
     straightforward as does not require feature-conditional code.

  3. Previously uninterpreted `nl_langinfo ()' special return values are
     now correctly interpreted.  For instance, the return value for
     `GROUPING' should not be interpreted as a string, as was the case
     with my earlier patch.

  4. No attempt is made to convert to a different encoding the strings
     returned by `nl_langinfo ()'.

  5. High-level procedures for locale-dependent number output are
     provided, namely `number->locale-string' and
     `monetary-amount->locale-string'.  Both of them are implemented in
     Scheme (which is relieving, really ;-)).

  6. Various bugs in the C code base were fixed, a few improvements were
     made (notably slight optimizations for systems without `locale_t').

  7. The test suite and documentation were augmented.

I'm well aware that the patch is hard to read.  Perhaps the most
important things to start with are the API in `i18n.scm' and the
corresponding documentation and test cases.

Thanks,
Ludovic.