lisp/international/utf-8.el

   1 ;;; utf-8.el --- UTF-8 decoding/encoding support -*- coding: iso-2022-7bit -*-
   2
   3 ;; Copyright (C) 2001, 2004 Electrotechnical Laboratory, JAPAN.
   4 ;; Licensed to the Free Software Foundation.
   5 ;; Copyright (C) 2001, 2002 Free Software Foundation, Inc.
   6
   7 ;; Author: TAKAHASHI Naoto  <ntakahas@m17n.org>
   8 ;; Maintainer: FSF
   9 ;; Keywords: multilingual, Unicode, UTF-8, i18n
  10
  11 ;; This file is part of GNU Emacs.
  12
  13 ;; GNU Emacs is free software; you can redistribute it and/or modify
  14 ;; it under the terms of the GNU General Public License as published by
  15 ;; the Free Software Foundation; either version 2, or (at your option)
  16 ;; any later version.
  17
  18 ;; GNU Emacs is distributed in the hope that it will be useful,
  19 ;; but WITHOUT ANY WARRANTY; without even the implied warranty of
  20 ;; MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  21 ;; GNU General Public License for more details.
  22
  23 ;; You should have received a copy of the GNU General Public License
  24 ;; along with GNU Emacs; see the file COPYING.  If not, write to the
  25 ;; Free Software Foundation, Inc., 59 Temple Place - Suite 330,
  26 ;; Boston, MA 02111-1307, USA.
  27
  28 ;;; Commentary:
  29
  30 ;; The coding-system `mule-utf-8' basically supports encoding/decoding
  31 ;; of the following character sets to and from UTF-8:
  32 ;;
  33 ;;   ascii
  34 ;;   eight-bit-control
  35 ;;   latin-iso8859-1
  36 ;;   mule-unicode-0100-24ff
  37 ;;   mule-unicode-2500-33ff
  38 ;;   mule-unicode-e000-ffff
  39 ;;
  40 ;; On decoding, Unicode characters that do not fit into the above
  41 ;; character sets are handled as `eight-bit-control' or
  42 ;; `eight-bit-graphic' characters to retain the information about the
  43 ;; original byte sequence and text properties record the corresponding
  44 ;; unicode.
  45 ;;
  46 ;; Fixme: note that reading and writing invalid utf-8 may not be
  47 ;; idempotent -- to represent the bytes to fix that needs a new charset.
  48 ;;
  49 ;; Characters from other character sets can be encoded with mule-utf-8
  50 ;; by populating the translation table
  51 ;; `utf-translation-table-for-encode'.  Hash tables
  52 ;; `utf-subst-table-for-decode' and `utf-subst-table-for-encode' are
  53 ;; used to support encoding and decoding of about a quarter of the CJK
  54 ;; space between U+3400 and U+DFFF.
  55
  56 ;; UTF-8 is defined in RFC 3629.  A sketch of the encoding is:
  57
  58 ;;        scalar       |               utf-8
  59 ;;        value        | 1st byte  | 2nd byte  | 3rd byte
  60 ;; --------------------+-----------+-----------+----------
  61 ;; 0000 0000 0xxx xxxx | 0xxx xxxx |           |
  62 ;; 0000 0yyy yyxx xxxx | 110y yyyy | 10xx xxxx |
  63 ;; zzzz yyyy yyxx xxxx | 1110 zzzz | 10yy yyyy | 10xx xxxx
  64
  65 ;;; Code:
  66
  67 (defvar ucs-mule-to-mule-unicode (make-char-table 'translation-table nil)
  68   "Char table mapping characters to latin-iso8859-1 or mule-unicode-*.
  69
  70 If `unify-8859-on-encoding-mode' is non-nil, this table populates the
  71 translation-table named `utf-translation-table-for-encode'.")
  72
  73 (define-translation-table 'utf-translation-table-for-encode)
  74
  75
  76 ;; Map Cyrillic and Greek to iso-8859 charsets, which take half the
  77 ;; space of mule-unicode.  For Latin scripts this isn't very
  78 ;; important.  Hebrew and Arabic might go here too when there's proper
  79 ;; support for them.
  80
  81 (defvar utf-fragmentation-table (make-char-table 'translation-table nil)
  82   "Char-table normally mapping non-Latin mule-unicode-* chars to iso-8859-*.
  83
  84 If `utf-fragment-on-decoding' is non-nil, this table populates the
  85 translation-table named `utf-translation-table-for-decode'")
  86
  87 (defvar utf-defragmentation-table (make-char-table 'translation-table nil)
  88   "Char-table for reverse mapping of `utf-fragmentation-table'.
  89
  90 If `utf-fragment-on-decoding' is non-nil and
  91 `unify-8859-on-encoding-mode' is nil, this table populates the
  92 translation-table named `utf-translation-table-for-encode'")
  93
  94 (define-translation-table 'utf-translation-table-for-decode)
  95
  96
  97 (defvar ucs-mule-cjk-to-unicode (make-hash-table :test 'eq)
  98   "Hash table mapping Emacs CJK character sets to Unicode code points.
  99
 100 If `utf-translate-cjk-mode' is non-nil, this table populates the
 101 translation-hash-table named `utf-subst-table-for-encode'.")
 102
 103 (define-translation-hash-table 'utf-subst-table-for-encode
 104   ucs-mule-cjk-to-unicode)
 105
 106 (defvar ucs-unicode-to-mule-cjk (make-hash-table :test 'eq)
 107   "Hash table mapping Unicode code points to Emacs CJK character sets.
 108
 109 If `utf-translate-cjk-mode' is non-nil, this table populates the
 110 translation-hash-table named `utf-subst-table-for-decode'.")
