d7482997 |
1 | /* |
2 | * misc.c: miscellaneous useful items |
3 | */ |
4 | |
e4ea58f8 |
5 | #include <stdarg.h> |
d7482997 |
6 | #include "halibut.h" |
7 | |
e4ea58f8 |
8 | char *adv(char *s) { |
9 | return s + 1 + strlen(s); |
10 | } |
11 | |
d7482997 |
12 | struct stackTag { |
13 | void **data; |
14 | int sp; |
15 | int size; |
16 | }; |
17 | |
18 | stack stk_new(void) { |
19 | stack s; |
20 | |
f1530049 |
21 | s = snew(struct stackTag); |
d7482997 |
22 | s->sp = 0; |
23 | s->size = 0; |
24 | s->data = NULL; |
25 | |
26 | return s; |
27 | } |
28 | |
29 | void stk_free(stack s) { |
30 | sfree(s->data); |
31 | sfree(s); |
32 | } |
33 | |
34 | void stk_push(stack s, void *item) { |
35 | if (s->size <= s->sp) { |
36 | s->size = s->sp + 32; |
f1530049 |
37 | s->data = sresize(s->data, s->size, void *); |
d7482997 |
38 | } |
39 | s->data[s->sp++] = item; |
40 | } |
41 | |
42 | void *stk_pop(stack s) { |
43 | if (s->sp > 0) |
44 | return s->data[--s->sp]; |
45 | else |
46 | return NULL; |
47 | } |
48 | |
7136a6c7 |
49 | void *stk_top(stack s) { |
50 | if (s->sp > 0) |
51 | return s->data[s->sp-1]; |
52 | else |
53 | return NULL; |
54 | } |
55 | |
d7482997 |
56 | /* |
57 | * Small routines to amalgamate a string from an input source. |
58 | */ |
59 | const rdstring empty_rdstring = {0, 0, NULL}; |
60 | const rdstringc empty_rdstringc = {0, 0, NULL}; |
61 | |
62 | void rdadd(rdstring *rs, wchar_t c) { |
63 | if (rs->pos >= rs->size-1) { |
64 | rs->size = rs->pos + 128; |
f1530049 |
65 | rs->text = sresize(rs->text, rs->size, wchar_t); |
d7482997 |
66 | } |
67 | rs->text[rs->pos++] = c; |
68 | rs->text[rs->pos] = 0; |
69 | } |
5dd44dce |
70 | void rdadds(rdstring *rs, wchar_t const *p) { |
d7482997 |
71 | int len = ustrlen(p); |
72 | if (rs->pos >= rs->size - len) { |
73 | rs->size = rs->pos + len + 128; |
f1530049 |
74 | rs->text = sresize(rs->text, rs->size, wchar_t); |
d7482997 |
75 | } |
76 | ustrcpy(rs->text + rs->pos, p); |
77 | rs->pos += len; |
78 | } |
79 | wchar_t *rdtrim(rdstring *rs) { |
f1530049 |
80 | rs->text = sresize(rs->text, rs->pos + 1, wchar_t); |
d7482997 |
81 | return rs->text; |
82 | } |
83 | |
84 | void rdaddc(rdstringc *rs, char c) { |
85 | if (rs->pos >= rs->size-1) { |
86 | rs->size = rs->pos + 128; |
f1530049 |
87 | rs->text = sresize(rs->text, rs->size, char); |
d7482997 |
88 | } |
89 | rs->text[rs->pos++] = c; |
90 | rs->text[rs->pos] = 0; |
91 | } |
5dd44dce |
92 | void rdaddsc(rdstringc *rs, char const *p) { |
7e2417cc |
93 | rdaddsn(rs, p, strlen(p)); |
94 | } |
95 | void rdaddsn(rdstringc *rs, char const *p, int len) { |
d7482997 |
96 | if (rs->pos >= rs->size - len) { |
97 | rs->size = rs->pos + len + 128; |
f1530049 |
98 | rs->text = sresize(rs->text, rs->size, char); |
d7482997 |
99 | } |
7e2417cc |
100 | memcpy(rs->text + rs->pos, p, len); |
d7482997 |
101 | rs->pos += len; |
7e2417cc |
102 | rs->text[rs->pos] = 0; |
d7482997 |
103 | } |
104 | char *rdtrimc(rdstringc *rs) { |
f1530049 |
105 | rs->text = sresize(rs->text, rs->pos + 1, char); |
d7482997 |
106 | return rs->text; |
107 | } |
108 | |
831da32e |
109 | static int compare_wordlists_literally(word *a, word *b) { |
d7482997 |
110 | int t; |
111 | while (a && b) { |
112 | if (a->type != b->type) |
113 | return (a->type < b->type ? -1 : +1); /* FIXME? */ |
114 | t = a->type; |
115 | if ((t != word_Normal && t != word_Code && |
116 | t != word_WeakCode && t != word_Emph) || |
117 | a->alt || b->alt) { |
118 | int c; |
119 | if (a->text && b->text) { |
120 | c = ustricmp(a->text, b->text); |
121 | if (c) |
122 | return c; |
123 | } |
