43341922 |
1 | /* |
2 | * Paper printing pre-backend for Halibut. |
3 | * |
4 | * This module does all the processing common to both PostScript |
5 | * and PDF output: selecting fonts, line wrapping and page breaking |
6 | * in accordance with font metrics, laying out the contents and |
7 | * index pages, generally doing all the page layout. After this, |
8 | * bk_ps.c and bk_pdf.c should only need to do linear translations |
9 | * into their literal output format. |
10 | */ |
11 | |
12 | /* |
13 | * To be done: |
14 | * |
43341922 |
15 | * - set up contents section now we know what sections begin on |
16 | * which pages |
17 | * |
43341922 |
18 | * - do PDF outline |
19 | * |
43341922 |
20 | * - index |
21 | * |
22 | * That should bring us to the same level of functionality that |
23 | * original-Halibut had, and the same in PDF plus the obvious |
24 | * interactive navigation features. After that, in future work: |
25 | * |
26 | * - linearised PDF, perhaps? |
27 | * |
28 | * - I'm uncertain of whether I need to include a ToUnicode CMap |
29 | * in each of my font definitions in PDF. Currently things (by |
30 | * which I mean cut and paste out of acroread) seem to be |
31 | * working fairly happily without it, but I don't know. |
32 | * |
33 | * - configurability |
34 | * |
35 | * - title pages |
36 | */ |
37 | |
38 | #include <assert.h> |
39 | #include <stdio.h> |
40 | |
41 | #include "halibut.h" |
42 | #include "paper.h" |
43 | |
44 | static font_data *make_std_font(font_list *fontlist, char const *name); |
45 | static void wrap_paragraph(para_data *pdata, word *words, |
46 | int w, int i1, int i2); |
47 | static page_data *page_breaks(line_data *first, line_data *last, |
48 | int page_height); |
138d7ffb |
49 | static void render_line(line_data *ldata, int left_x, int top_y, |
50 | xref_dest *dest, keywordlist *keywords); |
515d216b |
51 | static int paper_width_simple(para_data *pdata, word *text); |
52 | static void code_paragraph(para_data *pdata, |
53 | font_data *fn, font_data *fi, font_data *fb, |
54 | int font_size, int indent, word *words); |
43341922 |
55 | |
56 | void *paper_pre_backend(paragraph *sourceform, keywordlist *keywords, |
57 | indexdata *idx) { |
58 | paragraph *p; |
59 | document *doc; |
515d216b |
60 | int indent, extra_indent, firstline_indent, aux_indent; |
43341922 |
61 | para_data *pdata; |
62 | line_data *ldata, *firstline, *lastline; |
515d216b |
63 | font_data *tr, *ti, *hr, *hi, *cr, *co, *cb; |
43341922 |
64 | page_data *pages; |
65 | font_list *fontlist; |
515d216b |
66 | word *aux, *aux2; |
43341922 |
67 | |
68 | /* |
69 | * FIXME: All these things ought to become configurable. |
70 | */ |
71 | int paper_width = 595 * 4096; |
72 | int paper_height = 841 * 4096; |
73 | int left_margin = 72 * 4096; |
74 | int top_margin = 72 * 4096; |
75 | int right_margin = 72 * 4096; |
76 | int bottom_margin = 108 * 4096; |
77 | int indent_list_bullet = 6 * 4096; |
78 | int indent_list = 24 * 4096; |
79 | int indent_quote = 18 * 4096; |
80 | int base_leading = 4096; |
81 | int base_para_spacing = 10 * 4096; |
515d216b |
82 | int chapter_top_space = 72 * 4096; |
83 | int sect_num_left_space = 12 * 4096; |
43341922 |
84 | |
85 | int base_width = paper_width - left_margin - right_margin; |
86 | int page_height = paper_height - top_margin - bottom_margin; |
87 | |
88 | IGNORE(keywords); /* FIXME */ |
89 | IGNORE(idx); /* FIXME */ |
43341922 |
90 | |
91 | /* |
92 | * First, set up some font structures. |
93 | */ |
94 | fontlist = mknew(font_list); |
95 | fontlist->head = fontlist->tail = NULL; |
96 | tr = make_std_font(fontlist, "Times-Roman"); |