 111
 112 (define-translation-hash-table 'utf-subst-table-for-decode
 113   ucs-unicode-to-mule-cjk)
 114
 115 (mapc
 116  (lambda (pair)
 117    (aset utf-fragmentation-table (car pair) (cdr pair))
 118    (aset utf-defragmentation-table (cdr pair) (car pair)))
 119  '((?\e$,1&d\e(B . ?\e,F4\e(B) (?\e$,1&e\e(B . ?\e,F5\e(B) (?\e$,1&f\e(B . ?\e,F6\e(B) (?\e$,1&h\e(B . ?\e,F8\e(B) (?\e$,1&i\e(B . ?\e,F9\e(B)
 120    (?\e$,1&j\e(B . ?\e,F:\e(B) (?\e$,1&l\e(B . ?\e,F<\e(B) (?\e$,1&n\e(B . ?\e,F>\e(B) (?\e$,1&o\e(B . ?\e,F?\e(B) (?\e$,1&p\e(B . ?\e,F@\e(B)
 121    (?\e$,1&q\e(B . ?\e,FA\e(B) (?\e$,1&r\e(B . ?\e,FB\e(B) (?\e$,1&s\e(B . ?\e,FC\e(B) (?\e$,1&t\e(B . ?\e,FD\e(B) (?\e$,1&u\e(B . ?\e,FE\e(B)
 122    (?\e$,1&v\e(B . ?\e,FF\e(B) (?\e$,1&w\e(B . ?\e,FG\e(B) (?\e$,1&x\e(B . ?\e,FH\e(B) (?\e$,1&y\e(B . ?\e,FI\e(B) (?\e$,1&z\e(B . ?\e,FJ\e(B)
 123    (?\e$,1&{\e(B . ?\e,FK\e(B) (?\e$,1&|\e(B . ?\e,FL\e(B) (?\e$,1&}\e(B . ?\e,FM\e(B) (?\e$,1&~\e(B . ?\e,FN\e(B) (?\e$,1&\7f\e(B . ?\e,FO\e(B)
 124    (?\e$,1' \e(B . ?\e,FP\e(B) (?\e$,1'!\e(B . ?\e,FQ\e(B) (?\e$,1'#\e(B . ?\e,FS\e(B) (?\e$,1'$\e(B . ?\e,FT\e(B) (?\e$,1'%\e(B . ?\e,FU\e(B)
 125    (?\e$,1'&\e(B . ?\e,FV\e(B) (?\e$,1''\e(B . ?\e,FW\e(B) (?\e$,1'(\e(B . ?\e,FX\e(B) (?\e$,1')\e(B . ?\e,FY\e(B) (?\e$,1'*\e(B . ?\e,FZ\e(B)
 126    (?\e$,1'+\e(B . ?\e,F[\e(B) (?\e$,1',\e(B . ?\e,F\\e(B) (?\e$,1'-\e(B . ?\e,F]\e(B) (?\e$,1'.\e(B . ?\e,F^\e(B) (?\e$,1'/\e(B . ?\e,F_\e(B)
 127    (?\e$,1'0\e(B . ?\e,F`\e(B) (?\e$,1'1\e(B . ?\e,Fa\e(B) (?\e$,1'2\e(B . ?\e,Fb\e(B) (?\e$,1'3\e(B . ?\e,Fc\e(B) (?\e$,1'4\e(B . ?\e,Fd\e(B)
 128    (?\e$,1'5\e(B . ?\e,Fe\e(B) (?\e$,1'6\e(B . ?\e,Ff\e(B) (?\e$,1'7\e(B . ?\e,Fg\e(B) (?\e$,1'8\e(B . ?\e,Fh\e(B) (?\e$,1'9\e(B . ?\e,Fi\e(B)
 129    (?\e$,1':\e(B . ?\e,Fj\e(B) (?\e$,1';\e(B . ?\e,Fk\e(B) (?\e$,1'<\e(B . ?\e,Fl\e(B) (?\e$,1'=\e(B . ?\e,Fm\e(B) (?\e$,1'>\e(B . ?\e,Fn\e(B)
 130    (?\e$,1'?\e(B . ?\e,Fo\e(B) (?\e$,1'@\e(B . ?\e,Fp\e(B) (?\e$,1'A\e(B . ?\e,Fq\e(B) (?\e$,1'B\e(B . ?\e,Fr\e(B) (?\e$,1'C\e(B . ?\e,Fs\e(B)
 131    (?\e$,1'D\e(B . ?\e,Ft\e(B) (?\e$,1'E\e(B . ?\e,Fu\e(B) (?\e$,1'F\e(B . ?\e,Fv\e(B) (?\e$,1'G\e(B . ?\e,Fw\e(B) (?\e$,1'H\e(B . ?\e,Fx\e(B)
 132    (?\e$,1'I\e(B . ?\e,Fy\e(B) (?\e$,1'J\e(B . ?\e,Fz\e(B) (?\e$,1'K\e(B . ?\e,F{\e(B) (?\e$,1'L\e(B . ?\e,F|\e(B) (?\e$,1'M\e(B . ?\e,F}\e(B)
 133    (?\e$,1'N\e(B . ?\e,F~\e(B)
 134
 135    (?\e$,1(!\e(B . ?\e,L!\e(B) (?\e$,1("\e(B . ?\e,L"\e(B) (?\e$,1(#\e(B . ?\e,L#\e(B) (?\e$,1($\e(B . ?\e,L$\e(B)
 136    (?\e$,1(%\e(B . ?\e,L%\e(B) (?\e$,1(&\e(B . ?\e,L&\e(B) (?\e$,1('\e(B . ?\e,L'\e(B) (?\e$,1((\e(B . ?\e,L(\e(B) (?\e$,1()\e(B . ?\e,L)\e(B)
 137    (?\e$,1(*\e(B . ?\e,L*\e(B) (?\e$,1(+\e(B . ?\e,L+\e(B) (?\e$,1(,\e(B . ?\e,L,\e(B) (?\e$,1(.\e(B . ?\e,L.\e(B) (?\e$,1(/\e(B . ?\e,L/\e(B)
 138    (?\e$,1(0\e(B . ?\e,L0\e(B) (?\e$,1(1\e(B . ?\e,L1\e(B) (?\e$,1(2\e(B . ?\e,L2\e(B) (?\e$,1(3\e(B . ?\e,L3\e(B) (?\e$,1(4\e(B . ?\e,L4\e(B)
 139    (?\e$,1(5\e(B . ?\e,L5\e(B) (?\e$,1(6\e(B . ?\e,L6\e(B) (?\e$,1(7\e(B . ?\e,L7\e(B) (?\e$,1(8\e(B . ?\e,L8\e(B) (?\e$,1(9\e(B . ?\e,L9\e(B)
 140    (?\e$,1(:\e(B . ?\e,L:\e(B) (?\e$,1(;\e(B . ?\e,L;\e(B) (?\e$,1(<\e(B . ?\e,L<\e(B) (?\e$,1(=\e(B . ?\e,L=\e(B) (?\e$,1(>\e(B . ?\e,L>\e(B)
 141    (?\e$,1(?\e(B . ?\e,L?\e(B) (?\e$,1(@\e(B . ?\e,L@\e(B) (?\e$,1(A\e(B . ?\e,LA\e(B) (?\e$,1(B\e(B . ?\e,LB\e(B) (?\e$,1(C\e(B . ?\e,LC\e(B)
 142    (?\e$,1(D\e(B . ?\e,LD\e(B) (?\e$,1(E\e(B . ?\e,LE\e(B) (?\e$,1(F\e(B . ?\e,LF\e(B) (?\e$,1(G\e(B . ?\e,LG\e(B) (?\e$,1(H\e(B . ?\e,LH\e(B)
 143    (?\e$,1(I\e(B . ?\e,LI\e(B) (?\e$,1(J\e(B . ?\e,LJ\e(B) (?\e$,1(K\e(B . ?\e,LK\e(B) (?\e$,1(L\e(B . ?\e,LL\e(B) (?\e$,1(M\e(B . ?\e,LM\e(B)
 144    (?\e$,1(N\e(B . ?\e,LN\e(B) (?\e$,1(O\e(B . ?\e,LO\e(B) (?\e$,1(P\e(B . ?\e,LP\e(B) (?\e$,1(Q\e(B . ?\e,LQ\e(B) (?\e$,1(R\e(B . ?\e,LR\e(B)
 145    (?\e$,1(S\e(B . ?\e,LS\e(B) (?\e$,1(T\e(B . ?\e,LT\e(B) (?\e$,1(U\e(B . ?\e,LU\e(B) (?\e$,1(V\e(B . ?\e,LV\e(B) (?\e$,1(W\e(B . ?\e,LW\e(B)
 146    (?\e$,1(X\e(B . ?\e,LX\e(B) (?\e$,1(Y\e(B . ?\e,LY\e(B) (?\e$,1(Z\e(B . ?\e,LZ\e(B) (?\e$,1([\e(B . ?\e,L[\e(B) (?\e$,1(\\e(B . ?\e,L\\e(B)
 147    (?\e$,1(]\e(B . ?\e,L]\e(B) (?\e$,1(^\e(B . ?\e,L^\e(B) (?\e$,1(_\e(B . ?\e,L_\e(B) (?\e$,1(`\e(B . ?\e,L`\e(B) (?\e$,1(a\e(B . ?\e,La\e(B)
 148    (?\e$,1(b\e(B . ?\e,Lb\e(B) (?\e$,1(c\e(B . ?\e,Lc\e(B) (?\e$,1(d\e(B . ?\e,Ld\e(B) (?\e$,1(e\e(B . ?\e,Le\e(B) (?\e$,1(f\e(B . ?\e,Lf\e(B)
 149    (?\e$,1(g\e(B . ?\e,Lg\e(B) (?\e$,1(h\e(B . ?\e,Lh\e(B) (?\e$,1(i\e(B . ?\e,Li\e(B) (?\e$,1(j\e(B . ?\e,Lj\e(B) (?\e$,1(k\e(B . ?\e,Lk\e(B)
 150    (?\e$,1(l\e(B . ?\e,Ll\e(B) (?\e$,1(m\e(B . ?\e,Lm\e(B) (?\e$,1(n\e(B . ?\e,Ln\e(B) (?\e$,1(o\e(B . ?\e,Lo\e(B) (?\e$,1(q\e(B . ?\e,Lq\e(B)
 151    (?\e$,1(r\e(B . ?\e,Lr\e(B) (?\e$,1(s\e(B . ?\e,Ls\e(B) (?\e$,1(t\e(B . ?\e,Lt\e(B) (?\e$,1(u\e(B . ?\e,Lu\e(B) (?\e$,1(v\e(B . ?\e,Lv\e(B)
 152    (?\e$,1(w\e(B . ?\e,Lw\e(B) (?\e$,1(x\e(B . ?\e,Lx\e(B) (?\e$,1(y\e(B . ?\e,Ly\e(B) (?\e$,1(z\e(B . ?\e,Lz\e(B) (?\e$,1({\e(B . ?\e,L{\e(B)
 153    (?\e$,1(|\e(B . ?\e,L|\e(B) (?\e$,1(~\e(B . ?\e,L~\e(B) (?\e$,1(\7f\e(B . ?\e,L\7f\e(B)))
 154
 155
 156 (defcustom utf-fragment-on-decoding nil
 157   "Whether or not to decode some chars in UTF-8/16 text into iso8859 charsets.