831da32e |
124 | c = compare_wordlists_literally(a->alt, b->alt); |
d7482997 |
125 | if (c) |
126 | return c; |
127 | a = a->next; |
128 | b = b->next; |
129 | } else { |
130 | wchar_t *ap = a->text, *bp = b->text; |
131 | while (*ap && *bp) { |
da090173 |
132 | wchar_t ac = *ap, bc = *bp; |
d7482997 |
133 | if (ac != bc) |
134 | return (ac < bc ? -1 : +1); |
135 | if (!*++ap && a->next && a->next->type == t && !a->next->alt) |
136 | a = a->next, ap = a->text; |
137 | if (!*++bp && b->next && b->next->type == t && !b->next->alt) |
138 | b = b->next, bp = b->text; |
139 | } |
140 | if (*ap || *bp) |
141 | return (*ap ? +1 : -1); |
142 | a = a->next; |
143 | b = b->next; |
144 | } |
145 | } |
146 | |
147 | if (a || b) |
148 | return (a ? +1 : -1); |
149 | else |
150 | return 0; |
151 | } |
152 | |
831da32e |
153 | int compare_wordlists(word *a, word *b) { |
154 | /* |
155 | * First we compare only the alphabetic content of the word |
156 | * lists, with case not a factor. If that comes out equal, |
157 | * _then_ we compare the word lists literally. |
158 | */ |
159 | struct { |
160 | word *w; |
161 | int i; |
162 | wchar_t c; |
163 | } pos[2]; |
164 | |
165 | pos[0].w = a; |
166 | pos[1].w = b; |
167 | pos[0].i = pos[1].i = 0; |
168 | |
169 | while (1) { |
170 | /* |
171 | * Find the next alphabetic character in each word list. |
172 | */ |
173 | int k; |
174 | |
175 | for (k = 0; k < 2; k++) { |
176 | /* |
177 | * Advance until we hit either an alphabetic character |
178 | * or the end of the word list. |
179 | */ |
180 | while (1) { |
181 | if (!pos[k].w) { |
182 | /* End of word list. */ |
183 | pos[k].c = 0; |
184 | break; |
185 | } else if (!pos[k].w->text || !pos[k].w->text[pos[k].i]) { |
186 | /* No characters remaining in this word; move on. */ |
187 | pos[k].w = pos[k].w->next; |
188 | pos[k].i = 0; |
189 | } else if (!uisalpha(pos[k].w->text[pos[k].i])) { |
190 | /* This character isn't alphabetic; move on. */ |
191 | pos[k].i++; |
192 | } else { |
193 | /* We have an alphabetic! Lowercase it and continue. */ |
194 | pos[k].c = utolower(pos[k].w->text[pos[k].i]); |
195 | break; |
196 | } |
197 | } |
198 | } |
199 | |
68cff1f8 |
200 | #ifdef HAS_WCSCOLL |
201 | { |
202 | wchar_t a[2], b[2]; |
203 | int ret; |
204 | |
205 | a[0] = pos[0].c; |
206 | b[0] = pos[1].c; |
207 | a[1] = b[1] = L'\0'; |
208 | |
209 | ret = wcscoll(a, b); |
210 | if (ret) |
211 | return ret; |
212 | } |
213 | #else |
831da32e |
214 | if (pos[0].c < pos[1].c) |
215 | return -1; |
216 | else if (pos[0].c > pos[1].c) |
217 | return +1; |
68cff1f8 |
218 | #endif |
831da32e |
219 | |
220 | if (!pos[0].c) |
221 | break; /* they're equal */ |
222 | |
223 | pos[0].i++; |
224 | pos[1].i++; |
225 | } |
226 | |
227 | /* |
228 | * If we reach here, the strings were alphabetically equal, so |
229 | * compare in more detail. |
230 | */ |
231 | return compare_wordlists_literally(a, b); |
232 | } |
233 | |
bb9e7835 |
234 | void mark_attr_ends(word *words) |
235 | { |
d7482997 |
236 | word *w, *wp; |
bb9e7835 |
237 | |
238 | wp = NULL; |
239 | for (w = words; w; w = w->next) { |
b9e27ab6 |
240 | int both; |
241 | if (!isvis(w->type)) |
242 | /* Invisible elements should not affect this calculation */ |
243 | continue; |
244 | both = (isattr(w->type) && |
245 | wp && isattr(wp->type) && |
246 | sameattr(wp->type, w->type)); |
247 | w->aux |= both ? attr_Always : attr_First; |
248 | if (wp && !both) { |
249 | /* If previous considered word turns out to have been |
250 | * the end of a run, tidy it up. */ |
251 | int wp_attr = attraux(wp->aux); |