97 | ti = make_std_font(fontlist, "Times-Italic"); |
515d216b |
98 | hr = make_std_font(fontlist, "Helvetica-Bold"); |
99 | hi = make_std_font(fontlist, "Helvetica-BoldOblique"); |
43341922 |
100 | cr = make_std_font(fontlist, "Courier"); |
515d216b |
101 | co = make_std_font(fontlist, "Courier-Oblique"); |
102 | cb = make_std_font(fontlist, "Courier-Bold"); |
43341922 |
103 | |
104 | /* |
105 | * Go through and break up each paragraph into lines. |
106 | */ |
107 | indent = 0; |
108 | firstline = lastline = NULL; |
109 | for (p = sourceform; p; p = p->next) { |
110 | p->private_data = NULL; |
111 | |
112 | switch (p->type) { |
113 | /* |
114 | * These paragraph types are either invisible or don't |
115 | * define text in the normal sense. Either way, they |
116 | * don't require wrapping. |
117 | */ |
118 | case para_IM: |
119 | case para_BR: |
120 | case para_Rule: |
121 | case para_Biblio: |
122 | case para_NotParaType: |
123 | case para_Config: |
124 | case para_VersionID: |
125 | case para_NoCite: |
126 | break; |
127 | |
128 | /* |
129 | * These paragraph types don't require wrapping, but |
130 | * they do affect the line width to which we wrap the |
131 | * rest of the paragraphs, so we need to pay attention. |
132 | */ |
133 | case para_LcontPush: |
134 | indent += indent_list; break; |
135 | case para_LcontPop: |
136 | indent -= indent_list; assert(indent >= 0); break; |
137 | case para_QuotePush: |
138 | indent += indent_quote; break; |
139 | case para_QuotePop: |
140 | indent -= indent_quote; assert(indent >= 0); break; |
141 | |
142 | /* |
143 | * This paragraph type is special. Process it |
144 | * specially. |
145 | */ |
146 | case para_Code: |
515d216b |
147 | pdata = mknew(para_data); |
148 | code_paragraph(pdata, cr, co, cb, 12, indent, p->words); |
149 | p->private_data = pdata; |
43341922 |
150 | break; |
151 | |
152 | /* |
153 | * All of these paragraph types require wrapping in the |
154 | * ordinary way. So we must supply a set of fonts, a |
155 | * line width and auxiliary information (e.g. bullet |
156 | * text) for each one. |
157 | */ |
158 | case para_Chapter: |
159 | case para_Appendix: |
160 | case para_UnnumberedChapter: |
161 | case para_Heading: |
162 | case para_Subsect: |
163 | case para_Normal: |
164 | case para_BiblioCited: |
165 | case para_Bullet: |
166 | case para_NumberedList: |
167 | case para_DescribedThing: |
168 | case para_Description: |
169 | case para_Copyright: |
170 | case para_Title: |
171 | pdata = mknew(para_data); |
172 | |
173 | /* |
515d216b |
174 | * Choose fonts for this paragraph. |
175 | * |
176 | * FIXME: All of this ought to be completely |
177 | * user-configurable. |
43341922 |
178 | */ |
515d216b |
179 | switch (p->type) { |
180 | case para_Title: |
181 | pdata->fonts[FONT_NORMAL] = hr; |
182 | pdata->sizes[FONT_NORMAL] = 24; |
183 | pdata->fonts[FONT_EMPH] = hi; |
184 | pdata->sizes[FONT_EMPH] = 24; |
185 | pdata->fonts[FONT_CODE] = cb; |
186 | pdata->sizes[FONT_CODE] = 24; |
187 | break; |
188 | |
189 | case para_Chapter: |
190 | case para_Appendix: |
191 | case para_UnnumberedChapter: |
192 | pdata->fonts[FONT_NORMAL] = hr; |
193 | pdata->sizes[FONT_NORMAL] = 20; |
194 | pdata->fonts[FONT_EMPH] = hi; |
195 | pdata->sizes[FONT_EMPH] = 20; |
196 | pdata->fonts[FONT_CODE] = cb; |
197 | pdata->sizes[FONT_CODE] = 20; |
198 | break; |
199 | |
200 | case para_Heading: |
201 | case para_Subsect: |
202 | pdata->fonts[FONT_NORMAL] = hr; |
203 | pdata->fonts[FONT_EMPH] = hi; |
204 | pdata->fonts[FONT_CODE] = cb; |
205 | pdata->sizes[FONT_NORMAL] = |
206 | pdata->sizes[FONT_EMPH] = |
