| 1 | /* |
| 2 | * misc.c: miscellaneous useful items |
| 3 | */ |
| 4 | |
| 5 | #include "halibut.h" |
| 6 | |
| 7 | struct stackTag { |
| 8 | void **data; |
| 9 | int sp; |
| 10 | int size; |
| 11 | }; |
| 12 | |
| 13 | stack stk_new(void) { |
| 14 | stack s; |
| 15 | |
| 16 | s = mknew(struct stackTag); |
| 17 | s->sp = 0; |
| 18 | s->size = 0; |
| 19 | s->data = NULL; |
| 20 | |
| 21 | return s; |
| 22 | } |
| 23 | |
| 24 | void stk_free(stack s) { |
| 25 | sfree(s->data); |
| 26 | sfree(s); |
| 27 | } |
| 28 | |
| 29 | void stk_push(stack s, void *item) { |
| 30 | if (s->size <= s->sp) { |
| 31 | s->size = s->sp + 32; |
| 32 | s->data = resize(s->data, s->size); |
| 33 | } |
| 34 | s->data[s->sp++] = item; |
| 35 | } |
| 36 | |
| 37 | void *stk_pop(stack s) { |
| 38 | if (s->sp > 0) |
| 39 | return s->data[--s->sp]; |
| 40 | else |
| 41 | return NULL; |
| 42 | } |
| 43 | |
| 44 | void *stk_top(stack s) { |
| 45 | if (s->sp > 0) |
| 46 | return s->data[s->sp-1]; |
| 47 | else |
| 48 | return NULL; |
| 49 | } |
| 50 | |
| 51 | /* |
| 52 | * Small routines to amalgamate a string from an input source. |
| 53 | */ |
| 54 | const rdstring empty_rdstring = {0, 0, NULL}; |
| 55 | const rdstringc empty_rdstringc = {0, 0, NULL}; |
| 56 | |
| 57 | void rdadd(rdstring *rs, wchar_t c) { |
| 58 | if (rs->pos >= rs->size-1) { |
| 59 | rs->size = rs->pos + 128; |
| 60 | rs->text = resize(rs->text, rs->size); |
| 61 | } |
| 62 | rs->text[rs->pos++] = c; |
| 63 | rs->text[rs->pos] = 0; |
| 64 | } |
| 65 | void rdadds(rdstring *rs, wchar_t *p) { |
| 66 | int len = ustrlen(p); |
| 67 | if (rs->pos >= rs->size - len) { |
| 68 | rs->size = rs->pos + len + 128; |
| 69 | rs->text = resize(rs->text, rs->size); |
| 70 | } |
| 71 | ustrcpy(rs->text + rs->pos, p); |
| 72 | rs->pos += len; |
| 73 | } |
| 74 | wchar_t *rdtrim(rdstring *rs) { |
| 75 | rs->text = resize(rs->text, rs->pos + 1); |
| 76 | return rs->text; |
| 77 | } |
| 78 | |
| 79 | void rdaddc(rdstringc *rs, char c) { |
| 80 | if (rs->pos >= rs->size-1) { |
| 81 | rs->size = rs->pos + 128; |
| 82 | rs->text = resize(rs->text, rs->size); |
| 83 | } |
| 84 | rs->text[rs->pos++] = c; |
| 85 | rs->text[rs->pos] = 0; |
| 86 | } |
| 87 | void rdaddsc(rdstringc *rs, char *p) { |
| 88 | int len = strlen(p); |
| 89 | if (rs->pos >= rs->size - len) { |
| 90 | rs->size = rs->pos + len + 128; |
| 91 | rs->text = resize(rs->text, rs->size); |
| 92 | } |
| 93 | strcpy(rs->text + rs->pos, p); |
| 94 | rs->pos += len; |
| 95 | } |
| 96 | char *rdtrimc(rdstringc *rs) { |
| 97 | rs->text = resize(rs->text, rs->pos + 1); |
| 98 | return rs->text; |
| 99 | } |
| 100 | |
| 101 | static int compare_wordlists_literally(word *a, word *b) { |
| 102 | int t; |
| 103 | while (a && b) { |
| 104 | if (a->type != b->type) |
| 105 | return (a->type < b->type ? -1 : +1); /* FIXME? */ |
| 106 | t = a->type; |