252 | wp->aux = (wp->aux & ~attr_mask) | |
253 | ((wp_attr == attr_Always) ? attr_Last |
254 | /* attr_First */ : attr_Only); |
d7482997 |
255 | } |
bb9e7835 |
256 | wp = w; |
d7482997 |
257 | } |
b9e27ab6 |
258 | |
259 | /* Tidy up last word touched */ |
260 | if (wp) { |
261 | int wp_attr = attraux(wp->aux); |
262 | wp->aux = (wp->aux & ~attr_mask) | |
263 | ((wp_attr == attr_Always) ? attr_Last |
264 | /* attr_First */ : attr_Only); |
265 | } |
d7482997 |
266 | } |
267 | |
43341922 |
268 | /* |
269 | * This function implements the optimal paragraph wrapping |
270 | * algorithm, pretty much as used in TeX. A cost function is |
271 | * defined for each line of the wrapped paragraph (typically some |
272 | * convex function of the difference between the line's length and |
273 | * its desired length), and a dynamic programming approach is used |
274 | * to optimise globally across all possible layouts of the |
275 | * paragraph to find the one with the minimum total cost. |
276 | * |
277 | * The function as implemented here gives a choice of two options |
278 | * for the cost function: |
279 | * |
280 | * - If `natural_space' is zero, then the algorithm attempts to |
281 | * make each line the maximum possible width (either `width' or |
282 | * `subsequentwidth' depending on whether it's the first line of |
283 | * the paragraph or not), and the cost function is simply the |
284 | * square of the unused space at the end of each line. This is a |
285 | * simple mechanism suitable for use in fixed-pitch environments |
286 | * such as plain text displayed on a terminal. |
287 | * |
288 | * - However, if `natural_space' is positive, the algorithm |
289 | * assumes the medium is fully graphical and that the width of |
290 | * space characters can be adjusted finely, and it attempts to |
291 | * make each _space character_ the width given in |
292 | * `natural_space'. (The provided width function should return |
293 | * the _minimum_ acceptable width of a space character in this |
294 | * case.) Therefore, the cost function for a line is dependent |
295 | * on the number of spaces on that line as well as the amount by |
296 | * which the line width differs from the optimum. |
297 | */ |
d7482997 |
298 | wrappedline *wrap_para(word *text, int width, int subsequentwidth, |
43341922 |
299 | int (*widthfn)(void *, word *), void *ctx, |
300 | int natural_space) { |
d7482997 |
301 | wrappedline *head = NULL, **ptr = &head; |
302 | int nwords, wordsize; |
303 | struct wrapword { |
304 | word *begin, *end; |
305 | int width; |
306 | int spacewidth; |
307 | int cost; |
308 | int nwords; |
309 | } *wrapwords; |
310 | int i, j, n; |
311 | |
312 | /* |
313 | * Break the line up into wrappable components. |
314 | */ |
315 | nwords = wordsize = 0; |
316 | wrapwords = NULL; |
317 | while (text) { |
318 | if (nwords >= wordsize) { |
319 | wordsize = nwords + 64; |
320 | wrapwords = srealloc(wrapwords, wordsize * sizeof(*wrapwords)); |
321 | } |
322 | wrapwords[nwords].width = 0; |
323 | wrapwords[nwords].begin = text; |
324 | while (text) { |
43341922 |
325 | wrapwords[nwords].width += widthfn(ctx, text); |
d7482997 |
326 | wrapwords[nwords].end = text->next; |
327 | if (text->next && (text->next->type == word_WhiteSpace || |
328 | text->next->type == word_EmphSpace || |
329 | text->breaks)) |
330 | break; |
331 | text = text->next; |
332 | } |
333 | if (text && text->next && (text->next->type == word_WhiteSpace || |
334 | text->next->type == word_EmphSpace)) { |
43341922 |
335 | wrapwords[nwords].spacewidth = widthfn(ctx, text->next); |
d7482997 |