207 | pdata->sizes[FONT_CODE] = |
208 | (p->aux == 0 ? 16 : p->aux == 1 ? 14 : 13); |
209 | break; |
210 | |
211 | case para_Normal: |
212 | case para_BiblioCited: |
213 | case para_Bullet: |
214 | case para_NumberedList: |
215 | case para_DescribedThing: |
216 | case para_Description: |
217 | case para_Copyright: |
218 | pdata->fonts[FONT_NORMAL] = tr; |
219 | pdata->sizes[FONT_NORMAL] = 12; |
220 | pdata->fonts[FONT_EMPH] = ti; |
221 | pdata->sizes[FONT_EMPH] = 12; |
222 | pdata->fonts[FONT_CODE] = cr; |
223 | pdata->sizes[FONT_CODE] = 12; |
224 | break; |
225 | } |
43341922 |
226 | |
227 | /* |
515d216b |
228 | * Also select an indentation level depending on the |
229 | * paragraph type (list paragraphs other than |
43341922 |
230 | * para_DescribedThing need extra indent). |
231 | * |
515d216b |
232 | * (FIXME: Perhaps at some point we might even arrange |
233 | * for the user to be able to request indented first |
234 | * lines in paragraphs.) |
43341922 |
235 | */ |
515d216b |
236 | if (p->type == para_Bullet || |
237 | p->type == para_NumberedList || |
238 | p->type == para_Description) { |
239 | extra_indent = firstline_indent = indent_list; |
240 | } else { |
241 | extra_indent = firstline_indent = 0; |
242 | } |
243 | |
244 | /* |
245 | * Find the auxiliary text for this paragraph. |
246 | */ |
247 | aux = aux2 = NULL; |
248 | aux_indent = 0; |
249 | |
250 | switch (p->type) { |
251 | case para_Chapter: |
252 | case para_Appendix: |
253 | case para_Heading: |
254 | case para_Subsect: |
255 | /* |
256 | * For some heading styles (FIXME: be able to |
257 | * configure which), the auxiliary text contains |
258 | * the chapter number and is arranged to be |
259 | * right-aligned a few points left of the primary |
260 | * margin. For other styles, the auxiliary text is |
261 | * the full chapter _name_ and takes up space |
262 | * within the (wrapped) chapter title, meaning that |
263 | * we must move the first line indent over to make |
264 | * space for it. |
265 | */ |
266 | if (p->type == para_Heading || p->type == para_Subsect) { |
267 | int len; |
268 | |
269 | aux = p->kwtext2; |
270 | len = paper_width_simple(pdata, p->kwtext2); |
271 | aux_indent = -len - sect_num_left_space; |
272 | } else { |
273 | aux = p->kwtext; |
274 | aux2 = mknew(word); |
275 | aux2->next = NULL; |
276 | aux2->alt = NULL; |
277 | aux2->type = word_Normal; |
278 | aux2->text = ustrdup(L": "); |
279 | aux2->breaks = FALSE; |
280 | aux2->aux = 0; |
281 | aux_indent = 0; |
282 | |
283 | firstline_indent += paper_width_simple(pdata, aux); |
284 | firstline_indent += paper_width_simple(pdata, aux2); |
285 | } |
286 | break; |
287 | |
288 | case para_Bullet: |
289 | /* |
290 | * Auxiliary text consisting of a bullet. (FIXME: |
291 | * configurable bullet.) |
292 | */ |
293 | aux = mknew(word); |
294 | aux->next = NULL; |
295 | aux->alt = NULL; |
296 | aux->type = word_Normal; |
297 | aux->text = ustrdup(L"\x2022"); |
298 | aux->breaks = FALSE; |
299 | aux->aux = 0; |
300 | aux_indent = indent + indent_list_bullet; |
301 | break; |
302 | |
303 | case para_NumberedList: |
304 | /* |
305 | * Auxiliary text consisting of the number followed |
306 | * by a (FIXME: configurable) full stop. |
307 | */ |
308 | aux = p->kwtext; |
309 | aux2 = mknew(word); |
310 | aux2->next = NULL; |
311 | aux2->alt = NULL; |
312 | aux2->type = word_Normal; |
313 | aux2->text = ustrdup(L"."); |
314 | aux2->breaks = FALSE; |
315 | aux2->aux = 0; |
316 | aux_indent = indent + indent_list_bullet; |
317 | break; |
318 | |
319 | case para_BiblioCited: |
320 | /* |
321 | * Auxiliary text consisting of the bibliography |