| 107 | if ((t != word_Normal && t != word_Code && |
| 108 | t != word_WeakCode && t != word_Emph) || |
| 109 | a->alt || b->alt) { |
| 110 | int c; |
| 111 | if (a->text && b->text) { |
| 112 | c = ustricmp(a->text, b->text); |
| 113 | if (c) |
| 114 | return c; |
| 115 | } |
| 116 | c = compare_wordlists_literally(a->alt, b->alt); |
| 117 | if (c) |
| 118 | return c; |
| 119 | a = a->next; |
| 120 | b = b->next; |
| 121 | } else { |
| 122 | wchar_t *ap = a->text, *bp = b->text; |
| 123 | while (*ap && *bp) { |
| 124 | wchar_t ac = utolower(*ap), bc = utolower(*bp); |
| 125 | if (ac != bc) |
| 126 | return (ac < bc ? -1 : +1); |
| 127 | if (!*++ap && a->next && a->next->type == t && !a->next->alt) |
| 128 | a = a->next, ap = a->text; |
| 129 | if (!*++bp && b->next && b->next->type == t && !b->next->alt) |
| 130 | b = b->next, bp = b->text; |
| 131 | } |
| 132 | if (*ap || *bp) |
| 133 | return (*ap ? +1 : -1); |
| 134 | a = a->next; |
| 135 | b = b->next; |
| 136 | } |
| 137 | } |
| 138 | |
| 139 | if (a || b) |
| 140 | return (a ? +1 : -1); |
| 141 | else |
| 142 | return 0; |
| 143 | } |
| 144 | |
| 145 | int compare_wordlists(word *a, word *b) { |
| 146 | /* |
| 147 | * First we compare only the alphabetic content of the word |
| 148 | * lists, with case not a factor. If that comes out equal, |
| 149 | * _then_ we compare the word lists literally. |
| 150 | */ |
| 151 | struct { |
| 152 | word *w; |
| 153 | int i; |
| 154 | wchar_t c; |
| 155 | } pos[2]; |
| 156 | |
| 157 | pos[0].w = a; |
| 158 | pos[1].w = b; |
| 159 | pos[0].i = pos[1].i = 0; |
| 160 | |
| 161 | while (1) { |
| 162 | /* |
| 163 | * Find the next alphabetic character in each word list. |
| 164 | */ |
| 165 | int k; |
| 166 | |
| 167 | for (k = 0; k < 2; k++) { |
| 168 | /* |
| 169 | * Advance until we hit either an alphabetic character |
| 170 | * or the end of the word list. |
| 171 | */ |
| 172 | while (1) { |
| 173 | if (!pos[k].w) { |
| 174 | /* End of word list. */ |
| 175 | pos[k].c = 0; |
| 176 | break; |
| 177 | } else if (!pos[k].w->text || !pos[k].w->text[pos[k].i]) { |
| 178 | /* No characters remaining in this word; move on. */ |
| 179 | pos[k].w = pos[k].w->next; |
| 180 | pos[k].i = 0; |
| 181 | } else if (!uisalpha(pos[k].w->text[pos[k].i])) { |
| 182 | /* This character isn't alphabetic; move on. */ |
| 183 | pos[k].i++; |
| 184 | } else { |
| 185 | /* We have an alphabetic! Lowercase it and continue. */ |
| 186 | pos[k].c = utolower(pos[k].w->text[pos[k].i]); |
| 187 | break; |
| 188 | } |
| 189 | } |
| 190 | } |
| 191 | |
| 192 | if (pos[0].c < pos[1].c) |
| 193 | return -1; |
| 194 | else if (pos[0].c > pos[1].c) |
| 195 | return +1; |
| 196 | |
| 197 | if (!pos[0].c) |
| 198 | break; /* they're equal */ |
| 199 | |
| 200 | pos[0].i++; |
| 201 | pos[1].i++; |
| 202 | } |
| 203 | |
| 204 | /* |
