view lisp/international/utf-8.el @ 46496:395e5c46761b

(utf-8-subst-table) (utf-8-subst-rev-table, utf-8-translation-table-for-decode) (utf-8-fragment-on-decoding, ccl-untranslated-to-ucs) (utf-8-ccl-regs, utf-8-translate-cjk): New. (ccl-encode-mule-utf-8): Use utf-8-subst-rev-table. (ccl-decode-mule-utf-8, ccl-untranslated-to-ucs) (utf-8-untranslated-to-ucs, utf-8-compose): Rewritten. (mule-utf-8): Remove pre-write-conversion. (utf-8-post-read-conversion): Comment out.

;;; utf-8.el --- UTF-8 decoding/encoding support -*- coding: iso-2022-7bit -*-

;; Copyright (C) 2001 Electrotechnical Laboratory, JAPAN.
;; Licensed to the Free Software Foundation.
;; Copyright (C) 2001, 2002 Free Software Foundation, Inc.

;; Author: TAKAHASHI Naoto  <[e-mail]>
;; Maintainer: FSF
;; Keywords: multilingual, Unicode, UTF-8, i18n

;; This file is part of GNU Emacs.

;; GNU Emacs is free software; you can redistribute it and/or modify
;; it under the terms of the GNU General Public License as published by
;; the Free Software Foundation; either version 2, or (at your option)
;; any later version.

;; GNU Emacs is distributed in the hope that it will be useful,
;; but WITHOUT ANY WARRANTY; without even the implied warranty of
;; MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
;; GNU General Public License for more details.

;; You should have received a copy of the GNU General Public License
;; along with GNU Emacs; see the file COPYING.  If not, write to the
;; Free Software Foundation, Inc., 59 Temple Place - Suite 330,
;; Boston, MA 02111-1307, USA.

;;; Commentary:

;; The coding-system `mule-utf-8' basically supports encoding/decoding
;; of the following character sets to and from UTF-8:
;;
;;   ascii
;;   eight-bit-control
;;   latin-iso8859-1
;;   mule-unicode-0100-24ff
;;   mule-unicode-2500-33ff
;;   mule-unicode-e000-ffff
;;
;; On decoding, Unicode characters that do not fit into the above
;; character sets are handled as `eight-bit-control' or
;; `eight-bit-graphic' characters to retain the information about the
;; original byte sequence and text properties record the corresponding
;; unicode.
;;
;; Fixme: note that reading and writing invalid utf-8 may not be
;; idempotent -- to represent the bytes to fix that needs a new charset.
;;
;; Characters from other character sets can be encoded with
;; mule-utf-8 by populating the table `ucs-mule-to-mule-unicode' and
;; registering the translation with `register-char-codings'.  Hash
;; tables `utf-8-subst-table' and `utf-8-subst-rev-table' are used to
;; support encoding and decoding of about a quarter of the CJK space
;; between U+3400 and U+DFFF.

;; UTF-8 is defined in RFC 2279.  A sketch of the encoding is:

;;        scalar       |               utf-8
;;        value        | 1st byte  | 2nd byte  | 3rd byte
;; --------------------+-----------+-----------+----------
;; 0000 0000 0xxx xxxx | 0xxx xxxx |           |
;; 0000 0yyy yyxx xxxx | 110y yyyy | 10xx xxxx |
;; zzzz yyyy yyxx xxxx | 1110 zzzz | 10yy yyyy | 10xx xxxx

;;; Code:

(defvar ucs-mule-to-mule-unicode (make-translation-table)
  "Translation table for encoding to `mule-utf-8'.")
;; Could have been done by ucs-tables loaded before.
(unless (get 'ucs-mule-to-mule-unicode 'translation-table)
  (define-translation-table 'ucs-mule-to-mule-unicode
    ucs-mule-to-mule-unicode))

(defvar utf-8-subst-table (make-hash-table :test 'eq))
(defvar utf-8-subst-rev-table (make-hash-table :test 'eq))
(define-translation-hash-table 'utf-8-subst-table utf-8-subst-table)
(define-translation-hash-table 'utf-8-subst-rev-table utf-8-subst-rev-table)

(defvar utf-8-translation-table-for-decode (make-translation-table)
  "Translation table applied after decoding utf-8 to mule-unicode.
This is only actually applied to characters which would normally be
decoded into mule-unicode-0100-24ff.")
(define-translation-table 'utf-8-translation-table-for-decode
  utf-8-translation-table-for-decode)