322 | * reference text, and a trailing space. |
323 | */ |
324 | aux = p->kwtext; |
325 | aux2 = mknew(word); |
326 | aux2->next = NULL; |
327 | aux2->alt = NULL; |
328 | aux2->type = word_Normal; |
329 | aux2->text = ustrdup(L" "); |
330 | aux2->breaks = FALSE; |
331 | aux2->aux = 0; |
332 | aux_indent = indent; |
333 | firstline_indent += paper_width_simple(pdata, aux); |
334 | firstline_indent += paper_width_simple(pdata, aux2); |
335 | break; |
336 | } |
43341922 |
337 | |
338 | wrap_paragraph(pdata, p->words, base_width, |
339 | indent + firstline_indent, |
340 | indent + extra_indent); |
341 | |
43341922 |
342 | p->private_data = pdata; |
343 | |
515d216b |
344 | pdata->first->aux_text = aux; |
345 | pdata->first->aux_text_2 = aux2; |
346 | pdata->first->aux_left_indent = aux_indent; |
347 | |
348 | break; |
349 | } |
350 | |
351 | if (p->private_data) { |
352 | pdata = (para_data *)p->private_data; |
353 | |
43341922 |
354 | /* |
355 | * Set the line spacing for each line in this paragraph. |
356 | */ |
357 | for (ldata = pdata->first; ldata; ldata = ldata->next) { |
358 | if (ldata == pdata->first) |
359 | ldata->space_before = base_para_spacing / 2; |
360 | else |
361 | ldata->space_before = base_leading / 2; |
362 | if (ldata == pdata->last) |
363 | ldata->space_after = base_para_spacing / 2; |
364 | else |
365 | ldata->space_after = base_leading / 2; |
366 | ldata->page_break = FALSE; |
367 | } |
368 | |
369 | /* |
515d216b |
370 | * Some kinds of section heading do require a page |
371 | * break before them. |
43341922 |
372 | */ |
515d216b |
373 | if (p->type == para_Title || |
374 | p->type == para_Chapter || |
375 | p->type == para_Appendix || |
376 | p->type == para_UnnumberedChapter) { |
377 | pdata->first->page_break = TRUE; |
378 | pdata->first->space_before = chapter_top_space; |
379 | } |
43341922 |
380 | |
515d216b |
381 | /* |
382 | * Link all line structures together into a big list. |
383 | */ |
43341922 |
384 | if (pdata->first) { |
385 | if (lastline) { |
386 | lastline->next = pdata->first; |
387 | pdata->first->prev = lastline; |
388 | } else { |
389 | firstline = pdata->first; |
390 | pdata->first->prev = NULL; |
391 | } |
392 | lastline = pdata->last; |
393 | } |
394 | } |
395 | } |
396 | |
397 | /* |
398 | * Now we have an enormous linked list of every line of text in |
399 | * the document. Break it up into pages. |
400 | */ |
401 | pages = page_breaks(firstline, lastline, page_height); |
402 | |
403 | /* |
404 | * Now we're ready to actually lay out the pages. We do this by |
405 | * looping over _paragraphs_, since we may need to track cross- |
406 | * references between lines and even across pages. |
407 | */ |
408 | for (p = sourceform; p; p = p->next) { |
409 | pdata = (para_data *)p->private_data; |
410 | |
411 | if (pdata) { |
138d7ffb |
412 | xref_dest dest; |
413 | dest.type = NONE; |
43341922 |
414 | for (ldata = pdata->first; ldata; ldata = ldata->next) { |
138d7ffb |
415 | render_line(ldata, left_margin, paper_height - top_margin, |
416 | &dest, keywords); |
43341922 |
417 | if (ldata == pdata->last) |
418 | break; |
419 | } |
420 | } |
421 | } |
422 | |
423 | doc = mknew(document); |
424 | doc->fonts = fontlist; |
425 | doc->pages = pages; |
426 | doc->paper_width = paper_width; |
427 | doc->paper_height = paper_height; |
428 | return doc; |
429 | } |
430 | |
431 | static font_encoding *new_font_encoding(font_data *font) |
432 | { |
433 | font_encoding *fe; |
434 | int i; |
435 | |
436 | fe = mknew(font_encoding); |
437 | fe->next = NULL; |
438 | |
439 | if (font->list->tail) |
440 | font->list->tail->next = fe; |
