[PATCH] (Draft2) Make X11 server-side fonts and Mule suck less.

[stephen at xemacs.org]

Aidan Kehoe writes:

In the ChangeLog:

 > 	Stop using the `registry' charset property; use `registries' 
 > 	instead. The difference is that registries is an ordered vector of 
 > 	X11 registries and encodings rather than a regexp; this means we 
 > 	can leave the matching to the X11 server, avoiding transferring 
 > 	huge amounts of data (perhaps across the network!) in order to do 
 > 	a regexp search on it. 

Isn't this gratuitous difference with GNU?

There ought to be a way to get the best of both worlds.  First, since
`charset-registry' already matches multiple charsets in the general
case, we don't need to change the name.  Then

(1) Accept either a string (regexp) or vector of strings (explicit).

(2) If a regexp, use the old method, if a vector use yours.

(3) It's trivial to construct a regexp from a vector of strings, so
    that's what `charset-registry' should return.

(4) `make-charset', `set-charset-registry', and friends (if they have
    any! ;-) should warn on being handed a string.

(5) Bonus points for determining that a regexp is actually a
    disjunction of fixed strings and turning it into a vector
    internally.  This is mostly a joke except for the presumably very
    common and useful case of a single fixed string.  (If I thought
    GNU were going to stick with Mule indefinitely I'd consider it an
    important compatibility feature, but I doubt they will.)

In the ChangeLog:

 > 	* mule/mule-charset.el (charset-registries): New. 
 > 	* mule/mule-charset.el (set-charset-registry): New. 

Isn't `set-charset-registry' in the above a typo?

 > src/ChangeLog addition:
 > 
 > 2006-10-31  Aidan Kehoe  <kehoea at parhasard.net>
 > 
 > 	* charset.h (XCHARSET_CCL_PROGRAM):
 > 	* charset.h (XCHARSET_NAME):
 > 	Make dummy versions of these available in non-Mule.

Other things being equal, I doubt this is a good idea.  Why do you want it?

 > 	* console-impl.h:
 > 	* console-impl.h (struct console_methods):
 > 	Rename the last parameter to a couple of methods; reformat their
 > 	declarations. 

The above will cause some annoyance to CHISE, Carbon XEmacs, and
possibly SXEmacs someday.  Is it a good idea?

Can the rest of these be split out?  I want them!

 > 2006-10-31  Aidan Kehoe  <kehoea at parhasard.net>
 > 
 > 	* faces.el (face-property-matching-instance):
 > 	Simplify. 
 > 	* faces.el (face-font-instance):
 > 	Document CHARSET. 
 > 	* faces.el (set-face-font):
 >         Give more details on common values for font instantiators, 
 >         LOCALEs.
 > 	* unicode.el:
 > 	Remove a few comments that were only relevant to GNU Emacs.
 > 	* unicode.el (decode-char):
 > 	* unicode.el (encode-char):

 > src/ChangeLog addition:
 > 
 > 2006-10-31  Aidan Kehoe  <kehoea at parhasard.net>
 > 
 > 	
 > 	* faces.c:
 > 	* faces.c (face_property_matching_instance):
 > 	* faces.c (ensure_face_cachel_contains_charset):
 > 	* faces.c (merge_face_cachel_data):
 > 	* faces.c (reset_face_cachel):
 > 	* faces.c (mark_face_cachels_as_not_updated):
 > 	* faces.c (syms_of_faces):
 > 	* faces.c (vars_of_faces):
 > 	* faces.c (complex_vars_of_faces):
 > 	Provide a DEBUG_FACES macro; use it to make debugging output
 > 	available in debug builds.
 > 	Implement multi-stage font lookup, assigning the stages names, not
 > 	numbers.
 > 	Re-implement the cachel->font_specified cache using the
 > 	infrastructure for Lisp bit vectors. 
 > 	
 > 	* faces.h:
 > 	* faces.h (struct face_cachel):
 > 	* faces.h (FACE_CACHEL_FONT_UPDATED):
 > 	* faces.h (FACE_FONT):
 > 	Re-implement the cachel->font_specified cache using the
 > 	infrastructure for Lisp bit vectors. 
 > 
 > 	* font-mgr.h:
 > 	Move some XFT debug macros here from objects-x.c. 
 > 
 > 	* general-slots.h:
 > 	Provide a few new symbols for the multi-stage font resolution
 > 	process. 
 > 
 > 	* intl.c (init_intl):
 > 	Correct a comment. 
 > 	
 > 	* objects-gtk.c:
 > 	* objects-gtk.c (gtk_font_spec_matches_charset):
 > 	* objects-gtk.c (gtk_find_charset_font):
 > 	* objects-msw.c (mswindows_find_charset_font):
 > 	* objects-tty.c (tty_find_charset_font):
 > 	Redeclare various functions to work with the multi-stage lookup
 > 	process.
 > 
 > 	* objects-x.c (count_hyphens):
 > 	New. How many ASCII minus characters in a string? 
 > 
 > 	* objects.c (print_font_instance):
 > 	* objects.c (font_spec_matches_charset):
 > 	* objects.c (font_validate_matchspec):
 > 	* objects.c (font_instantiate):
 > 	Redeclare some methods to take enums rather than numeric stages. 
 > 
 > 	* objects.h (EXFUN):
 > 	Make Fregexp_quote available to mule-charset.c
 > 	
 > 	* redisplay-x.c:
 > 	* redisplay-x.c (separate_textual_runs):
 > 	Make this slightly faster, cleaner. Make it accept a face cachel
 > 	pointer argument, and check it as to whether a given charset
 > 	should be translated to UCS-2 before redisplay. 
 > 	
 > 	* specifier.c:
 > 	* specifier.c (charset_matches_specifier_tag_set_p):
 > 	* specifier.c (define_specifier_tag):
 > 	* specifier.c (Fdefine_specifier_tag):
 > 	* specifier.c (setup_device_initial_specifier_tags):
 > 	* specifier.c (setup_charset_initial_specifier_tags):
 > 	* specifier.c (specifier_instance_from_inst_list):
 > 	* specifier.c (syms_of_specifier):
 > 	* specifier.c (vars_of_specifier):
 > 	* specifier.h:
 > 
 > 	Extend specifiers to allow limiting their applicability by using
 > 	charset predicates. Document this. 
 > 
 > 	* unicode.c (unicode_to_ichar):
 > 	* unicode.c (syms_of_unicode):
 > 	* unicode.c (vars_of_unicode):
 > 	Use unicode-registries, a dumped vector, as the charset-registries
 > 	of the on-the-fly JIT charsets.