;; Map Cyrillic and Greek to iso-8859 charsets, which take half the
;; space of mule-unicode.  For Latin scripts this isn't very
;; important.  Hebrew and Arabic might go here too when there's proper
;; support for them.
(mapc
 (lambda (pair)
   (aset utf-8-translation-table-for-decode (car pair) (cdr pair)))
 '((?$,1&d(B . ?,F4(B) (?$,1&e(B . ?,F5(B) (?$,1&f(B . ?,F6(B) (?$,1&h(B . ?,F8(B) (?$,1&i(B . ?,F9(B)
   (?$,1&j(B . ?,F:(B) (?$,1&l(B . ?,F<(B) (?$,1&n(B . ?,F>(B) (?$,1&o(B . ?,F?(B) (?$,1&p(B . ?,F@(B)
   (?$,1&q(B . ?,FA(B) (?$,1&r(B . ?,FB(B) (?$,1&s(B . ?,FC(B) (?$,1&t(B . ?,FD(B) (?$,1&u(B . ?,FE(B)
   (?$,1&v(B . ?,FF(B) (?$,1&w(B . ?,FG(B) (?$,1&x(B . ?,FH(B) (?$,1&y(B . ?,FI(B) (?$,1&z(B . ?,FJ(B)
   (?$,1&{(B . ?,FK(B) (?$,1&|(B . ?,FL(B) (?$,1&}(B . ?,FM(B) (?$,1&~(B . ?,FN(B) (?$,1&(B . ?,FO(B)
   (?$,1' (B . ?,FP(B) (?$,1'!(B . ?,FQ(B) (?$,1'#(B . ?,FS(B) (?$,1'$(B . ?,FT(B) (?$,1'%(B . ?,FU(B)
   (?$,1'&(B . ?,FV(B) (?$,1''(B . ?,FW(B) (?$,1'((B . ?,FX(B) (?$,1')(B . ?,FY(B) (?$,1'*(B . ?,FZ(B)
   (?$,1'+(B . ?,F[(B) (?$,1',(B . ?,F\(B) (?$,1'-(B . ?,F](B) (?$,1'.(B . ?,F^(B) (?$,1'/(B . ?,F_(B)
   (?$,1'0(B . ?,F`(B) (?$,1'1(B . ?,Fa(B) (?$,1'2(B . ?,Fb(B) (?$,1'3(B . ?,Fc(B) (?$,1'4(B . ?,Fd(B)
   (?$,1'5(B . ?,Fe(B) (?$,1'6(B . ?,Ff(B) (?$,1'7(B . ?,Fg(B) (?$,1'8(B . ?,Fh(B) (?$,1'9(B . ?,Fi(B)
   (?$,1':(B . ?,Fj(B) (?$,1';(B . ?,Fk(B) (?$,1'<(B . ?,Fl(B) (?$,1'=(B . ?,Fm(B) (?$,1'>(B . ?,Fn(B)
   (?$,1'?(B . ?,Fo(B) (?$,1'@(B . ?,Fp(B) (?$,1'A(B . ?,Fq(B) (?$,1'B(B . ?,Fr(B) (?$,1'C(B . ?,Fs(B)
   (?$,1'D(B . ?,Ft(B) (?$,1'E(B . ?,Fu(B) (?$,1'F(B . ?,Fv(B) (?$,1'G(B . ?,Fw(B) (?$,1'H(B . ?,Fx(B)
   (?$,1'I(B . ?,Fy(B) (?$,1'J(B . ?,Fz(B) (?$,1'K(B . ?,F{(B) (?$,1'L(B . ?,F|(B) (?$,1'M(B . ?,F}(B)
   (?$,1'N(B . ?,F~(B)