441 | else |
442 | font->list->head = fe; |
443 | font->list->tail = fe; |
444 | |
445 | fe->font = font; |
446 | fe->free_pos = 0x21; |
447 | |
448 | for (i = 0; i < 256; i++) { |
449 | fe->vector[i] = NULL; |
450 | fe->indices[i] = -1; |
451 | fe->to_unicode[i] = 0xFFFF; |
452 | } |
453 | |
454 | return fe; |
455 | } |
456 | |
457 | static font_data *make_std_font(font_list *fontlist, char const *name) |
458 | { |
459 | const int *widths; |
460 | int nglyphs; |
461 | font_data *f; |
462 | font_encoding *fe; |
463 | int i; |
464 | |
465 | widths = ps_std_font_widths(name); |
466 | if (!widths) |
467 | return NULL; |
468 | |
469 | for (nglyphs = 0; ps_std_glyphs[nglyphs] != NULL; nglyphs++); |
470 | |
471 | f = mknew(font_data); |
472 | |
473 | f->list = fontlist; |
474 | f->name = name; |
475 | f->nglyphs = nglyphs; |
476 | f->glyphs = ps_std_glyphs; |
477 | f->widths = widths; |
478 | f->subfont_map = mknewa(subfont_map_entry, nglyphs); |
479 | |
480 | /* |
481 | * Our first subfont will contain all of US-ASCII. This isn't |
482 | * really necessary - we could just create custom subfonts |
483 | * precisely as the whim of render_string dictated - but |
484 | * instinct suggests that it might be nice to have the text in |
485 | * the output files look _marginally_ recognisable. |
486 | */ |
487 | fe = new_font_encoding(f); |
488 | fe->free_pos = 0xA1; /* only the top half is free */ |
489 | f->latest_subfont = fe; |
490 | |
491 | for (i = 0; i < (int)lenof(f->bmp); i++) |
492 | f->bmp[i] = 0xFFFF; |
493 | |
494 | for (i = 0; i < nglyphs; i++) { |
495 | wchar_t ucs; |
496 | ucs = ps_glyph_to_unicode(f->glyphs[i]); |
497 | assert(ucs != 0xFFFF); |
498 | f->bmp[ucs] = i; |
499 | if (ucs >= 0x20 && ucs <= 0x7E) { |
500 | fe->vector[ucs] = f->glyphs[i]; |
501 | fe->indices[ucs] = i; |
502 | fe->to_unicode[ucs] = ucs; |
503 | f->subfont_map[i].subfont = fe; |
504 | f->subfont_map[i].position = ucs; |
505 | } else { |
506 | /* |
507 | * This character is not yet assigned to a subfont. |
508 | */ |
509 | f->subfont_map[i].subfont = NULL; |
510 | f->subfont_map[i].position = 0; |
511 | } |
512 | } |
513 | |
514 | return f; |
515 | } |
516 | |
517 | static int string_width(font_data *font, wchar_t const *string, int *errs) |
518 | { |
519 | int width = 0; |
520 | |
521 | if (errs) |
522 | *errs = 0; |
523 | |
524 | for (; *string; string++) { |
525 | int index; |
526 | |
527 | index = font->bmp[(unsigned short)*string]; |
528 | if (index == 0xFFFF) { |
529 | if (errs) |
530 | *errs = 1; |
531 | } else { |
532 | width += font->widths[index]; |
533 | } |
534 | } |
535 | |
536 | return width; |
537 | } |
538 | |
faad4952 |
539 | static int paper_width_internal(void *vctx, word *word, int *nspaces); |
43341922 |
540 | |
541 | struct paper_width_ctx { |
542 | int minspacewidth; |
543 | para_data *pdata; |
544 | }; |
545 | |
faad4952 |
546 | static int paper_width_list(void *vctx, word *text, word *end, int *nspaces) { |
43341922 |
547 | int w = 0; |
faad4952 |
548 | while (text && text != end) { |
549 | w += paper_width_internal(vctx, text, nspaces); |
43341922 |
550 | text = text->next; |
551 | } |
552 | return w; |
553 | } |
554 | |
faad4952 |
555 | static int paper_width_internal(void *vctx, word *word, int *nspaces) |
43341922 |
556 | { |
557 | struct paper_width_ctx *ctx = (struct paper_width_ctx *)vctx; |
558 | int style, type, findex, width, errs; |
559 | wchar_t *str; |
560 | |
561 | switch (word->type) { |
562 | case word_HyperLink: |
563 | case word_HyperEnd: |
564 | case word_UpperXref: |
565 | case word_LowerXref: |
566 | case word_XrefEnd: |