 158 Setting this means that the relevant Cyrillic and Greek characters are
 159 decoded into the iso8859 charsets rather than into
 160 mule-unicode-0100-24ff.  The iso8859 charsets take half as much space
 161 in the buffer, but using them may affect how the buffer can be re-encoded
 162 and may require a different input method to search for them, for instance.
 163 See `unify-8859-on-decoding-mode' and `unify-8859-on-encoding-mode'
 164 for mechanisms to make this largely transparent.
 165
 166 Setting this variable outside customize has no effect."
 167   :set (lambda (s v)
 168          (if v
 169              (progn
 170                (define-translation-table 'utf-translation-table-for-decode
 171                  utf-fragmentation-table)
 172                ;; Even if unify-8859-on-encoding-mode is off, make
 173                ;; mule-utf-* encode characters in
 174                ;; utf-fragmentation-table.
 175                (unless (eq (get 'utf-translation-table-for-encode
 176                                 'translation-table)
 177                            ucs-mule-to-mule-unicode)
 178                  (define-translation-table 'utf-translation-table-for-encode
 179                    utf-defragmentation-table)))
 180            (define-translation-table 'utf-translation-table-for-decode)
 181            ;; When unify-8859-on-encoding-mode is off, be sure to make
 182            ;; mule-utf-* disabled for characters in
 183            ;; utf-fragmentation-table.
 184            (unless (eq (get 'utf-translation-table-for-encode
 185                             'translation-table)
 186                        ucs-mule-to-mule-unicode)
 187              (define-translation-table 'utf-translation-table-for-encode)))
 188          (set-default s v))
 189   :version "21.4"
 190   :type 'boolean
 191   :group 'mule)
 192
 193
 194 (defconst utf-translate-cjk-charsets '(chinese-gb2312
 195                                        chinese-big5-1 chinese-big5-2
 196                                        japanese-jisx0208 japanese-jisx0212
 197                                        korean-ksc5601)
 198   "List of charsets supported by `utf-translate-cjk-mode'.")
 199
 200 (defvar utf-translate-cjk-lang-env nil
 201   "Language environment in which tables for `utf-translate-cjk-mode' is loaded.
 202 The value nil means that the tables are not yet loaded.")
 203
 204 (defvar utf-translate-cjk-unicode-range)
 205
 206 ;; String generated from utf-translate-cjk-unicode-range.  It is
 207 ;; suitable for an argument to skip-chars-forward.
 208 (defvar utf-translate-cjk-unicode-range-string nil)
 209
 210 (defun utf-translate-cjk-set-unicode-range (range)
 211   (setq utf-translate-cjk-unicode-range range)
 212   (setq utf-translate-cjk-unicode-range-string
 213         (let ((decode-char-no-trans
 214                #'(lambda (x)
 215                    (cond ((< x #x100) (make-char 'latin-iso8859-1 x))
 216                          ((< x #x2500)
 217                           (setq x (- x #x100))
 218                           (make-char 'mule-unicode-0100-24ff
 219                                      (+ (/ x 96) 32) (+ (% x 96) 32)))
 220                          ((< x #x3400)
 221                           (setq x (- x #x2500))
 222                           (make-char 'mule-unicode-2500-33ff
 223                                      (+ (/ x 96) 32) (+ (% x 96) 32)))
 224                          (t
 225                           (setq x (- x #xe000))
 226                           (make-char 'mule-unicode-e000-ffff
 227                                      (+ (/ x 96) 32) (+ (% x 96) 32))))))
 228               ranges from to)
 229           (dolist (elt range)
 230             (setq from (max #xA0 (car elt)) to (min #xffff (cdr elt)))
 231             (if (and (>= to #x3400) (< to #xE000))
 232                 (setq to #x33FF))
 233             (cond ((< from #x100)
 234                    (if (>= to #xE000)
 235                        (setq ranges (cons (cons #xE000 to) ranges)
 236                              to #x33FF))
 237                    (if (>= to #x2500)
 238                        (setq ranges (cons (cons #x2500 to) ranges)
 239                              to #x24FF))
 240                    (if (>= to #x100)
 241                        (setq ranges (cons (cons #x100 to) ranges)
 242                              to #xFF)))
 243                   ((< from #x2500)
 244                    (if (>= to #xE000)
 245                        (setq ranges (cons (cons #xE000 to) ranges)
 246                              to #x33FF))
 247                    (if (>= to #x2500)
 248                        (setq ranges (cons (cons #x2500 to) ranges)
 249                              to #x24FF)))
 250                   ((< from #x3400)
 251                    (if (>= to #xE000)
 252                        (setq ranges (cons (cons #xE000 to) ranges)
 253                              to #x33FF))))
 254             (if (<= from to)
 255                 (setq ranges (cons (cons from to) ranges))))
 256           (mapconcat #'(lambda (x)
 257                          (format "%c-%c"
 258                                  (funcall decode-char-no-trans (car x))
 259                                  (funcall decode-char-no-trans (cdr x))))
 260                      ranges "")))
 261   ;; These forces loading and settting tables for
 262   ;; utf-translate-cjk-mode.
 263   (setq utf-translate-cjk-lang-env nil
 264         ucs-mule-cjk-to-unicode (make-hash-table :test 'eq)
 265         ucs-unicode-to-mule-cjk (make-hash-table :test 'eq)))
 266
 267 (defcustom utf-translate-cjk-unicode-range '((#x2e80 . #xd7a3)
 268                                              (#xff00 . #xffef))
 269   "List of Unicode code ranges supported by `utf-translate-cjk-mode'.
 270 Setting this variable directly does not take effect;
 271 use either \\[customize] or the function
 272 `utf-translate-cjk-set-unicode-range'."