   (?$,1(!(B . ?,L!(B) (?$,1("(B . ?,L"(B) (?$,1(#(B . ?,L#(B) (?$,1($(B . ?,L$(B)
   (?$,1(%(B . ?,L%(B) (?$,1(&(B . ?,L&(B) (?$,1('(B . ?,L'(B) (?$,1(((B . ?,L((B) (?$,1()(B . ?,L)(B)
   (?$,1(*(B . ?,L*(B) (?$,1(+(B . ?,L+(B) (?$,1(,(B . ?,L,(B) (?$,1(.(B . ?,L.(B) (?$,1(/(B . ?,L/(B)
   (?$,1(0(B . ?,L0(B) (?$,1(1(B . ?,L1(B) (?$,1(2(B . ?,L2(B) (?$,1(3(B . ?,L3(B) (?$,1(4(B . ?,L4(B)
   (?$,1(5(B . ?,L5(B) (?$,1(6(B . ?,L6(B) (?$,1(7(B . ?,L7(B) (?$,1(8(B . ?,L8(B) (?$,1(9(B . ?,L9(B)
   (?$,1(:(B . ?,L:(B) (?$,1(;(B . ?,L;(B) (?$,1(<(B . ?,L<(B) (?$,1(=(B . ?,L=(B) (?$,1(>(B . ?,L>(B)
   (?$,1(?(B . ?,L?(B) (?$,1(@(B . ?,L@(B) (?$,1(A(B . ?,LA(B) (?$,1(B(B . ?,LB(B) (?$,1(C(B . ?,LC(B)
   (?$,1(D(B . ?,LD(B) (?$,1(E(B . ?,LE(B) (?$,1(F(B . ?,LF(B) (?$,1(G(B . ?,LG(B) (?$,1(H(B . ?,LH(B)
   (?$,1(I(B . ?,LI(B) (?$,1(J(B . ?,LJ(B) (?$,1(K(B . ?,LK(B) (?$,1(L(B . ?,LL(B) (?$,1(M(B . ?,LM(B)
   (?$,1(N(B . ?,LN(B) (?$,1(O(B . ?,LO(B) (?$,1(P(B . ?,LP(B) (?$,1(Q(B . ?,LQ(B) (?$,1(R(B . ?,LR(B)
   (?$,1(S(B . ?,LS(B) (?$,1(T(B . ?,LT(B) (?$,1(U(B . ?,LU(B) (?$,1(V(B . ?,LV(B) (?$,1(W(B . ?,LW(B)
   (?$,1(X(B . ?,LX(B) (?$,1(Y(B . ?,LY(B) (?$,1(Z(B . ?,LZ(B) (?$,1([(B . ?,L[(B) (?$,1(\(B . ?,L\(B)
   (?$,1(](B . ?,L](B) (?$,1(^(B . ?,L^(B) (?$,1(_(B . ?,L_(B) (?$,1(`(B . ?,L`(B) (?$,1(a(B . ?,La(B)
   (?$,1(b(B . ?,Lb(B) (?$,1(c(B . ?,Lc(B) (?$,1(d(B . ?,Ld(B) (?$,1(e(B . ?,Le(B) (?$,1(f(B . ?,Lf(B)
   (?$,1(g(B . ?,Lg(B) (?$,1(h(B . ?,Lh(B) (?$,1(i(B . ?,Li(B) (?$,1(j(B . ?,Lj(B) (?$,1(k(B . ?,Lk(B)
   (?$,1(l(B . ?,Ll(B) (?$,1(m(B . ?,Lm(B) (?$,1(n(B . ?,Ln(B) (?$,1(o(B . ?,Lo(B) (?$,1(q(B . ?,Lq(B)
   (?$,1(r(B . ?,Lr(B) (?$,1(s(B . ?,Ls(B) (?$,1(t(B . ?,Lt(B) (?$,1(u(B . ?,Lu(B) (?$,1(v(B . ?,Lv(B)
   (?$,1(w(B . ?,Lw(B) (?$,1(x(B . ?,Lx(B) (?$,1(y(B . ?,Ly(B) (?$,1(z(B . ?,Lz(B) (?$,1({(B . ?,L{(B)
   (?$,1(|(B . ?,L|(B) (?$,1(~(B . ?,L~(B) (?$,1((B . ?,L(B)))

(defcustom utf-8-fragment-on-decoding nil
  "Whether or not to decode some scripts in UTF-8 text into 8-bit characters.
Setting this means that the relevant Cyrillic and Greek characters are
decoded into the iso8859 charsets rather than into
mule-unicode-0100-24ff.  The 8-bit characters take half as much space
in the buffer, but using them may affect how the buffer can be re-encoded
and may require a different input method to search for them, for instance.
See `unify-8859-on-decoding-mode' and `unify-8859-on-encoding-mode'
for mechanisms to make this largely transparent."
  :set (lambda (s v)
	 (if v
	     (define-translation-table 'utf-8-translation-table-for-decode
	       utf-8-translation-table-for-decode)
	   (define-translation-table 'utf-8-translation-table-for-decode))
	 (set-default s v))
  :version "21.4"
  :type 'boolean
  :group 'mule)

(defcustom utf-8-translate-cjk nil
  "Whether the `mule-utf-8' coding system should encode many CJK characters.

Enabling this loads tables which enable the coding system to encode
characters in the charsets `korean-ksc5601', `chinese-gb2312' and
`japanese-jisx0208', and to decode the corresponding unicodes into
such characters.  This works by loading the library `utf-8-subst'; see
its commentary.  The tables are fairly large (about 33000 entries), so this
option is not the default."
  :link '(emacs-commentary-link "utf-8-subst")
  :set (lambda (s v)
	 (when v
	   (require 'utf-8-subst)
	   (let ((table (make-char-table 'translation-table)))
	     (coding-system-put 'mule-utf-8 'safe-charsets
				(append (coding-system-get 'mule-utf-8
							   'safe-charsets)
					'(korean-ksc5601 chinese-gb2312
							 japanese-jisx0208)))
	     (maphash (lambda (k v)
			(aset table k v))
		      utf-8-subst-rev-table)
	     (register-char-codings 'mule-utf-8 table)))
	 (set-default s v))
  :version "21.4"
  :type 'boolean
  :group 'mule)