567 | case word_IndexRef: |
568 | return 0; |
569 | } |
570 | |
571 | style = towordstyle(word->type); |
572 | type = removeattr(word->type); |
573 | |
574 | findex = (style == word_Normal ? FONT_NORMAL : |
575 | style == word_Emph ? FONT_EMPH : |
576 | FONT_CODE); |
577 | |
578 | if (type == word_Normal) { |
579 | str = word->text; |
580 | } else if (type == word_WhiteSpace) { |
faad4952 |
581 | if (findex != FONT_CODE) { |
582 | if (nspaces) |
583 | (*nspaces)++; |
43341922 |
584 | return ctx->minspacewidth; |
faad4952 |
585 | } else |
43341922 |
586 | str = L" "; |
587 | } else /* if (type == word_Quote) */ { |
588 | if (word->aux == quote_Open) |
589 | str = L"\x2018"; /* FIXME: configurability! */ |
590 | else |
591 | str = L"\x2019"; /* FIXME: configurability! */ |
592 | } |
593 | |
594 | width = string_width(ctx->pdata->fonts[findex], str, &errs); |
595 | |
596 | if (errs && word->alt) |
faad4952 |
597 | return paper_width_list(vctx, word->alt, NULL, nspaces); |
43341922 |
598 | else |
599 | return ctx->pdata->sizes[findex] * width; |
600 | } |
601 | |
faad4952 |
602 | static int paper_width(void *vctx, word *word) |
603 | { |
604 | return paper_width_internal(vctx, word, NULL); |
605 | } |
606 | |
515d216b |
607 | static int paper_width_simple(para_data *pdata, word *text) |
608 | { |
609 | struct paper_width_ctx ctx; |
610 | |
611 | ctx.pdata = pdata; |
612 | ctx.minspacewidth = |
613 | (pdata->sizes[FONT_NORMAL] * |
614 | string_width(pdata->fonts[FONT_NORMAL], L" ", NULL)); |
615 | |
616 | return paper_width_list(&ctx, text, NULL, NULL); |
617 | } |
618 | |
43341922 |
619 | static void wrap_paragraph(para_data *pdata, word *words, |
620 | int w, int i1, int i2) |
621 | { |
622 | wrappedline *wrapping, *p; |
623 | int spacewidth; |
624 | struct paper_width_ctx ctx; |
625 | int line_height; |
626 | |
627 | /* |
628 | * We're going to need to store the line height in every line |
629 | * structure we generate. |
630 | */ |
631 | { |
632 | int i; |
633 | line_height = 0; |
634 | for (i = 0; i < NFONTS; i++) |
635 | if (line_height < pdata->sizes[i]) |
636 | line_height = pdata->sizes[i]; |
637 | line_height *= 4096; |
638 | } |
639 | |
640 | spacewidth = (pdata->sizes[FONT_NORMAL] * |
641 | string_width(pdata->fonts[FONT_NORMAL], L" ", NULL)); |
642 | if (spacewidth == 0) { |
643 | /* |
644 | * A font without a space?! Disturbing. I hope this never |
645 | * comes up, but I'll make a random guess anyway and set my |
646 | * space width to half the point size. |
647 | */ |
648 | spacewidth = pdata->sizes[FONT_NORMAL] * 4096 / 2; |
649 | } |
650 | |
651 | /* |
652 | * I'm going to set the _minimum_ space width to 3/5 of the |
653 | * standard one, and use the standard one as the optimum. |
654 | */ |
655 | ctx.minspacewidth = spacewidth * 3 / 5; |
656 | ctx.pdata = pdata; |
657 | |
658 | wrapping = wrap_para(words, w - i1, w - i2, paper_width, &ctx, spacewidth); |
659 | |
660 | /* |
661 | * Having done the wrapping, we now concoct a set of line_data |
662 | * structures. |
663 | */ |
664 | pdata->first = pdata->last = NULL; |
665 | |
666 | for (p = wrapping; p; p = p->next) { |
667 | line_data *ldata; |
668 | word *wd; |
669 | int len, wid, spaces; |
670 | |
671 | ldata = mknew(line_data); |
672 | |
673 | ldata->pdata = pdata; |
674 | ldata->first = p->begin; |
faad4952 |
675 | ldata->end = p->end; |
43341922 |
676 | ldata->line_height = line_height; |
677 | |
678 | ldata->xpos = (p == wrapping ? i1 : i2); |
679 | |
680 | if (pdata->last) { |
681 | pdata->last->next = ldata; |
682 | ldata->prev = pdata->last; |
683 | } else { |
684 | pdata->first = ldata; |