336 | text = text->next; |
337 | } else { |
338 | wrapwords[nwords].spacewidth = 0; |
339 | } |
340 | nwords++; |
341 | if (text) |
342 | text = text->next; |
343 | } |
344 | |
345 | /* |
346 | * Perform the dynamic wrapping algorithm: work backwards from |
347 | * nwords-1, determining the optimal wrapping for each terminal |
348 | * subsequence of the paragraph. |
349 | */ |
350 | for (i = nwords; i-- ;) { |
351 | int best = -1; |
352 | int bestcost = 0; |
353 | int cost; |
43341922 |
354 | int linelen = 0, spacewidth = 0, minspacewidth = 0; |
355 | int nspaces; |
d7482997 |
356 | int thiswidth = (i == 0 ? width : subsequentwidth); |
357 | |
358 | j = 0; |
43341922 |
359 | nspaces = 0; |
d7482997 |
360 | while (i+j < nwords) { |
361 | /* |
362 | * See what happens if we put j+1 words on this line. |
363 | */ |
43341922 |
364 | if (spacewidth) { |
365 | nspaces++; |
366 | minspacewidth = spacewidth; |
367 | } |
d7482997 |
368 | linelen += spacewidth + wrapwords[i+j].width; |
369 | spacewidth = wrapwords[i+j].spacewidth; |
370 | j++; |
371 | if (linelen > thiswidth) { |
372 | /* |
373 | * If we're over the width limit, abandon ship, |
374 | * _unless_ there is no best-effort yet (which will |
375 | * only happen if the first word is too long all by |
376 | * itself). |
377 | */ |
378 | if (best > 0) |
379 | break; |
380 | } |
43341922 |
381 | |
382 | /* |
383 | * Compute the cost of this line. The method of doing |
384 | * this differs hugely depending on whether |
385 | * natural_space is nonzero or not. |
386 | */ |
387 | if (natural_space) { |
388 | if (!nspaces && linelen > thiswidth) { |
389 | /* |
390 | * Special case: if there are no spaces at all |
391 | * on the line because one single word is too |
392 | * long for its line, cost is zero because |
393 | * there's nothing we can do about it anyway. |
394 | */ |
395 | cost = 0; |
396 | } else { |
397 | int shortfall = thiswidth - linelen; |
398 | int spaceextra = shortfall / (nspaces ? nspaces : 1); |
399 | int spaceshortfall = natural_space - |
400 | (minspacewidth + spaceextra); |
401 | |
402 | if (i+j == nwords && spaceshortfall < 0) { |
403 | /* |
404 | * Special case: on the very last line of |
405 | * the paragraph, we don't score penalty |
406 | * points for having to _stretch_ the line, |
407 | * since we won't stretch it anyway. |
408 | * However, we score penalties as normal |
409 | * for having to squeeze it. |
410 | */ |
411 | cost = 0; |
412 | } else { |
413 | /* |
414 | * Squaring this number is tricky since |
415 | * it's liable to be quite big. Let's |
416 | * divide it through by 256. |
417 | */ |
418 | int x = spaceshortfall >> 8; |
419 | int xf = spaceshortfall & 0xFF; |
420 | |
421 | /* |
422 | * Not counting strange variable-fixed- |
423 | * point oddities, we are computing |
424 | * |
425 | * (x+xf)^2 = x^2 + 2*x*xf + xf*xf |
426 | * |
427 | * except that _our_ xf is 256 times the |
428 | * one listed there. |
429 | */ |
430 | |
431 | cost = x * x; |
432 | cost += (2 * x * xf) >> 8; |
433 | } |
434 | } |
d7482997 |
435 | } else { |
43341922 |
436 | if (i+j == nwords) { |
437 | /* |
438 | * Special case: if we're at the very end of the |
439 | * paragraph, we don't score penalty points for the |
440 | * white space left on the line. |
441 | */ |
442 | cost = 0; |
443 | } else { |
444 | cost = (thiswidth-linelen) * (thiswidth-linelen); |
445 | } |
d7482997 |
446 | } |
43341922 |
447 | |
448 | /* |
449 | * Add in the cost of wrapping all lines after this |
450 | * point too. |
451 | */ |
452 | if (i+j < nwords) |
453 | cost += wrapwords[i+j].cost; |