 273   :version "21.4"
 274   :type '(repeat (cons integer integer))
 275   :set (lambda (symbol value)
 276          (utf-translate-cjk-set-unicode-range value))
 277   :group 'mule)
 278
 279 ;; Return non-nil if CODE-POINT is in `utf-translate-cjk-unicode-range'.
 280 (defsubst utf-translate-cjk-substitutable-p (code-point)
 281   (let ((tail utf-translate-cjk-unicode-range)
 282         elt)
 283     (while tail
 284       (setq elt (car tail) tail (cdr tail))
 285       (if (and (>= code-point (car elt)) (<= code-point (cdr elt)))
 286           (setq tail nil)
 287         (setq elt nil)))
 288     elt))
 289
 290 (defun utf-translate-cjk-load-tables ()
 291   "Load tables for `utf-translate-cjk-mode'."
 292   ;; Fixme: Allow the use of the CJK charsets to be
 293   ;; customized by reordering and possible omission.
 294   (let ((redefined (< (hash-table-size ucs-mule-cjk-to-unicode) 43000)))
 295     (if redefined
 296         ;; Redefine them with realistic initial sizes and a
 297         ;; smallish rehash size to avoid wasting significant
 298         ;; space after they're built.
 299         (setq ucs-mule-cjk-to-unicode
 300               (make-hash-table :test 'eq :size 43000 :rehash-size 1000)
 301               ucs-unicode-to-mule-cjk
 302               (make-hash-table :test 'eq :size 21500 :rehash-size 1000)))
 303
 304     ;; Load the files explicitly, to avoid having to keep
 305     ;; around the large tables they contain (as well as the
 306     ;; ones which get built).
 307     (cond ((string= "Korean" current-language-environment)
 308            (load "subst-jis")
 309            (load "subst-big5")
 310            (load "subst-gb2312")
 311            (load "subst-ksc"))
 312           ((string= "Chinese-BIG5" current-language-environment)
 313            (load "subst-jis")
 314            (load "subst-ksc")
 315            (load "subst-gb2312")
 316            (load "subst-big5"))
 317           ((string= "Chinese-GB" current-language-environment)
 318            (load "subst-jis")
 319            (load "subst-ksc")
 320            (load "subst-big5")
 321            (load "subst-gb2312"))
 322           (t
 323            (load "subst-ksc")
 324            (load "subst-gb2312")
 325            (load "subst-big5")
 326            (load "subst-jis")))   ; jis covers as much as big5, gb2312
 327
 328     (when redefined
 329       (define-translation-hash-table 'utf-subst-table-for-decode
 330         ucs-unicode-to-mule-cjk)
 331       (define-translation-hash-table 'utf-subst-table-for-encode
 332         ucs-mule-cjk-to-unicode)
 333       (set-char-table-extra-slot (get 'utf-translation-table-for-encode
 334                                       'translation-table)
 335                                  1 ucs-mule-cjk-to-unicode))
 336
 337     (setq utf-translate-cjk-lang-env current-language-environment)))
 338
 339 (defun utf-lookup-subst-table-for-decode (code-point)
 340   (if (and utf-translate-cjk-mode
 341            (not utf-translate-cjk-lang-env)
 342            (utf-translate-cjk-substitutable-p code-point))
 343       (utf-translate-cjk-load-tables))
 344   (gethash code-point
 345            (get 'utf-subst-table-for-decode 'translation-hash-table)))
 346
 347
 348 (defun utf-lookup-subst-table-for-encode (char)
 349   (if (and utf-translate-cjk-mode
 350            (not utf-translate-cjk-lang-env)
 351            (memq (char-charset char) utf-translate-cjk-charsets))
 352       (utf-translate-cjk-load-tables))
 353   (gethash char
 354            (get 'utf-subst-table-for-encode 'translation-hash-table)))
 355
 356 (define-minor-mode utf-translate-cjk-mode
 357   "Toggle whether UTF based coding systems de/encode CJK characters.
 358 If ARG is an integer, enable if ARG is positive and disable if
 359 zero or negative.  This is a minor mode.
 360 Enabling this allows the coding systems mule-utf-8,
 361 mule-utf-16le and mule-utf-16be to encode characters in the charsets
 362 `korean-ksc5601', `chinese-gb2312', `chinese-big5-1',
 363 `chinese-big5-2', `japanese-jisx0208' and `japanese-jisx0212', and to
 364 decode the corresponding unicodes into such characters.
 365
 366 Where the charsets overlap, the one preferred for decoding is chosen
 367 according to the language environment in effect when this option is
 368 turned on: ksc5601 for Korean, gb2312 for Chinese-GB, big5 for
 369 Chinese-Big5 and jisx for other environments.
 370
 371 This mode is on by default.  If you are not interested in CJK
 372 characters and want to avoid some overhead on encoding/decoding
 373 by the above coding systems, you can customize the user option
 374 `utf-translate-cjk-mode' to nil."
 375   :init-value t
 376   :version "21.4"
 377   :type 'boolean
 378   :group 'mule
 379   :global t
 380   (if utf-translate-cjk-mode
 381       (progn
 382         (define-translation-hash-table 'utf-subst-table-for-decode
 383           ucs-unicode-to-mule-cjk)
 384         (define-translation-hash-table 'utf-subst-table-for-encode
 385           ucs-mule-cjk-to-unicode)
 386         (set-char-table-extra-slot (get 'utf-translation-table-for-encode
 387                                         'translation-table)
 388                                    1 ucs-mule-cjk-to-unicode))
 389     (define-translation-hash-table 'utf-subst-table-for-decode
 390       (make-hash-table :test 'eq))
 391     (define-translation-hash-table 'utf-subst-table-for-encode
 392       (make-hash-table :test 'eq))
 393     (set-char-table-extra-slot (get 'utf-translation-table-for-encode
 394                                     'translation-table)
 395                                1 nil))
 396
 397   ;; Update safe-chars of mule-utf-* coding systems.
 398   (dolist (elt (coding-system-list t))
 399     (if (string-match "^mule-utf" (symbol-name elt))
 400         (let ((safe-charsets (coding-system-get elt 'safe-charsets))
 401               (safe-chars (coding-system-get elt 'safe-chars))
 402               (need-update nil))
 403           (dolist (charset utf-translate-cjk-charsets)
 404             (unless (eq utf-translate-cjk-mode (memq charset safe-charsets))
 405               (setq safe-charsets
 406                     (if utf-translate-cjk-mode
 407                         (cons charset safe-charsets)
 408                       (delq charset safe-charsets))
 409                     need-update t)
 410               (aset safe-chars (make-char charset) utf-translate-cjk-mode)))
 411           (when need-update
 412             (coding-system-put elt 'safe-charsets safe-charsets)
 413             (define-coding-system-internal elt))))))
 414
 415 (define-ccl-program ccl-mule-utf-untrans
 416   ;; R0 is an untranslatable Unicode code-point (U+3500..U+DFFF or
 417   ;; U+10000..U+10FFFF) or an invaid byte (#x00..#xFF).  Write
 418   ;; eight-bit-control/graphic sequence (2 to 4 chars) representing
 419   ;; UTF-8 sequence of r0.  Registers r4, r5, r6 are modified.
 420   ;;
 421   ;; This is a subrountine because we assume that this is called very
 422   ;; rarely (so we don't have to worry about the overhead of the
 423   ;; call).