(define-ccl-program ccl-decode-mule-utf-8
  ;;
  ;;        charset         | bytes in utf-8 | bytes in emacs
  ;; -----------------------+----------------+---------------
  ;;         ascii          |       1        |       1
  ;; -----------------------+----------------+---------------
  ;;    eight-bit-control   |       2        |       2
  ;;    eight-bit-graphic   |       2        |       1
  ;;     latin-iso8859-1    |       2        |       2
  ;; -----------------------+----------------+---------------
  ;; mule-unicode-0100-24ff |       2        |       4
  ;;        (< 0800)        |                |
  ;; -----------------------+----------------+---------------
  ;; mule-unicode-0100-24ff |       3        |       4
  ;;        (>= 8000)       |                |
  ;; mule-unicode-2500-33ff |       3        |       4
  ;; mule-unicode-e000-ffff |       3        |       4
  ;;
  ;; Thus magnification factor is two.
  ;;
  `(2
    ((r5 = ,(charset-id 'eight-bit-control))
     (r6 = ,(charset-id 'eight-bit-graphic))
     (loop
      (read r0)

      ;; 1byte encoding, i.e., ascii
      (if (r0 < #x80)
	  (write r0)
	(if (r0 < #xc0)		    ; continuation byte (invalid here)
	    (if (r0 < #xa0)
		(write-multibyte-character r5 r0)
	      (write-multibyte-character r6 r0))
	  ;; 2 byte encoding 00000yyyyyxxxxxx = 110yyyyy 10xxxxxx
	  (if (r0 < #xe0)
	      ((read r1)

	       (if ((r1 & #b11000000) != #b10000000)
		   ;; Invalid 2-byte sequence
		   ((if (r0 < #xa0)
			(write-multibyte-character r5 r0)
		      (write-multibyte-character r6 r0))
		    (if (r1 < #x80)
			(write r1)
		      (if (r1 < #xa0)
			  (write-multibyte-character r5 r1)
			(write-multibyte-character r6 r1))))

		 ((r3 = r0)	   ; save in case of overlong sequence
		  (r2 = r1)
		  (r0 &= #x1f)
		  (r0 <<= 6)
		  (r2 = r1)	   ; save in case of overlong sequence
		  (r1 &= #x3f)
		  (r1 += r0)
		  ;; Now r1 holds scalar value

		  (if (r1 < 128)	; `overlong sequence'
		      ((if (r3 < #xa0)
			   (write-multibyte-character r5 r3)
			 (write-multibyte-character r6 r3))
		       (if (r2 < #x80)
			   (write r2)
			 (if (r2 < #xa0)
			     (write-multibyte-character r5 r2)
			   (write-multibyte-character r6 r2))))

		    ;; eight-bit-control
		    (if (r1 < 160)
			((write-multibyte-character r5 r1))

		      ;; latin-iso8859-1
		      (if (r1 < 256)
			  ((r0 = ,(charset-id 'latin-iso8859-1))
			   (r1 -= 128)
			   (write-multibyte-character r0 r1))

			;; mule-unicode-0100-24ff (< 0800)
			((r0 = ,(charset-id 'mule-unicode-0100-24ff))
			 (r1 -= #x0100)
			 (r2 = (((r1 / 96) + 32) << 7))
			 (r1 %= 96)
			 (r1 += (r2 + 32))
			 (translate-character
			  utf-8-translation-table-for-decode r0 r1)
			 (write-multibyte-character r0 r1))))))))

	    ;; 3byte encoding
	    ;; zzzzyyyyyyxxxxxx = 1110zzzz 10yyyyyy 10xxxxxx
	    (if (r0 < #xf0)
		((read r1 r2)

		 ;; This is set to 1 if the encoding is invalid.
		 (r4 = 0)

		 (r3 = (r1 & #b11000000))
		 (r3 |= ((r2 >> 2) & #b00110000))
		 (if (r3 != #b10100000)
		     (r4 = 1)
		   ((r3 = ((r0 & #x0f) << 12))
		    (r3 += ((r1 & #x3f) << 6))
		    (r3 += (r2 & #x3f))
		    (if (r3 < #x0800)
			(r4 = 1))))

		 (if (r4 != 0)
		     ;; Invalid 3-byte sequence
		     ((if (r0 < #xa0)
			  (write-multibyte-character r5 r0)
			(write-multibyte-character r6 r0))
		      (if (r1 < #x80)
			  (write r1)
			(if (r1 < #xa0)
			    (write-multibyte-character r5 r1)
			  (write-multibyte-character r6 r1)))
		      (if (r2 < #x80)
			  (write r2)
			(if (r2 < #xa0)
			    (write-multibyte-character r5 r2)
			  (write-multibyte-character r6 r2))))
		 