| 205 | * If we reach here, the strings were alphabetically equal, so |
| 206 | * compare in more detail. |
| 207 | */ |
| 208 | return compare_wordlists_literally(a, b); |
| 209 | } |
| 210 | |
| 211 | void mark_attr_ends(paragraph *sourceform) { |
| 212 | paragraph *p; |
| 213 | word *w, *wp; |
| 214 | for (p = sourceform; p; p = p->next) { |
| 215 | wp = NULL; |
| 216 | for (w = p->words; w; w = w->next) { |
| 217 | if (isattr(w->type)) { |
| 218 | int before = (wp && isattr(wp->type) && |
| 219 | sameattr(wp->type, w->type)); |
| 220 | int after = (w->next && isattr(w->next->type) && |
| 221 | sameattr(w->next->type, w->type)); |
| 222 | w->aux |= (before ? |
| 223 | (after ? attr_Always : attr_Last) : |
| 224 | (after ? attr_First : attr_Only)); |
| 225 | } |
| 226 | wp = w; |
| 227 | } |
| 228 | } |
| 229 | } |
| 230 | |
| 231 | wrappedline *wrap_para(word *text, int width, int subsequentwidth, |
| 232 | int (*widthfn)(word *)) { |
| 233 | wrappedline *head = NULL, **ptr = &head; |
| 234 | int nwords, wordsize; |
| 235 | struct wrapword { |
| 236 | word *begin, *end; |
| 237 | int width; |
| 238 | int spacewidth; |
| 239 | int cost; |
| 240 | int nwords; |
| 241 | } *wrapwords; |
| 242 | int i, j, n; |
| 243 | |
| 244 | /* |
| 245 | * Break the line up into wrappable components. |
| 246 | */ |
| 247 | nwords = wordsize = 0; |
| 248 | wrapwords = NULL; |
| 249 | while (text) { |
| 250 | if (nwords >= wordsize) { |
| 251 | wordsize = nwords + 64; |
| 252 | wrapwords = srealloc(wrapwords, wordsize * sizeof(*wrapwords)); |
| 253 | } |
| 254 | wrapwords[nwords].width = 0; |
| 255 | wrapwords[nwords].begin = text; |
| 256 | while (text) { |
| 257 | wrapwords[nwords].width += widthfn(text); |
| 258 | wrapwords[nwords].end = text->next; |
| 259 | if (text->next && (text->next->type == word_WhiteSpace || |
| 260 | text->next->type == word_EmphSpace || |
| 261 | text->breaks)) |
| 262 | break; |
| 263 | text = text->next; |
| 264 | } |
| 265 | if (text && text->next && (text->next->type == word_WhiteSpace || |
| 266 | text->next->type == word_EmphSpace)) { |
| 267 | wrapwords[nwords].spacewidth = widthfn(text->next); |
| 268 | text = text->next; |
| 269 | } else { |
| 270 | wrapwords[nwords].spacewidth = 0; |
| 271 | } |
| 272 | nwords++; |
| 273 | if (text) |
| 274 | text = text->next; |
| 275 | } |
| 276 | |
| 277 | /* |
| 278 | * Perform the dynamic wrapping algorithm: work backwards from |
| 279 | * nwords-1, determining the optimal wrapping for each terminal |
| 280 | * subsequence of the paragraph. |
| 281 | */ |
| 282 | for (i = nwords; i-- ;) { |
| 283 | int best = -1; |
| 284 | int bestcost = 0; |
| 285 | int cost; |
| 286 | int linelen = 0, spacewidth = 0; |
| 287 | int seenspace; |
| 288 | int thiswidth = (i == 0 ? width : subsequentwidth); |
| 289 | |
| 290 | j = 0; |
| 291 | seenspace = 0; |
| 292 | while (i+j < nwords) { |