454 | |
d7482997 |
455 | /* |
456 | * We compare bestcost >= cost, not bestcost > cost, |
457 | * because in cases where the costs are identical we |
458 | * want to try to look like the greedy algorithm, |
459 | * because readers are likely to have spent a lot of |
460 | * time looking at greedy-wrapped paragraphs and |
461 | * there's no point violating the Principle of Least |
462 | * Surprise if it doesn't actually gain anything. |
463 | */ |
464 | if (best < 0 || bestcost >= cost) { |
465 | bestcost = cost; |
466 | best = j; |
467 | } |
468 | } |
469 | /* |
470 | * Now we know the optimal answer for this terminal |
471 | * subsequence, so put it in wrapwords. |
472 | */ |
473 | wrapwords[i].cost = bestcost; |
474 | wrapwords[i].nwords = best; |
475 | } |
476 | |
477 | /* |
478 | * We've wrapped the paragraph. Now build the output |
479 | * `wrappedline' list. |
480 | */ |
481 | i = 0; |
482 | while (i < nwords) { |
f1530049 |
483 | wrappedline *w = snew(wrappedline); |
d7482997 |
484 | *ptr = w; |
485 | ptr = &w->next; |
486 | w->next = NULL; |
487 | |
488 | n = wrapwords[i].nwords; |
489 | w->begin = wrapwords[i].begin; |
490 | w->end = wrapwords[i+n-1].end; |
491 | |
492 | /* |
493 | * Count along the words to find nspaces and shortfall. |
494 | */ |
495 | w->nspaces = 0; |
496 | w->shortfall = width; |
497 | for (j = 0; j < n; j++) { |
498 | w->shortfall -= wrapwords[i+j].width; |
499 | if (j < n-1 && wrapwords[i+j].spacewidth) { |
500 | w->nspaces++; |
501 | w->shortfall -= wrapwords[i+j].spacewidth; |
502 | } |
503 | } |
504 | i += n; |
505 | } |
506 | |
507 | sfree(wrapwords); |
508 | |
509 | return head; |
510 | } |
511 | |
512 | void wrap_free(wrappedline *w) { |
513 | while (w) { |
514 | wrappedline *t = w->next; |
515 | sfree(w); |
516 | w = t; |
517 | } |
518 | } |
e4ea58f8 |
519 | |
520 | void cmdline_cfg_add(paragraph *cfg, char *string) |
521 | { |
522 | wchar_t *ustring; |
523 | int upos, ulen, pos, len; |
524 | |
525 | ulen = 0; |
526 | while (cfg->keyword[ulen]) |
527 | ulen += 1 + ustrlen(cfg->keyword+ulen); |
528 | len = 0; |
529 | while (cfg->origkeyword[len]) |
530 | len += 1 + strlen(cfg->origkeyword+len); |
531 | |
7e976207 |
532 | ustring = ufroma_locale_dup(string); |
e4ea58f8 |
533 | |
534 | upos = ulen; |
535 | ulen += 2 + ustrlen(ustring); |
f1530049 |
536 | cfg->keyword = sresize(cfg->keyword, ulen, wchar_t); |
e4ea58f8 |
537 | ustrcpy(cfg->keyword+upos, ustring); |
538 | cfg->keyword[ulen-1] = L'\0'; |
539 | |
540 | pos = len; |
541 | len += 2 + strlen(string); |
f1530049 |
542 | cfg->origkeyword = sresize(cfg->origkeyword, len, char); |
e4ea58f8 |
543 | strcpy(cfg->origkeyword+pos, string); |
544 | cfg->origkeyword[len-1] = '\0'; |
545 | |
546 | sfree(ustring); |
547 | } |
548 | |
549 | paragraph *cmdline_cfg_new(void) |
550 | { |
551 | paragraph *p; |
552 | |
f1530049 |
553 | p = snew(paragraph); |
e4ea58f8 |
554 | memset(p, 0, sizeof(*p)); |
555 | p->type = para_Config; |
556 | p->next = NULL; |
557 | p->fpos.filename = "<command line>"; |
558 | p->fpos.line = p->fpos.col = -1; |
559 | p->keyword = ustrdup(L"\0"); |
560 | p->origkeyword = dupstr("\0"); |
561 | |
562 | return p; |
563 | } |
564 | |
565 | paragraph *cmdline_cfg_simple(char *string, ...) |
566 | { |
567 | va_list ap; |
568 | char *s; |
569 | paragraph *p; |
570 | |
571 | p = cmdline_cfg_new(); |
572 | cmdline_cfg_add(p, string); |
573 | |
574 | va_start(ap, string); |
575 | while ((s = va_arg(ap, char *)) != NULL) |
576 | cmdline_cfg_add(p, s); |
577 | va_end(ap); |
578 | |
579 | return p; |
580 | } |