		   ;; mule-unicode-0100-24ff (>= 0800)
		   ((if (r3 < #x2500)
			((r0 = ,(charset-id 'mule-unicode-0100-24ff))
			 (r3 -= #x0100)
			 (r3 //= 96)
			 (r1 = (r7 + 32))
			 (r1 += ((r3 + 32) << 7))
			 (translate-character
			  utf-8-translation-table-for-decode r0 r1)
			 (write-multibyte-character r0 r1))
		    
		      ;; mule-unicode-2500-33ff
		      ;; Fixme: Perhaps allow translation via
		      ;; utf-8-subst-table for #x2e80 up, so that we use
		      ;; consistent charsets for all of CJK.  Would need
		      ;; corresponding change to encoding tables.
		      (if (r3 < #x3400)
			  ((r0 = ,(charset-id 'mule-unicode-2500-33ff))
			   (r3 -= #x2500)
			   (r3 //= 96)
			   (r1 = (r7 + 32))
			   (r1 += ((r3 + 32) << 7))
			   (write-multibyte-character r0 r1))

			;; U+3400 .. U+D7FF
			;; Try to convert to CJK chars, else keep
			;; them as eight-bit-{control|graphic}.
			(if (r3 < #xd800)
			    ((r4 = r3)	; don't zap r3
			     (lookup-integer utf-8-subst-table r4 r5)
			     (if r7
				 ;; got a translation
				 ((write-multibyte-character r4 r5)
				  ;; Zapped through register starvation.
				  (r5 = ,(charset-id 'eight-bit-control)))
			       ;; #xe0 <= r0 < #xf0, so r0 is eight-bit-graphic
			       ((r3 = r6)
				(write-multibyte-character r3 r0)
				(if (r1 < #xa0)
				    (r3 = r5))
				(write-multibyte-character r3 r1)
				(if (r2 < #xa0)
				    (r3 = r5)
				  (r3 = r6))
				(write-multibyte-character r3 r2))))

			  ;; Surrogates, U+D800 .. U+DFFF
			  ;; Fixme: process them properly.
			  (if (r3 < #xe000)
			      ((r3 = r6)
			       (write-multibyte-character r3 r0) ; eight-bit-graphic
			       (if (r1 < #xa0)
				   (r3 = r5))
			       (write-multibyte-character r3 r1)
			       (if (r2 < #xa0)
				   (r3 = r5)
				 (r3 = r6))
			       (write-multibyte-character r3 r2))
			
			    ;; mule-unicode-e000-ffff
			    ;; Fixme: fffe and ffff are invalid.
			    ((r0 = ,(charset-id 'mule-unicode-e000-ffff))
			     (r3 -= #xe000)
			     (r3 //= 96)
			     (r1 = (r7 + 32))
			     (r1 += ((r3 + 32) << 7))
			     (write-multibyte-character r0 r1)))))))))

	      (if (r0 < #xfe)
		  ;; 4byte encoding
		  ;; keep those bytes as eight-bit-{control|graphic}
		  ;; Fixme: allow lookup in utf-8-subst-table.
		  ((read r1 r2 r3)
		   ;; r0 > #xf0, thus eight-bit-graphic
		   (write-multibyte-character r6 r0)
		   (if (r1 < #xa0)
		       (if (r1 < #x80)	; invalid byte
			   (write r1)
			 (write-multibyte-character r5 r1))
		     (write-multibyte-character r6 r1))
		   (if (r2 < #xa0)
		       (if (r2 < #x80)	; invalid byte
			   (write r2)
			 (write-multibyte-character r5 r2))
		     (write-multibyte-character r6 r2))
		   (if (r3 < #xa0)
		       (if (r3 < #x80)	; invalid byte
			   (write r3)
			 (write-multibyte-character r5 r3))
		     (write-multibyte-character r6 r3))
		   (if (r0 >= #xf8)	; 5- or 6-byte encoding
		       ((read r1)
			(if (r1 < #xa0)
			    (if (r1 < #x80) ; invalid byte
				(write r1)
			      (write-multibyte-character r5 r1))
			  (write-multibyte-character r6 r1))
			(if (r0 >= #xfc) ; 6-byte
			    ((read r1)
			     (if (r1 < #xa0)
				 (if (r1 < #x80) ; invalid byte
				     (write r1)
				   (write-multibyte-character r5 r1))
			       (write-multibyte-character r6 r1)))))))
		;; else invalid byte >= #xfe
		(write-multibyte-character r6 r0))))))
      (repeat))))

  "CCL program to decode UTF-8.
Basic decoding is done into the charsets ascii, latin-iso8859-1 and
mule-unicode-*, but see also `utf-8-translation-table-for-decode' and
`utf-8-subst-table'.
Encodings of un-representable Unicode characters are decoded asis into
eight-bit-control and eight-bit-graphic characters.")