 424   `(0
 425     ((r5 = ,(charset-id 'eight-bit-control))
 426      (r6 = ,(charset-id 'eight-bit-graphic))
 427      (if (r0 < #x100)
 428          ((r4 = ((r0 >> 6) | #xC0))
 429           (write-multibyte-character r6 r4))
 430        ((if (r0 < #x10000)
 431             ((r4 = ((r0 >> 12) | #xE0))
 432              (write-multibyte-character r6 r4))
 433           ((r4 = ((r0 >> 18) | #xF0))
 434            (write-multibyte-character r6 r4)
 435            (r4 = (((r0 >> 12) & #x3F) | #x80))
 436            (if (r4 < #xA0)
 437                (write-multibyte-character r5 r4)
 438              (write-multibyte-character r6 r4))))
 439         (r4 = (((r0 >> 6) & #x3F) | #x80))
 440         (if (r4 < #xA0)
 441             (write-multibyte-character r5 r4)
 442           (write-multibyte-character r6 r4))))
 443      (r4 = ((r0 & #x3F) | #x80))
 444      (if (r4 < #xA0)
 445          (write-multibyte-character r5 r4)
 446        (write-multibyte-character r6 r4)))))
 447
 448 (define-ccl-program ccl-decode-mule-utf-8
 449   ;;
 450   ;;        charset         | bytes in utf-8 | bytes in emacs
 451   ;; -----------------------+----------------+---------------
 452   ;;         ascii          |       1        |       1
 453   ;; -----------------------+----------------+---------------
 454   ;;    eight-bit-control   |       2        |       2
 455   ;;    eight-bit-graphic   |       2        |       1
 456   ;;     latin-iso8859-1    |       2        |       2
 457   ;; -----------------------+----------------+---------------
 458   ;; mule-unicode-0100-24ff |       2        |       4
 459   ;;        (< 0800)        |                |
 460   ;; -----------------------+----------------+---------------
 461   ;; mule-unicode-0100-24ff |       3        |       4
 462   ;;        (>= 8000)       |                |
 463   ;; mule-unicode-2500-33ff |       3        |       4
 464   ;; mule-unicode-e000-ffff |       3        |       4
 465   ;; -----------------------+----------------+---------------
 466   ;;      invalid byte      |       1        |       2
 467   ;;
 468   ;; Thus magnification factor is two.
 469   ;;
 470   `(2
 471     ((r6 = ,(charset-id 'latin-iso8859-1))
 472      (read r0)
 473      (loop
 474       (if (r0 < #x80)
 475           ;; 1-byte encoding, i.e., ascii
 476           (write-read-repeat r0))
 477       (if (r0 < #xc2)
 478           ;; continuation byte (invalid here) or 1st byte of overlong
 479           ;; 2-byte sequence.
 480           ((call ccl-mule-utf-untrans)
 481            (r6 = ,(charset-id 'latin-iso8859-1))
 482            (read r0)
 483            (repeat)))
 484
 485       ;; Read the 2nd byte.
 486       (read r1)
 487       (if ((r1 & #b11000000) != #b10000000) ; Invalid 2nd byte
 488           ((call ccl-mule-utf-untrans)
 489            (r6 = ,(charset-id 'latin-iso8859-1))
 490            ;; Handle it in the next loop.
 491            (r0 = r1)
 492            (repeat)))
 493
 494       (if (r0 < #xe0)
 495           ;; 2-byte encoding 00000yyyyyxxxxxx = 110yyyyy 10xxxxxx
 496           ((r1 &= #x3F)
 497            (r1 |= ((r0 & #x1F) << 6))
 498            ;; Now r1 holds scalar value.  We don't have to check
 499            ;; `overlong sequence' because r0 >= 0xC2.
 500
 501            (if (r1 >= 256)
 502                ;; mule-unicode-0100-24ff (< 0800)
 503                ((r0 = r1)
 504                 (lookup-integer utf-subst-table-for-decode r0 r1)
 505                 (if (r7 == 0)
 506                     ((r0 = ,(charset-id 'mule-unicode-0100-24ff))
 507                      (r1 -= #x0100)
 508                      (r2 = (((r1 / 96) + 32) << 7))
 509                      (r1 %= 96)
 510                      (r1 += (r2 + 32))
 511                      (translate-character
 512                       utf-translation-table-for-decode r0 r1)))
 513                 (write-multibyte-character r0 r1)
 514                 (read r0)
 515                 (repeat))
 516              (if (r1 >= 160)
 517                  ;; latin-iso8859-1
 518                  ((r0 = r1)
 519                   (lookup-integer utf-subst-table-for-decode r0 r1)
 520                   (if (r7 == 0)
 521                       ((r1 -= 128)
 522                        (write-multibyte-character r6 r1))
 523                     ((write-multibyte-character r0 r1)))
 524                   (read r0)
 525                   (repeat))
 526                ;; eight-bit-control
 527                ((r0 = ,(charset-id 'eight-bit-control))
 528                 (write-multibyte-character r0 r1)
 529                 (read r0)
 530                 (repeat))))))
 531
 532       ;; Read the 3rd bytes.
 533       (read r2)
 534       (if ((r2 & #b11000000) != #b10000000) ; Invalid 3rd byte
 535           ((call ccl-mule-utf-untrans)
 536            (r0 = r1)
 537            (call ccl-mule-utf-untrans)
 538            (r6 = ,(charset-id 'latin-iso8859-1))
 539            ;; Handle it in the next loop.
 540            (r0 = r2)
 541            (repeat)))
 542
 543       (if (r0 < #xF0)
 544           ;; 3byte encoding
 545           ;; zzzzyyyyyyxxxxxx = 1110zzzz 10yyyyyy 10xxxxxx
 546           ((r3 = ((r0 & #xF) << 12))
 547            (r3 |= ((r1 & #x3F) << 6))
 548            (r3 |= (r2 & #x3F))
 549
 550            (if (r3 < #x800)             ; `overlong sequence'
 551                ((call ccl-mule-utf-untrans)
 552                 (r0 = r1)
 553                 (call ccl-mule-utf-untrans)
 554                 (r0 = r2)
 555                 (call ccl-mule-utf-untrans)
 556                 (r6 = ,(charset-id 'latin-iso8859-1))
 557                 (read r0)
 558                 (repeat)))
 559
 560            (if (r3 < #x2500)
 561                ;; mule-unicode-0100-24ff (>= 0800)
 562                ((r0 = r3)
 563                 (lookup-integer utf-subst-table-for-decode r0 r1)
 564                 (if (r7 == 0)
 565                     ((r0 = ,(charset-id 'mule-unicode-0100-24ff))
 566                      (r3 -= #x0100)
 567                      (r3 //= 96)
 568                      (r1 = (r7 + 32))
 569                      (r1 += ((r3 + 32) << 7))
 570                      (translate-character
 571                       utf-translation-table-for-decode r0 r1)))
 572                 (write-multibyte-character r0 r1)
 573                 (read r0)
 574                 (repeat)))
 575
 576            (if (r3 < #x3400)
 577                ;; mule-unicode-2500-33ff
 578                ((r0 = r3)               ; don't zap r3
 579                 (lookup-integer utf-subst-table-for-decode r0 r1)
 580                 (if (r7 == 0)
 581                     ((r0 = ,(charset-id 'mule-unicode-2500-33ff))
 582                      (r3 -= #x2500)
 583                      (r3 //= 96)
 584                      (r1 = (r7 + 32))
 585                      (r1 += ((r3 + 32) << 7))))
 586                 (write-multibyte-character r0 r1)
 587                 (read r0)
 588                 (repeat)))
 589
 590            (if (r3 < #xE000)
 591                ;; Try to convert to CJK chars, else
 592                ;; keep them as eight-bit-{control|graphic}.
