chartrans.c

   1 /*
   2  * chartrans.c
   3  * Copyright (C) 1999,2000 A.J. van Os
   4  *
   5  * Description:
   6  * Translate characters to ISO-8859-1
   7  */
   8
   9 #include <ctype.h>
  10 #include <limits.h>
  11 #include "antiword.h"
  12
  13 /*
  14  * iTranslateCharacters - Translate characters to ISO-8859-1
  15  *
  16  * Translate Macintosh characters, Window characters and some Unicode
  17  * characters into ISO-8859-1 (aka Latin1).
  18  *
  19  * returns the translated character or EOF if the input character should
  20  *      be ignored
  21  */
  22 int
  23 iTranslateCharacters(int iChar, int iFileOffset, BOOL bMacFile)
  24 {
  25         if (bMacFile) {
  26                 /* Translate special Macintosh characters */
  27                 switch (iChar) {
  28                 case MAC_CAPTAL_A_DIAERESIS:
  29                         return OUR_CAPTAL_A_DIAERESIS;
  30                 case MAC_CAPTAL_A_RING_ABOVE:
  31                         return OUR_CAPTAL_A_RING_ABOVE;
  32                 case MAC_CAPTAL_C_CEDILLA:
  33                         return OUR_CAPTAL_C_CEDILLA;
  34                 case MAC_CAPTAL_O_ACUTE_ACCENT:
  35                         return OUR_CAPTAL_O_ACUTE_ACCENT;
  36                 case MAC_CAPTAL_N_TILDE:
  37                         return OUR_CAPTAL_N_TILDE;
  38                 case MAC_CAPTAL_O_DIAERESIS:
  39                         return OUR_CAPTAL_O_DIAERESIS;
  40                 case MAC_CAPTAL_U_DIAERESIS:
  41                         return OUR_CAPTAL_U_DIAERESIS;
  42                 case MAC_SMALL_A_ACUTE_ACCENT:
  43                         return OUR_SMALL_A_ACUTE_ACCENT;
  44                 case MAC_SMALL_A_GRAVE_ACCENT:
  45                         return OUR_SMALL_A_GRAVE_ACCENT;
  46                 case MAC_SMALL_A_CIRCUMFLEX_ACCENT:
  47                         return OUR_SMALL_A_CIRCUMFLEX_ACCENT;
  48                 case MAC_SMALL_A_DIAERESIS:
  49                         return OUR_SMALL_A_DIAERESIS;
  50                 case MAC_SMALL_A_TILDE:
  51                         return OUR_SMALL_A_TILDE;
  52                 case MAC_SMALL_A_RING_ABOVE:
  53                         return OUR_SMALL_A_RING_ABOVE;
  54                 case MAC_SMALL_C_CEDILLA:
  55                         return OUR_SMALL_C_CEDILLA;
  56                 case MAC_SMALL_E_ACUTE_ACCENT:
  57                         return OUR_SMALL_E_ACUTE_ACCENT;
  58                 case MAC_SMALL_E_GRAVE_ACCENT:
  59                         return OUR_SMALL_E_GRAVE_ACCENT;
  60                 case MAC_SMALL_E_CIRCUMFLEX_ACCENT:
  61                         return OUR_SMALL_E_CIRCUMFLEX_ACCENT;
  62                 case MAC_SMALL_E_DIAERESIS:
  63                         return OUR_SMALL_E_DIAERESIS;
  64                 case MAC_SMALL_I_ACUTE_ACCENT:
  65                         return OUR_SMALL_I_ACUTE_ACCENT;
  66                 case MAC_SMALL_I_GRAVE_ACCENT:
  67                         return OUR_SMALL_I_GRAVE_ACCENT;
  68                 case MAC_SMALL_I_CIRCUMFLEX_ACCENT:
  69                         return OUR_SMALL_I_CIRCUMFLEX_ACCENT;
  70                 case MAC_SMALL_I_DIAERESIS:
  71                         return OUR_SMALL_I_DIAERESIS;
  72                 case MAC_SMALL_N_TILDE:
  73                         return OUR_SMALL_N_TILDE;
  74                 case MAC_SMALL_O_ACUTE_ACCENT:
  75                         return OUR_SMALL_O_ACUTE_ACCENT;
  76                 case MAC_SMALL_O_GRAVE_ACCENT:
  77                         return OUR_SMALL_O_GRAVE_ACCENT;
  78                 case MAC_SMALL_O_CIRCUMFLEX_ACCENT:
  79                         return OUR_SMALL_O_CIRCUMFLEX_ACCENT;
  80                 case MAC_SMALL_O_DIAERESIS:
  81                         return OUR_SMALL_O_DIAERESIS;
  82                 case MAC_SMALL_O_TILDE:
  83                         return OUR_SMALL_O_TILDE;
  84                 case MAC_SMALL_U_ACUTE_ACCENT:
  85                         return OUR_SMALL_U_ACUTE_ACCENT;
  86                 case MAC_SMALL_U_GRAVE_ACCENT:
  87                         return OUR_SMALL_U_GRAVE_ACCENT;
  88                 case MAC_SMALL_U_CIRCUMFLEX_ACCENT:
  89                         return OUR_SMALL_U_CIRCUMFLEX_ACCENT;
  90                 case MAC_SMALL_U_DIAERESIS:
  91                         return OUR_SMALL_U_DIAERESIS;
  92                 case MAC_SMALL_SHARP_S:
  93                         return OUR_SMALL_SHARP_S;
  94                 case MAC_LEFT_DOUBLE_QMARK:
  95                         return OUR_LEFT_DOUBLE_QMARK;
  96                 case MAC_RIGHT_DOUBLE_QMARK:
  97                         return OUR_RIGHT_DOUBLE_QMARK;
  98                 case MAC_EN_DASH:
  99                         return OUR_EN_DASH;
 100                 case MAC_EM_DASH:
 101                         return OUR_EM_DASH;
 102                 case MAC_OPENING_DOUBLE_QUOTE:
 103                         return OUR_OPENING_DOUBLE_QUOTE;
 104                 case MAC_CLOSING_DOUBLE_QUOTE:
 105                         return OUR_CLOSING_DOUBLE_QUOTE;
 106                 case MAC_LEFT_SINGLE_QUOTE:
 107                         return OUR_LEFT_SINGLE_QUOTE;
 108                 case MAC_RIGHT_SINGLE_QUOTE:
 109                         return OUR_RIGHT_SINGLE_QUOTE;
 110                 default:
 111                         break;
 112                 }
 113         }
 114
 115         /* Translate characters to ISO-8859-1 */
 116         switch (iChar) {
 117         case IGNORE_CHAR:
 118         case ANNOTATION:
 119         case FRAME:
 120         case WORD_SOFT_HYPHEN:
 121         case UNICODE_HYPHENATION_POINT:
 122                 return EOF;
 123         case PICTURE:
 124         case TABLE_SEPARATOR:
 125         case TAB:
 126         case HARD_RETURN:
 127         case FORM_FEED:
 128         case PAR_END:
 129         case COLUMN_FEED:
 130                 return iChar;
 131         case FOOTNOTE_OR_ENDNOTE:
 132                 NO_DBG_HEX(iFileOffset);
 133                 switch (eGetNotetype(iFileOffset)) {
 134                 case notetype_is_footnote:
 135                         return FOOTNOTE_CHAR;
 136                 case notetype_is_endnote:
 137                         return ENDNOTE_CHAR;
 138                 default:
 139                         return UNKNOWN_NOTE_CHAR;
 140                 }
 141         case WORD_UNBREAKABLE_JOIN:
 142         case UNICODE_NON_BREAKING_HYPHEN:
 143                 return OUR_UNBREAKABLE_JOIN;
 144         case WORD_EURO_SIGN:
 145         case UNICODE_EURO_SIGN:
 146                 return OUR_EURO_SIGN;
 147         case WORD_CEDILLA:
 148                 return OUR_CEDILLA;
 149         case WORD_DUTCH_GUILDER_SIGN:
 150                 return OUR_DUTCH_GUILDER_SIGN;
 151         case WORD_LOW_DOUBLE_QUOTE:
 152         case UNICODE_LOW_DOUBLE_QUOTE:
 153                 return OUR_LOW_DOUBLE_QUOTE;
 154         case WORD_ELLIPSIS:
 155         case UNICODE_ELLIPSIS:
 156                 return OUR_ELLIPSIS;
 157         case WORD_DAGGER:
 158         case UNICODE_DAGGER:
 159                 return OUR_DAGGER;
 160         case WORD_DOUBLE_DAGGER:
 161         case UNICODE_DOUBLE_DAGGER:
 162                 return OUR_DOUBLE_DAGGER;
 163         case WORD_NON_SPACING_CIRCUMFLEX_ACCENT:
 164                 return OUR_NON_SPACING_CIRCUMFLEX_ACCENT;
 165         case WORD_PER_MILLE_SIGN:
 166         case UNICODE_PER_MILLE_SIGN:
 167                 return OUR_PER_MILLE_SIGN;
 168         case WORD_LEFT_SINGLE_QMARK:
 169         case UNICODE_LEFT_SINGLE_QMARK:
 170                 return OUR_LEFT_SINGLE_QMARK;
 171         case WORD_CAPITAL_LIGATURE_OE:
 172                 return OUR_CAPITAL_LIGATURE_OE;
 173         case WORD_LEFT_SINGLE_QUOTE:
 174         case UNICODE_LEFT_SINGLE_QUOTE:
 175         case UNICODE_LEFT_SINGLE_QUOTE_ALT:
 176                 return OUR_LEFT_SINGLE_QUOTE;
 177         case WORD_RIGHT_SINGLE_QUOTE:
 178         case UNICODE_RIGHT_SINGLE_QUOTE:
 179         case UNICODE_RIGHT_SINGLE_QUOTE_ALT:
 180                 return OUR_RIGHT_SINGLE_QUOTE;
 181         case WORD_OPENING_DOUBLE_QUOTE:
 182         case UNICODE_OPENING_DOUBLE_QUOTE:
 183                 return OUR_OPENING_DOUBLE_QUOTE;
 184         case WORD_CLOSING_DOUBLE_QUOTE:
 185         case UNICODE_CLOSING_DOUBLE_QUOTE:
 186                 return OUR_CLOSING_DOUBLE_QUOTE;
 187         case WORD_BULLET:
 188         case UNICODE_BULLET:
 189         case UNICODE_BULLET_ALT:
 190                 return OUR_BULLET;
 191         case WORD_EM_DASH:
 192         case UNICODE_EM_DASH:
 193         case UNICODE_HORIZONTAL_BAR:
 194                 return OUR_EM_DASH;
 195         case WORD_EN_DASH:
 196         case UNICODE_EN_DASH:
 197         case UNICODE_FIGURE_DASH:
 198                 return OUR_EN_DASH;
 199         case WORD_NON_SPACING_TILDE:
 200                 return OUR_NON_SPACING_TILDE;
 201         case WORD_TRADEMARK:
 202         case UNICODE_TRADEMARK:
 203                 return OUR_TRADEMARK;
 204         case WORD_RIGHT_SINGLE_QMARK:
 205         case UNICODE_RIGHT_SINGLE_QMARK:
 206                 return OUR_RIGHT_SINGLE_QMARK;
 207         case WORD_SMALL_LIGATURE_OE:
 208                 return OUR_SMALL_LIGATURE_OE;
 209         case UNICODE_CAPITAL_W_CIRCUMFLEX_ACCENT:
 210                 return OUR_CAPITAL_W_CIRCUMFLEX_ACCENT;
 211         case UNICODE_SMALL_W_CIRCUMFLEX_ACCENT:
 212                 return OUR_SMALL_W_CIRCUMFLEX_ACCENT;
 213         case UNICODE_CAPITAL_Y_CIRCUMFLEX_ACCENT:
 214                 return OUR_CAPITAL_Y_CIRCUMFLEX_ACCENT;
 215         case UNICODE_SMALL_Y_CIRCUMFLEX_ACCENT:
 216                 return OUR_SMALL_Y_CIRCUMFLEX_ACCENT;
 217         case UNICODE_HYPHEN:
 218                 return OUR_HYPHEN;
 219         case UNICODE_DOUBLE_VERTICAL_LINE:
 220                 return OUR_DOUBLE_VERTICAL_LINE;
 221         case UNICODE_DOUBLE_LOW_LINE:
 222                 return OUR_DOUBLE_LOW_LINE;
 223         case UNICODE_FRACTION_SLASH:
 224                 return OUR_FRACTION_SLASH;
 225         case UNICODE_WHITE_SMILING_FACE:
 226                 return OUR_WHITE_SMILING_FACE;
 227         case UNICODE_BLACK_SMILING_FACE:
 228                 return OUR_BLACK_SMILING_FACE;
 229         case UNICODE_DIAMOND:
 230                 return OUR_DIAMOND;
 231         case UNICODE_COPYRIGHT:
 232                 return OUR_COPYRIGHT;
 233         default:
 234                 DBG_HEX_C(iChar >= 0x80 && iChar <= 0x9f, iFileOffset);
 235                 DBG_HEX_C(iChar >= 0x80 && iChar <= 0x9f, iChar);
 236                 DBG_HEX_C(iChar < 0x20 || iChar > 0xff, iFileOffset);
 237                 DBG_HEX_C(iChar < 0x20 || iChar > 0xff, iChar);
 238                 if (iChar < 0x20) {
 239                         /* A control character slipped through */
 240                         return EOF;
 241                 }
 242                 if (iChar > 0xff) {
 243                         /* Untranslated Unicode character */
 244                         return '?';
 245                 }
 246                 return iChar;
 247         }
 248 } /* end of iTranslateCharacters */
 249
 250 /*
 251  * iToUpper -  convert letter to upper case
 252  *
 253  * This function converts a letter to upper case. Unlike toupper(3) this
 254  * function is independent from the settings of locale. This comes in handy
 255  * for people who have to read Word documents in more than one language or
 256  * contain more than one language.
 257  *
 258  * returns the converted letter, or iChar if the conversion was not possible.
 259  */
 260 int iToUpper(int iChar)
 261 {
 262         if ((iChar & ~0x7f) == 0) {
 263                 /* US ASCII: use standard function */
 264                 return toupper(iChar);
 265         }
 266         if (iChar >= 0xe0 && iChar <= 0xfe && iChar != 0xf7) {
 267                 /*
 268                  * Lower case accented characters
 269                  * 0xf7 is Division sign; 0xd7 is Multiplication sign
 270                  * 0xff is y with diaeresis; 0xdf is Sharp s
 271                  */
 272                 return iChar & ~0x20;
 273         }
 274         return iChar;
 275 } /* end of iToUpper */