(define-ccl-program ccl-encode-mule-utf-8
  `(1
    ((r5 = -1)
     (loop
      (if (r5 < 0)
	  ((r1 = -1)
	   (read-multibyte-character r0 r1)
	   (translate-character ucs-mule-to-mule-unicode r0 r1))
	(;; We have already done read-multibyte-character.
	 (r0 = r5)
	 (r1 = r6)
	 (r5 = -1)))

      (if (r0 == ,(charset-id 'ascii))
	  (write r1)

	(if (r0 == ,(charset-id 'latin-iso8859-1))
	    ;; r1          scalar                  utf-8
	    ;;       0000 0yyy yyxx xxxx    110y yyyy 10xx xxxx
	    ;; 20    0000 0000 1010 0000    1100 0010 1010 0000
	    ;; 7f    0000 0000 1111 1111    1100 0011 1011 1111
	    ((r0 = (((r1 & #x40) >> 6) | #xc2))
	     (r1 &= #x3f)
	     (r1 |= #x80)
	     (write r0 r1))

	  (if (r0 == ,(charset-id 'mule-unicode-0100-24ff))
	      ((r0 = ((((r1 & #x3f80) >> 7) - 32) * 96))
	       ;; #x3f80 == (0011 1111 1000 0000)b
	       (r1 &= #x7f)
	       (r1 += (r0 + 224))	; 240 == -32 + #x0100
	       ;; now r1 holds scalar value
	       (if (r1 < #x0800)
		   ;; 2byte encoding
		   ((r0 = (((r1 & #x07c0) >> 6) | #xc0))
		    ;; #x07c0 == (0000 0111 1100 0000)b
		    (r1 &= #x3f)
		    (r1 |= #x80)
		    (write r0 r1))
		 ;; 3byte encoding
		 ((r0 = (((r1 & #xf000) >> 12) | #xe0))
		  (r2 = ((r1 & #x3f) | #x80))
		  (r1 &= #x0fc0)
		  (r1 >>= 6)
		  (r1 |= #x80)
		  (write r0 r1 r2))))

	    (if (r0 == ,(charset-id 'mule-unicode-2500-33ff))
		((r0 = ((((r1 & #x3f80) >> 7) - 32) * 96))
		 (r1 &= #x7f)
		 (r1 += (r0 + 9440))	; 9440 == -32 + #x2500
		 (r0 = (((r1 & #xf000) >> 12) | #xe0))
		 (r2 = ((r1 & #x3f) | #x80))
		 (r1 &= #x0fc0)
		 (r1 >>= 6)
		 (r1 |= #x80)
		 (write r0 r1 r2))

	      (if (r0 == ,(charset-id 'mule-unicode-e000-ffff))
		  ((r0 = ((((r1 & #x3f80) >> 7) - 32) * 96))
		   (r1 &= #x7f)
		   (r1 += (r0 + 57312))	; 57312 == -32 + #xe000
		   (r0 = (((r1 & #xf000) >> 12) | #xe0))
		   (r2 = ((r1 & #x3f) | #x80))
		   (r1 &= #x0fc0)
		   (r1 >>= 6)
		   (r1 |= #x80)
		   (write r0 r1 r2))

		(if (r0 == ,(charset-id 'eight-bit-control))
		    ;; r1          scalar                  utf-8
		    ;;       0000 0yyy yyxx xxxx    110y yyyy 10xx xxxx
		    ;; 80    0000 0000 1000 0000    1100 0010 1000 0000
		    ;; 9f    0000 0000 1001 1111    1100 0010 1001 1111
		    ((write #xc2)
		     (write r1))

		  (if (r0 == ,(charset-id 'eight-bit-graphic))
		      ;; r1          scalar                  utf-8
		      ;;       0000 0yyy yyxx xxxx    110y yyyy 10xx xxxx
		      ;; a0    0000 0000 1010 0000    1100 0010 1010 0000
		      ;; ff    0000 0000 1111 1111    1101 1111 1011 1111
		      ((write r1)
		       (r1 = -1)
		       (read-multibyte-character r0 r1)
		       (if (r0 != ,(charset-id 'eight-bit-graphic))
			   (if (r0 != ,(charset-id 'eight-bit-control))
			       ((r5 = r0)
				(r6 = r1))))
		       (if (r5 < 0)
			   ((read-multibyte-character r0 r2)
			    (if (r0 != ,(charset-id 'eight-bit-graphic))
				(if (r0 != ,(charset-id 'eight-bit-control))
				    ((r5 = r0)
				     (r6 = r2))))
			    (if (r5 < 0)
				(write r1 r2)
			      (if (r1 < #xa0)
				  (write r1)
				((write #xc2)
				 (write r1)))))))