 593                ((r0 = r3)
 594                 (lookup-integer utf-subst-table-for-decode r3 r1)
 595                 (if r7
 596                     ;; got a translation
 597                     ((write-multibyte-character r3 r1)
 598                      (read r0)
 599                      (repeat))
 600                   ((call ccl-mule-utf-untrans)
 601                    (r6 = ,(charset-id 'latin-iso8859-1))
 602                    (read r0)
 603                    (repeat)))))
 604
 605            ;; mule-unicode-e000-ffff
 606            ;; Fixme: fffe and ffff are invalid.
 607            (r0 = r3)            ; don't zap r3
 608            (lookup-integer utf-subst-table-for-decode r0 r1)
 609            (if (r7 == 0)
 610                ((r0 = ,(charset-id 'mule-unicode-e000-ffff))
 611                 (r3 -= #xe000)
 612                 (r3 //= 96)
 613                 (r1 = (r7 + 32))
 614                 (r1 += ((r3 + 32) << 7))))
 615            (write-multibyte-character r0 r1)
 616            (read r0)
 617            (repeat)))
 618
 619       ;; Read the 4th bytes.
 620       (read r3)
 621       (if ((r3 & #b11000000) != #b10000000) ; Invalid 4th byte
 622           ((call ccl-mule-utf-untrans)
 623            (r0 = r1)
 624            (call ccl-mule-utf-untrans)
 625            (r0 = r2)
 626            (call ccl-mule-utf-untrans)
 627            (r6 = ,(charset-id 'latin-iso8859-1))
 628            ;; Handle it in the next loop.
 629            (r0 = r3)
 630            (repeat)))
 631
 632       (if (r0 < #xF8)
 633           ;; 4-byte encoding:
 634           ;; wwwzzzzzzyyyyyyxxxxxx = 11110www 10zzzzzz 10yyyyyy 10xxxxxx
 635           ;; keep those bytes as eight-bit-{control|graphic}
 636           ;; Fixme: allow lookup in utf-subst-table-for-decode.
 637           ((r4 = ((r0 & #x7) << 18))
 638            (r4 |= ((r1 & #x3F) << 12))
 639            (r4 |= ((r2 & #x3F) << 6))
 640            (r4 |= (r3 & #x3F))
 641
 642            (if (r4 < #x10000)           ; `overlong sequence'
 643                ((call ccl-mule-utf-untrans)
 644                 (r0 = r1)
 645                 (call ccl-mule-utf-untrans)
 646                 (r0 = r2)
 647                 (call ccl-mule-utf-untrans)
 648                 (r0 = r3)
 649                 (call ccl-mule-utf-untrans))
 650              ((r0 = r4)
 651               (call ccl-mule-utf-untrans))))
 652
 653         ;; Unsupported sequence.
 654         ((call ccl-mule-utf-untrans)
 655          (r0 = r1)
 656          (call ccl-mule-utf-untrans)
 657          (r0 = r2)
 658          (call ccl-mule-utf-untrans)
 659          (r0 = r3)
 660          (call ccl-mule-utf-untrans)))
 661       (r6 = ,(charset-id 'latin-iso8859-1))
 662       (read r0)
 663       (repeat)))
 664
 665
 666     ;; At EOF...
 667     (if (r0 >= 0)
 668         ;; r0 >= #x80
 669         ((call ccl-mule-utf-untrans)
 670          (if (r1 >= 0)
 671              ((r0 = r1)
 672               (call ccl-mule-utf-untrans)
 673               (if (r2 >= 0)
 674                   ((r0 = r2)
 675                    (call ccl-mule-utf-untrans)
 676                    (if (r3 >= 0)
 677                        ((r0 = r3)
 678                         (call ccl-mule-utf-untrans))))))))))
 679
 680   "CCL program to decode UTF-8.
 681 Basic decoding is done into the charsets ascii, latin-iso8859-1 and
 682 mule-unicode-*, but see also `utf-fragmentation-table' and
 683 `ucs-mule-cjk-to-unicode'.
 684 Encodings of un-representable Unicode characters are decoded asis into
 685 eight-bit-control and eight-bit-graphic characters.")
 686
 687 (define-ccl-program ccl-mule-utf-8-encode-untrans
 688   ;; UTF-8 decoder generates an UTF-8 sequence represented by a
 689   ;; sequence eight-bit-control/graphic chars for an untranslatable
 690   ;; character and an invalid byte.
 691   ;;
 692   ;; This CCL parses that sequence (the first byte is already in r1),
 693   ;; writes out the original bytes of that sequence, and sets r5 to
 694   ;; -1.
 695   ;;
 696   ;; If the eight-bit-control/graphic sequence is shorter than what r1
 697   ;; suggests, it sets r5 and r6 to the last character read that
 698   ;; should be handled by the next loop of a caller.
 699   ;;
 700   ;; Note: For UTF-8 validation, we only check if a character is
 701   ;; eight-bit-control/graphic or not.  It may result in incorrect
 702   ;; handling of random binary data, but such a data can't be encoded
 703   ;; by UTF-8 anyway.  At least, UTF-8 decoders doesn't generate such
 704   ;; a sequence even if a source contains invalid byte-sequence.
 705   `(0
 706     (;; Read the 2nd byte.
 707      (read-multibyte-character r5 r6)
 708      (r0 = (r5 != ,(charset-id 'eight-bit-control)))
 709      (if ((r5 != ,(charset-id 'eight-bit-graphic)) & r0)
 710          ((write r1)                    ; invalid UTF-8
 711           (r1 = -1)
 712           (end)))
 713
 714      (if (r1 <= #xC3)
 715          ;; 2-byte sequence for an originally invalid byte.
 716          ((r6 &= #x3F)
 717           (r6 |= ((r1 & #x1F) << 6))
 718           (write r6)
 719           (r5 = -1)
 720           (end)))
 721
 722      (write r1 r6)
 723      (r2 = r1)
 724      (r1 = -1)
 725      ;; Read the 3rd byte.
 726      (read-multibyte-character r5 r6)
 727      (r0 = (r5 != ,(charset-id 'eight-bit-control)))
 728      (if ((r5 != ,(charset-id 'eight-bit-graphic)) & r0)
 729          (end))                         ; invalid UTF-8
 730      (write r6)
 731      (if (r2 < #xF0)
 732          ;; 3-byte sequence for an untranslated character.
 733          ((r5 = -1)
 734           (end)))
 735      ;; Read the 4th byte.
 736      (read-multibyte-character r5 r6)
 737      (r0 = (r5 != ,(charset-id 'eight-bit-control)))
 738      (if ((r5 != ,(charset-id 'eight-bit-graphic)) & r0)
 739          (end))                 ; invalid UTF-8
 740      ;; 4-byte sequence for an untranslated character.
 741      (write r6)
 742      (r5 = -1)
 743      (end))
 744
 745     ;; At EOF...
 746     ((r5 = -1)
 747      (if (r1 >= 0)
 748          (write r1)))))
 749
 750 (define-ccl-program ccl-encode-mule-utf-8
 751   `(1
 752     ((r5 = -1)
 753      (loop
 754       (if (r5 < 0)
 755           (read-multibyte-character r0 r1)
 756         ;; Pre-read character is in r5 (charset-ID) and r6 (code-point).