685 | ldata->prev = NULL; |
686 | } |
687 | ldata->next = NULL; |
688 | pdata->last = ldata; |
689 | |
43341922 |
690 | spaces = 0; |
faad4952 |
691 | len = paper_width_list(&ctx, ldata->first, ldata->end, &spaces); |
692 | wid = (p == wrapping ? w - i1 : w - i2); |
43341922 |
693 | wd = ldata->first; |
43341922 |
694 | |
faad4952 |
695 | ldata->hshortfall = wid - len; |
696 | ldata->nspaces = spaces; |
697 | /* |
698 | * This tells us how much the space width needs to |
699 | * change from _min_spacewidth. But we want to store |
700 | * its difference from the _natural_ space width, to |
701 | * make the text rendering easier. |
702 | */ |
703 | ldata->hshortfall += ctx.minspacewidth * spaces; |
704 | ldata->hshortfall -= spacewidth * spaces; |
705 | /* |
706 | * Special case: on the last line of a paragraph, we |
707 | * never stretch spaces. |
708 | */ |
709 | if (ldata->hshortfall > 0 && !p->next) |
710 | ldata->hshortfall = 0; |
43341922 |
711 | |
712 | ldata->aux_text = NULL; |
515d216b |
713 | ldata->aux_text_2 = NULL; |
43341922 |
714 | ldata->aux_left_indent = 0; |
715 | } |
716 | |
717 | } |
718 | |
719 | static page_data *page_breaks(line_data *first, line_data *last, |
720 | int page_height) |
721 | { |
722 | line_data *l, *m; |
723 | page_data *ph, *pt; |
724 | |
725 | /* |
726 | * Page breaking is done by a close analogue of the optimal |
727 | * paragraph wrapping algorithm used by wrap_para(). We work |
728 | * backwards from the end of the document line by line; for |
729 | * each line, we contemplate every possible number of lines we |
730 | * could put on a page starting with that line, determine a |
731 | * cost function for each one, add it to the pre-computed cost |
732 | * function for optimally page-breaking everything after that |
733 | * page, and pick the best option. |
734 | * |
735 | * Since my line_data structures are only used for this |
736 | * purpose, I might as well just store the algorithm data |
737 | * directly in them. |
738 | */ |
739 | |
740 | for (l = last; l; l = l->prev) { |
741 | int minheight, text = 0, space = 0; |
742 | int cost; |
743 | |
744 | l->bestcost = -1; |
745 | for (m = l; m; m = m->next) { |
746 | if (m != l && m->page_break) |
747 | break; /* we've gone as far as we can */ |
748 | |
749 | if (m != l) |
750 | space += m->prev->space_after; |
751 | if (m != l || m->page_break) |
752 | space += m->space_before; |
753 | text += m->line_height; |
754 | minheight = text + space; |
755 | |
756 | if (m != l && minheight > page_height) |
757 | break; |
758 | |
759 | /* |
760 | * Compute the cost of this arrangement, as the square |
761 | * of the amount of wasted space on the page. |
762 | * Exception: if this is the last page before a |
763 | * mandatory break or the document end, we don't |
764 | * penalise a large blank area. |
765 | */ |
766 | if (m->next && !m->next->page_break) |
767 | { |
768 | int x = page_height - minheight; |
769 | int xf; |
770 | |
771 | xf = x & 0xFF; |
772 | x >>= 8; |
773 | |
774 | cost = x*x; |
775 | cost += (x * xf) >> 8; |
776 | } else |
777 | cost = 0; |
778 | |
779 | /* |
780 | * FIXME: here I should introduce penalties for |
781 | * breaking in mid-paragraph, particularly very close |
782 | * to one end of a paragraph and particularly in code |
783 | * paragraphs. |
784 | */ |
785 | |
786 | if (m->next && !m->next->page_break) |
787 | cost += m->next->bestcost; |
788 | |
789 | if (l->bestcost == -1 || l->bestcost > cost) { |
790 | /* |
791 | * This is the best option yet for this starting |
792 | * point. |
793 | */ |
794 | l->bestcost = cost; |
795 | if (m->next && !m->next->page_break) |