		    ((lookup-character utf-8-subst-rev-table r0 r1)
		     (if r7		; lookup succeeded
			 ((r1 = (((r0 & #xf000) >> 12) | #xe0))
			  (r2 = ((r0 & #x3f) | #x80))
			  (r0 &= #x0fc0)
			  (r0 >>= 6)
			  (r0 |= #x80)
			  (write r1 r0 r2))
		       ;; Unsupported character.
		       ;; Output U+FFFD, which is `ef bf bd' in UTF-8.
		       ((write #xef)
			(write #xbf)
			(write #xbd)))))))))))
      (repeat)))
    (if (r1 >= #xa0)
	(write r1)
      (if (r1 >= #x80)
	  ((write #xc2)
	   (write r1)))))

  "CCL program to encode into UTF-8.")

;; Dummy definition so that the CCL can be checked correctly; the
;; actual data are loaded on demand.
(unless (boundp 'ucs-mule-8859-to-mule-unicode)	; don't zap it
  (define-translation-table 'ucs-mule-8859-to-mule-unicode))

(define-ccl-program ccl-untranslated-to-ucs
  `(0
    (if (r0 < #xf0)			; 3-byte encoding, as above
	((r4 = 0)
	 (r3 = (r1 & #b11000000))
	 (r3 |= ((r2 >> 2) & #b00110000))
	 (if (r3 != #b10100000)
	     (r4 = 1)
	   ((r3 = ((r0 & #x0f) << 12))
	    (r3 += ((r1 & #x3f) << 6))
	    (r3 += (r2 & #x3f))
	    (if (r3 < #x0800)
		(r4 = 1))))
	 (if (r4 != 0)
	     (r0 = 0)
	   (r0 = r3)))
      (if (r0 < #xf8)			; 4-byte (Mule-UCS recipe)
	  ((r4 = (r1 >> 6))
	   (if (r4 != #b10)
	       (r0 = 0)
	     ((r4 = (r2 >> 6))
	      (if (r4 != #b10)
		  (r0 = 0)
		((r4 = (r3 >> 6))
		 (if (r4 != #b10)
		     (r0 = 0)
		   ((r1 = ((r1  & #x3F) << 12))
		    (r2 = ((r2  & #x3F) << 6))
		    (r3 &= #x3F)
		    (r0 = (((((r0 & #x07) << 18) | r1) | r2) | r3)))))))))
	(r0 = 0))))
  "Decode 3- or 4-byte sequences in r0, r1, r2 [,r3] to unicodes in r0.
r0 == 0 for invalid sequence.")

(defvar utf-8-ccl-regs (make-vector 8 0))

(defsubst utf-8-untranslated-to-ucs ()
  "Return the UCS code for an untranslated sequence of raw bytes t point.
Only for 3- or 4-byte sequences."
  (aset utf-8-ccl-regs 0 (or (char-after) 0))
  (aset utf-8-ccl-regs 1 (or (char-after (1+ (point))) 0))
  (aset utf-8-ccl-regs 2 (or (char-after (+ 2 (point))) 0))
  (aset utf-8-ccl-regs 3 (or (char-after (+ 3 (point))) 0))
  (ccl-execute 'ccl-untranslated-to-ucs utf-8-ccl-regs)
  (aref utf-8-ccl-regs 0))

(defun utf-8-help-echo (window object position)
  (format "Untranslated Unicode U+%04X"
	  (get-char-property position 'untranslated-utf-8 object)))

;; We compose the untranslatable sequences into a single character.
;; This is infelicitous for editing, because there's currently no
;; mechanism for treating compositions as atomic, but is OK for
;; display.  They are composed to U+FFFD with help-echo which
;; indicates the unicodes they represent.  This function GCs too much.
(defsubst utf-8-compose ()
  "Put a suitable composition on an untranslatable sequence.
Return the sequence's length."
  (let* ((u (utf-8-untranslated-to-ucs))
	 (l (unless (zerop u)
	      (if (>= u #x10000)
		       4
		     3))))
    (when l
      (put-text-property (point) (min (point-max) (+ l (point)))
			 'untranslated-utf-8 u)
      (put-text-property (point) (min (point-max) (+ l (point)))
			 'help-echo 'utf-8-help-echo)
      (compose-region (point) (+ l (point)) ?$,3u=(B)
      l)))

(defcustom utf-8-compose-scripts nil
  "*Non-nil means compose various scripts on decoding utf-8 text."
  :group 'mule
  :version "21.4"
  :type 'boolean)

