Re: gunicode.h

[Owen Taylor <otaylor redhat com>]

Havoc Pennington <hp@redhat.com> writes:

> Owen Taylor <otaylor@redhat.com> writes:
> > gunichar g_utf8_get_char        (gchar       *iter);
> > gchar *  g_utf8_iter_from_index (const gchar *str,
> > 				 gint         index);
> > gint     g_utf8_iter_to_index   (const gchar *str,
> > 				 const gchar *pos);
> 
> I think the use of "iter" is cryptic here, I like "pointer" more
> i.e. g_utf8_pointer_from_index() g_utf8_pointer_to_index().

Probably. I originally called all these functions:

 g_iter_get_char(), etc.

But decided that was not very clear.
 
> > gchar *  g_utf8_next_char       (gchar       *iter);
> > gchar *  g_utf8_prev_char       (gchar       *iter);
> 
> These args should be const if "pos" above is const, right?

Yes.
 
> > gchar *  g_utf8_find_next_char  (gchar       *iter,
> > 				 const gchar *bound);
> > gchar *  g_utf8_find_prev_char  (const gchar *str,
> > 				 gchar       *iter);
> > 
> 
> Why is bound const but not iter?

No particular reason. 

> > /* Copies n characters from src to dest */
> > gchar *g_utf8_strncpy (gchar       *dest,
> > 		       const gchar *src,
> > 		       size_t       n);
> >
> 
> Use of size_t here seems weird since size_t is almost always used to
> count bytes. I would think just int, there's no sense in which this n
> is a memory offset or size.

The reason why this is size_t is basically just copying strncpy. Note
that that 'n' counts things which are at smallest just bytes.

Note that size_t is extensively used for things other than bytes in 
in the standard C lib.

 void qsort (void *base, size_t nmemb, size_t size,
             int (*compar)(const void *, const void *))

Note the size_t nmemb.

An int could be smaller than the possible size of 'n'. If you have
a 5 gig string on a 64-bit machine, size_t, works but not char.

Of course we extensively use int in places like this throughout glib,
so this argument doesn't hold a lot of water.
 
> > gunichar2 *g_utf8_to_ucs2 (const char *str,
> > 			 gint        len);
> > gunichar *g_utf8_to_ucs4 (const char *str,
> > 			 int         len);
> >
> 
> Include the inverse of these, maybe? (g_ucs4_to_utf8() etc.)

Yes. I think so. (Though that gives 6 functions.) I do want to a malloc'ing iconv 
wrapper, so you can do:

 char *result = g_convert ("utf-8", "ucs-2", (char *)input);

 if (!result)
  g_warning ("Cannot convert string");

Or something like that. But for efficiency and convenient 

Thanks for the feedback,
                                              Owen