 757         ((r0 = r5)
 758          (r1 = r6)
 759          (r5 = -1)))
 760       (translate-character utf-translation-table-for-encode r0 r1)
 761
 762       (if (r0 == ,(charset-id 'ascii))
 763           (write-repeat r1))
 764
 765       (if (r0 == ,(charset-id 'latin-iso8859-1))
 766           ;; r1          scalar                  utf-8
 767           ;;       0000 0yyy yyxx xxxx    110y yyyy 10xx xxxx
 768           ;; 20    0000 0000 1010 0000    1100 0010 1010 0000
 769           ;; 7f    0000 0000 1111 1111    1100 0011 1011 1111
 770           ((write ((r1 >> 6) | #xc2))
 771            (r1 &= #x3f)
 772            (r1 |= #x80)
 773            (write-repeat r1)))
 774
 775       (if (r0 == ,(charset-id 'mule-unicode-0100-24ff))
 776           ((r0 = ((((r1 & #x3f80) >> 7) - 32) * 96))
 777            ;; #x3f80 == (0011 1111 1000 0000)b
 778            (r1 &= #x7f)
 779            (r1 += (r0 + 224))           ; 240 == -32 + #x0100
 780            ;; now r1 holds scalar value
 781            (if (r1 < #x0800)
 782                ;; 2byte encoding
 783                ((write ((r1 >> 6) | #xC0))
 784                 (r1 &= #x3F)
 785                 (r1 |= #x80)
 786                 (write-repeat r1))
 787              ;; 3byte encoding
 788              ((write ((r1 >> 12) | #xE0))
 789               (write  (((r1 & #x0FC0) >> 6) | #x80))
 790               (r1 &= #x3F)
 791               (r1 |= #x80)
 792               (write-repeat r1)))))
 793
 794       (if (r0 == ,(charset-id 'mule-unicode-2500-33ff))
 795           ((r0 = ((((r1 & #x3f80) >> 7) - 32) * 96))
 796            (r1 &= #x7f)
 797            (r1 += (r0 + 9440))          ; 9440 == -32 + #x2500
 798            ;; now r1 holds scalar value
 799            (write ((r1 >> 12) | #xE0))
 800            (write  (((r1 & #x0FC0) >> 6) | #x80))
 801            (r1 &= #x3F)
 802            (r1 |= #x80)
 803            (write-repeat r1)))
 804
 805       (if (r0 == ,(charset-id 'mule-unicode-e000-ffff))
 806           ((r0 = ((((r1 & #x3f80) >> 7) - 32) * 96))
 807            (r1 &= #x7f)
 808            (r1 += (r0 + 57312))         ; 57312 == -32 + #xe000
 809            ;; now r1 holds scalar value
 810            (write ((r1 >> 12) | #xE0))
 811            (write  (((r1 & #x0FC0) >> 6) | #x80))
 812            (r1 &= #x3F)
 813            (r1 |= #x80)
 814            (write-repeat r1)))
 815
 816       (if (r0 == ,(charset-id 'eight-bit-control))
 817           ;; r1          scalar                  utf-8
 818           ;;       0000 0yyy yyxx xxxx    110y yyyy 10xx xxxx
 819           ;; 80    0000 0000 1000 0000    1100 0010 1000 0000
 820           ;; 9f    0000 0000 1001 1111    1100 0010 1001 1111
 821           ((write #xC2)
 822            (write-repeat r1)))
 823
 824       (if (r0 == ,(charset-id 'eight-bit-graphic))
 825           ;; r1          scalar                  utf-8
 826           ;;       0000 0yyy yyxx xxxx    110y yyyy 10xx xxxx
 827           ;; a0    0000 0000 1010 0000    1100 0010 1010 0000
 828           ;; ff    0000 0000 1111 1111    1101 1111 1011 1111
 829           ((r0 = (r1 >= #xC0))
 830            (r0 &= (r1 <= #xC3))
 831            (r4 = (r1 >= #xE1))
 832            (r4 &= (r1 <= #xF7))
 833            (r0 |= r4)
 834            (if r0
 835                ((call ccl-mule-utf-8-encode-untrans)
 836                 (repeat))
 837              (write-repeat r1))))
 838
 839       (lookup-character utf-subst-table-for-encode r0 r1)
 840       (if r7            ; lookup succeeded
 841           (if (r0 < #x800)
 842               ;; 2byte encoding
 843               ((write ((r0 >> 6) | #xC0))
 844                (r0 = ((r0 & #x3F) | #x80))
 845                (write-repeat r0))
 846             ;; 3byte encoding
 847             ((write ((r0 >> 12) | #xE0))
 848              (write  (((r0 & #x0FC0) >> 6) | #x80))
 849              (r0 = ((r0 & #x3F) | #x80))
 850              (write-repeat r0))))
 851
 852       ;; Unsupported character.
 853       ;; Output U+FFFD, which is `ef bf bd' in UTF-8.
 854       (write #xef)
 855       (write #xbf)
 856       (write-repeat #xbd))))
 857   "CCL program to encode into UTF-8.")
 858
 859
 860 (define-ccl-program ccl-untranslated-to-ucs
 861   `(0
 862     (if (r1 == 0)
 863         nil
 864       (if (r0 <= #xC3)                  ; 2-byte encoding
 865           ((r0 = ((r0 & #x3) << 6))
 866            (r0 |= (r1 & #x3F))
 867            (r1 = 2))
 868         (if (r2 == 0)
 869             (r1 = 0)
 870           (if (r0 < #xF0)               ; 3-byte encoding, as above
 871               ((r0 = ((r0 & #xF) << 12))
 872                (r0 |= ((r1 & #x3F) << 6))
 873                (r0 |= (r2 & #x3F))
 874                (r1 = 3))
 875             (if (r3 == 0)
 876                 (r1 = 0)
 877               ((r0 = ((r0 & #x7) << 18))
 878                (r0 |= ((r1 & #x3F) << 12))
 879                (r0 |= ((r2 & #x3F) << 6))
 880                (r0 |= (r3 & #x3F))
 881                (r1 = 4))))))))
 882   "Decode 2-, 3-, or 4-byte sequences in r0, r1, r2 [,r3] to unicodes in r0.
 883 Set r1 to the byte length.  r0 == 0 for invalid sequence.")
 884
 885 (defvar utf-8-ccl-regs (make-vector 8 0))
 886
 887 (defsubst utf-8-untranslated-to-ucs ()
 888   "Return the UCS code for an untranslated sequence of raw bytes t point.
 889 Only for 3- or 4-byte sequences."
 890   (aset utf-8-ccl-regs 0 (or (char-after) 0))
 891   (aset utf-8-ccl-regs 1 (or (char-after (1+ (point))) 0))
 892   (aset utf-8-ccl-regs 2 (or (char-after (+ 2 (point))) 0))
 893   (aset utf-8-ccl-regs 3 (or (char-after (+ 3 (point))) 0))
 894   (ccl-execute 'ccl-untranslated-to-ucs utf-8-ccl-regs))
 895
 896 (defun utf-8-help-echo (window object position)
 897   (format "Untranslated Unicode U+%04X"
 898           (get-char-property position 'untranslated-utf-8 object)))
 899
 900 ;; We compose the untranslatable sequences into a single character,
 901 ;; and move point to the next character.
 902 ;; This is infelicitous for editing, because there's currently no
 903 ;; mechanism for treating compositions as atomic, but is OK for
 904 ;; display.  They are composed to U+FFFD with help-echo which
 905 ;; indicates the unicodes they represent.  This function GCs too much.
 906
 907 ;; If utf-translate-cjk-mode is non-nil, this function is called with
 908 ;; HASH-TABLE which translates CJK characters into some of CJK
 909 ;; charsets.
 910
 911 (defsubst utf-8-compose (hash-table)
 912   "Put a suitable composition on an untranslatable sequence at point.