(defun utf-8-post-read-conversion (length)
  "Compose untranslated utf-8 sequences into single characters.
Also compose particular scripts if `utf-8-compose-scripts' is non-nil."
  (save-excursion
    ;; Can't do eval-when-compile to insert a multibyte constant
    ;; version of the string in the loop, since it's always loaded as
    ;; unibyte from a byte-compiled file.
    (let ((range (string-as-multibyte "^\xe1-\xf7")))
      (while (and (skip-chars-forward range)
		  (not (eobp)))
	(forward-char (utf-8-compose)))))
  ;; Fixme: Takahashi-san implies it may not work this easily.  I
  ;; asked why but didn't get a reply. -- fx
  (when (and utf-8-compose-scripts (> length 1))
    ;; These currently have definitions which cover the relevant
    ;; unicodes.  We could avoid loading thai-util &c by checking
    ;; whether the region contains any characters with the appropriate
    ;; categories.  There aren't yet Unicode-based rules for Tibetan.
    (save-excursion (setq length (diacritic-post-read-conversion length)))
    (save-excursion (setq length (thai-post-read-conversion length)))
    (save-excursion (setq length (lao-post-read-conversion length)))
    (save-excursion
      (setq length (in-is13194-devanagari-post-read-conversion length))))
  length)

;; ucs-tables is preloaded
;; (defun utf-8-pre-write-conversion (beg end)
;;   "Semi-dummy pre-write function effectively to autoload ucs-tables."
;;   ;; Ensure translation table is loaded.
;;   (require 'ucs-tables)
;;   ;; Don't do this again.
;;   (coding-system-put 'mule-utf-8 'pre-write-conversion nil)
;;   nil)

(make-coding-system
 'mule-utf-8 4 ?u
 "UTF-8 encoding for Emacs-supported Unicode characters.
The supported Emacs character sets are the following, plus any other
characters included in the tables `ucs-mule-to-mule-unicode' and
`utf-8-subst-rev-table':
 ascii
 eight-bit-control
 eight-bit-graphic
 latin-iso8859-1
 latin-iso8859-2
 latin-iso8859-3
 latin-iso8859-4
 cyrillic-iso8859-5
 greek-iso8859-7
 hebrew-iso8859-8
 latin-iso8859-9
 latin-iso8859-14
 latin-iso8859-15
 mule-unicode-0100-24ff
 mule-unicode-2500-33ff
 mule-unicode-e000-ffff

Unicode characters out of the ranges U+0000-U+33FF and U+E200-U+FFFF
may be decoded into korean-ksc5601, chinese-gb2312, japanese-jisx0208
\(see user option `utf-8-translate-cjk'); otherwise, sequences of
eight-bit-control and eight-bit-graphic characters are used to
preserve their byte sequences, and these are composed to display as a
single character.  Emacs characters that otherwise can't be encoded
are encoded as U+FFFD."

 '(ccl-decode-mule-utf-8 . ccl-encode-mule-utf-8)
 '((safe-charsets
    ascii
    eight-bit-control
    eight-bit-graphic
    latin-iso8859-1
    latin-iso8859-15
    latin-iso8859-14
    latin-iso8859-9
    hebrew-iso8859-8
    greek-iso8859-7
    cyrillic-iso8859-5
    latin-iso8859-4
    latin-iso8859-3
    latin-iso8859-2
    vietnamese-viscii-lower
    vietnamese-viscii-upper
    thai-tis620
    ipa
    ethiopic
    indian-is13194
    katakana-jisx0201
    chinese-sisheng
    lao
    mule-unicode-0100-24ff
    mule-unicode-2500-33ff
    mule-unicode-e000-ffff)
   (mime-charset . utf-8)
   (coding-category . coding-category-utf-8)
   (valid-codes (0 . 255))
;;    (pre-write-conversion . utf-8-pre-write-conversion)
   (post-read-conversion . utf-8-post-read-conversion)))

(define-coding-system-alias 'utf-8 'mule-utf-8)

;; I think this needs special private charsets defined for the
;; untranslated sequences, if it's going to work well.

;;; (defun utf-8-compose-function (pos to pattern &optional string)
;;;   (let* ((prop (get-char-property pos 'composition string))
;;; 	 (l (and prop (- (cadr prop) (car prop)))))
;;;     (cond ((and l (> l (- to pos)))
;;; 	   (delete-region pos to))
;;; 	  ((and (> (char-after pos) 224)
;;; 		(< (char-after pos) 256)
;;; 		(save-restriction
;;; 		  (narrow-to-region pos to)
;;; 		  (utf-8-compose)))
;;; 	   t))))

;;; (dotimes (i 96)
;;;   (aset composition-function-table
;;; 	(+ 128 i)
;;; 	`((,(string-as-multibyte "[\200-\237\240-\377]")
;;; 	   . utf-8-compose-function))))

;;; utf-8.el ends here