| 293 | /* |
| 294 | * See what happens if we put j+1 words on this line. |
| 295 | */ |
| 296 | if (spacewidth) |
| 297 | seenspace = 1; |
| 298 | linelen += spacewidth + wrapwords[i+j].width; |
| 299 | spacewidth = wrapwords[i+j].spacewidth; |
| 300 | j++; |
| 301 | if (linelen > thiswidth) { |
| 302 | /* |
| 303 | * If we're over the width limit, abandon ship, |
| 304 | * _unless_ there is no best-effort yet (which will |
| 305 | * only happen if the first word is too long all by |
| 306 | * itself). |
| 307 | */ |
| 308 | if (best > 0) |
| 309 | break; |
| 310 | } |
| 311 | if (i+j == nwords) { |
| 312 | /* |
| 313 | * Special case: if we're at the very end of the |
| 314 | * paragraph, we don't score penalty points for the |
| 315 | * white space left on the line. |
| 316 | */ |
| 317 | cost = 0; |
| 318 | } else { |
| 319 | cost = (thiswidth-linelen) * (thiswidth-linelen); |
| 320 | cost += wrapwords[i+j].cost; |
| 321 | } |
| 322 | /* |
| 323 | * We compare bestcost >= cost, not bestcost > cost, |
| 324 | * because in cases where the costs are identical we |
| 325 | * want to try to look like the greedy algorithm, |
| 326 | * because readers are likely to have spent a lot of |
| 327 | * time looking at greedy-wrapped paragraphs and |
| 328 | * there's no point violating the Principle of Least |
| 329 | * Surprise if it doesn't actually gain anything. |
| 330 | */ |
| 331 | if (best < 0 || bestcost >= cost) { |
| 332 | bestcost = cost; |
| 333 | best = j; |
| 334 | } |
| 335 | } |
| 336 | /* |
| 337 | * Now we know the optimal answer for this terminal |
| 338 | * subsequence, so put it in wrapwords. |
| 339 | */ |
| 340 | wrapwords[i].cost = bestcost; |
| 341 | wrapwords[i].nwords = best; |
| 342 | } |
| 343 | |
| 344 | /* |
| 345 | * We've wrapped the paragraph. Now build the output |
| 346 | * `wrappedline' list. |
| 347 | */ |
| 348 | i = 0; |
| 349 | while (i < nwords) { |
| 350 | wrappedline *w = mknew(wrappedline); |
| 351 | *ptr = w; |
| 352 | ptr = &w->next; |
| 353 | w->next = NULL; |
| 354 | |
| 355 | n = wrapwords[i].nwords; |
| 356 | w->begin = wrapwords[i].begin; |
| 357 | w->end = wrapwords[i+n-1].end; |
| 358 | |
| 359 | /* |
| 360 | * Count along the words to find nspaces and shortfall. |
| 361 | */ |
| 362 | w->nspaces = 0; |
| 363 | w->shortfall = width; |
| 364 | for (j = 0; j < n; j++) { |
| 365 | w->shortfall -= wrapwords[i+j].width; |
| 366 | if (j < n-1 && wrapwords[i+j].spacewidth) { |
| 367 | w->nspaces++; |
| 368 | w->shortfall -= wrapwords[i+j].spacewidth; |
| 369 | } |
| 370 | } |
| 371 | i += n; |
| 372 | } |
| 373 | |
| 374 | sfree(wrapwords); |
| 375 | |
| 376 | return head; |
| 377 | } |
| 378 | |
| 379 | void wrap_free(wrappedline *w) { |
| 380 | while (w) { |
| 381 | wrappedline *t = w->next; |
| 382 | sfree(w); |
| 383 | w = t; |
| 384 | } |
| 385 | } |