 913 If HASH-TABLE is non-nil, try to translate CJK characters by it at first.
 914 Move point to the end of the sequence."
 915   (utf-8-untranslated-to-ucs)
 916   (let ((l (aref utf-8-ccl-regs 1))
 917         ch)
 918     (if (> l 0)
 919         (if (and hash-table
 920                  (setq ch (gethash (aref utf-8-ccl-regs 0)  hash-table)))
 921             (progn
 922               (insert ch)
 923               (delete-region (point) (min (point-max) (+ l (point)))))
 924           (setq ch (aref utf-8-ccl-regs 0))
 925           (put-text-property (point) (min (point-max) (+ l (point)))
 926                              'untranslated-utf-8 ch)
 927           (put-text-property (point) (min (point-max) (+ l (point)))
 928                              'help-echo 'utf-8-help-echo)
 929           (if (= l 2)
 930               (put-text-property (point) (min (point-max) (+ l (point)))
 931                                  'display (format "\\%03o" ch))
 932             (compose-region (point) (+ l (point)) ?\e$,3u=\e(B))
 933           (forward-char l))
 934       (forward-char 1))))
 935
 936 (defcustom utf-8-compose-scripts nil
 937   "*Non-nil means compose various scripts on decoding utf-8 text."
 938   :group 'mule
 939   :version "21.4"
 940   :type 'boolean)
 941
 942 (defun utf-8-post-read-conversion (length)
 943   "Compose untranslated utf-8 sequences into single characters.
 944 If `utf-translate-cjk-mode' is non-nil, tries to translate CJK characters.
 945 Also compose particular scripts if `utf-8-compose-scripts' is non-nil."
 946   (save-excursion
 947     (save-restriction
 948       (narrow-to-region (point) (+ (point) length))
 949       ;; Can't do eval-when-compile to insert a multibyte constant
 950       ;; version of the string in the loop, since it's always loaded as
 951       ;; unibyte from a byte-compiled file.
 952       (let ((range (string-as-multibyte "^\xc0-\xc3\xe1-\xf7"))
 953             (buffer-multibyte enable-multibyte-characters)
 954             hash-table ch)
 955         (set-buffer-multibyte t)
 956         (when utf-translate-cjk-mode
 957           (unless utf-translate-cjk-lang-env
 958             ;; Check these characters in utf-translate-cjk-range.
 959             ;; We may have to translate them to CJK charsets.
 960             (skip-chars-forward
 961              (concat range utf-translate-cjk-unicode-range-string))
 962             (unless (eobp)
 963               (utf-translate-cjk-load-tables)
 964               (setq range
 965                     (concat range utf-translate-cjk-unicode-range-string)))
 966             (setq hash-table (get 'utf-subst-table-for-decode
 967                                   'translation-hash-table))))
 968         (while (and (skip-chars-forward range)
 969                     (not (eobp)))
 970           (setq ch (following-char))
 971           (if (< ch 256)
 972               (utf-8-compose hash-table)
 973             (if (and hash-table
 974                      (setq ch (gethash (encode-char ch 'ucs) hash-table)))
 975                 (progn
 976                   (insert ch)
 977                   (delete-char 1))
 978               (forward-char 1))))
 979         (or buffer-multibyte
 980             (set-buffer-multibyte nil)))
 981
 982       (when (and utf-8-compose-scripts (> length 1))
 983         ;; These currently have definitions which cover the relevant
 984         ;; unicodes.  We could avoid loading thai-util &c by checking
 985         ;; whether the region contains any characters with the appropriate
 986         ;; categories.  There aren't yet Unicode-based rules for Tibetan.
 987         (diacritic-compose-region (point-max) (point-min))
 988         (thai-compose-region (point-max) (point-min))
 989         (lao-compose-region (point-max) (point-min))
 990         (devanagari-compose-region (point-max) (point-min))
 991         (malayalam-compose-region (point-max) (point-min))
 992         (tamil-compose-region (point-max) (point-min)))
 993       (- (point-max) (point-min)))))
 994
 995 (defun utf-8-pre-write-conversion (beg end)
 996   "Prepare for `utf-translate-cjk-mode' to encode text between BEG and END.
 997 This is used as a post-read-conversion of utf-8 coding system."
 998   (if (and utf-translate-cjk-mode
 999            (not utf-translate-cjk-lang-env)
1000            (save-excursion
1001              (goto-char beg)
1002              (re-search-forward "\\cc\\|\\cj\\|\\ch" end t)))
1003       (utf-translate-cjk-load-tables))
1004   nil)
1005
1006 (make-coding-system
1007  'mule-utf-8 4 ?u
1008  "UTF-8 encoding for Emacs-supported Unicode characters.
1009 It supports Unicode characters of these ranges:
1010     U+0000..U+33FF, U+E000..U+FFFF.
1011 They correspond to these Emacs character sets:
1012     ascii, latin-iso8859-1, mule-unicode-0100-24ff,
1013     mule-unicode-2500-33ff, mule-unicode-e000-ffff
1014
1015 On decoding (e.g. reading a file), Unicode characters not in the above
1016 ranges are decoded into sequences of eight-bit-control and
1017 eight-bit-graphic characters to preserve their byte sequences.  The
1018 byte sequence is preserved on i/o for valid utf-8, but not necessarily
1019 for invalid utf-8.
1020
1021 On encoding (e.g. writing a file), Emacs characters not belonging to
1022 any of the character sets listed above are encoded into the UTF-8 byte
1023 sequence representing U+FFFD (REPLACEMENT CHARACTER)."
1024
1025  '(ccl-decode-mule-utf-8 . ccl-encode-mule-utf-8)
1026  `((safe-charsets
1027     ascii
1028     eight-bit-control
1029     eight-bit-graphic
1030     latin-iso8859-1
1031     mule-unicode-0100-24ff
1032     mule-unicode-2500-33ff
1033     mule-unicode-e000-ffff
1034     ,@(if utf-translate-cjk-mode
1035           utf-translate-cjk-charsets))
1036    (mime-charset . utf-8)
1037    (coding-category . coding-category-utf-8)
1038    (valid-codes (0 . 255))
1039    (pre-write-conversion . utf-8-pre-write-conversion)
1040    (post-read-conversion . utf-8-post-read-conversion)
1041    (translation-table-for-encode . utf-translation-table-for-encode)
1042    (dependency unify-8859-on-encoding-mode
1043                unify-8859-on-decoding-mode
1044                utf-fragment-on-decoding
1045                utf-translate-cjk-mode)))
1046
1047 (define-coding-system-alias 'utf-8 'mule-utf-8)
1048
1049 ;; I think this needs special private charsets defined for the
1050 ;; untranslated sequences, if it's going to work well.
1051
1052 ;;; (defun utf-8-compose-function (pos to pattern &optional string)
1053 ;;;   (let* ((prop (get-char-property pos 'composition string))
1054 ;;;      (l (and prop (- (cadr prop) (car prop)))))
1055 ;;;     (cond ((and l (> l (- to pos)))
1056 ;;;        (delete-region pos to))
1057 ;;;       ((and (> (char-after pos) 224)
1058 ;;;             (< (char-after pos) 256)
1059 ;;;             (save-restriction
1060 ;;;               (narrow-to-region pos to)
1061 ;;;               (utf-8-compose)))
1062 ;;;        t))))
1063
1064 ;;; (dotimes (i 96)
1065 ;;;   (aset composition-function-table
1066 ;;;     (+ 128 i)
1067 ;;;     `((,(string-as-multibyte "[\200-\237\240-\377]")
1068 ;;;        . utf-8-compose-function))))
1069
1070 ;;; arch-tag: b08735b7-753b-4ae6-b754-0f3efe4515c5
1071 ;;; utf-8.el ends here