faad4952 |
796 | l->vshortfall = page_height - minheight; |
43341922 |
797 | else |
faad4952 |
798 | l->vshortfall = 0; |
43341922 |
799 | l->text = text; |
800 | l->space = space; |
801 | l->page_last = m; |
802 | } |
803 | } |
804 | } |
805 | |
806 | /* |
807 | * Now go through the line list forwards and assemble the |
808 | * actual pages. |
809 | */ |
810 | ph = pt = NULL; |
811 | |
812 | l = first; |
813 | while (l) { |
814 | page_data *page; |
815 | int text, space; |
816 | |
817 | page = mknew(page_data); |
818 | page->next = NULL; |
819 | page->prev = pt; |
820 | if (pt) |
821 | pt->next = page; |
822 | else |
823 | ph = page; |
824 | pt = page; |
825 | |
826 | page->first_line = l; |
827 | page->last_line = l->page_last; |
828 | |
829 | page->first_text = page->last_text = NULL; |
830 | |
138d7ffb |
831 | page->first_xref = page->last_xref = NULL; |
832 | |
43341922 |
833 | /* |
834 | * Now assign a y-coordinate to each line on the page. |
835 | */ |
836 | text = space = 0; |
837 | for (l = page->first_line; l; l = l->next) { |
838 | if (l != page->first_line) |
839 | space += l->prev->space_after; |
840 | if (l != page->first_line || l->page_break) |
841 | space += l->space_before; |
842 | text += l->line_height; |
843 | |
844 | l->page = page; |
845 | l->ypos = text + space + |
faad4952 |
846 | space * (float)page->first_line->vshortfall / |
43341922 |
847 | page->first_line->space; |
848 | |
849 | if (l == page->last_line) |
850 | break; |
851 | } |
852 | |
853 | l = page->last_line->next; |
854 | } |
855 | |
856 | return ph; |
857 | } |
858 | |
859 | static void add_string_to_page(page_data *page, int x, int y, |
860 | font_encoding *fe, int size, char *text) |
861 | { |
862 | text_fragment *frag; |
863 | |
864 | frag = mknew(text_fragment); |
865 | frag->next = NULL; |
866 | |
867 | if (page->last_text) |
868 | page->last_text->next = frag; |
869 | else |
870 | page->first_text = frag; |
871 | page->last_text = frag; |
872 | |
873 | frag->x = x; |
874 | frag->y = y; |
875 | frag->fe = fe; |
876 | frag->fontsize = size; |
877 | frag->text = dupstr(text); |
878 | } |
879 | |
880 | /* |
881 | * Returns the updated x coordinate. |
882 | */ |
883 | static int render_string(page_data *page, font_data *font, int fontsize, |
884 | int x, int y, wchar_t *str) |
885 | { |
886 | char *text; |
887 | int textpos, textwid, glyph; |
888 | font_encoding *subfont = NULL, *sf; |
889 | |
890 | text = mknewa(char, 1 + ustrlen(str)); |
891 | textpos = textwid = 0; |
892 | |
893 | while (*str) { |
894 | glyph = font->bmp[*str]; |
895 | |
896 | if (glyph == 0xFFFF) |
897 | continue; /* nothing more we can do here */ |
898 | |
899 | /* |
900 | * Find which subfont this character is going in. |
901 | */ |
902 | sf = font->subfont_map[glyph].subfont; |
903 | |
904 | if (!sf) { |
905 | int c; |
906 | |
907 | /* |
908 | * This character is not yet in a subfont. Assign one. |
909 | */ |
910 | if (font->latest_subfont->free_pos >= 0x100) |
911 | font->latest_subfont = new_font_encoding(font); |
912 | |
913 | c = font->latest_subfont->free_pos++; |
914 | if (font->latest_subfont->free_pos == 0x7F) |
915 | font->latest_subfont->free_pos = 0xA1; |
916 | |
917 | font->subfont_map[glyph].subfont = font->latest_subfont; |
918 | font->subfont_map[glyph].position = c; |
919 | font->latest_subfont->vector[c] = font->glyphs[glyph]; |
920 | font->latest_subfont->indices[c] = glyph; |
921 | font->latest_subfont->to_unicode[c] = *str; |
922 | |
923 | sf = font->latest_subfont; |
924 | } |
925 | |
926 | if (!subfont || sf != subfont) { |
927 | if (subfont) { |
928 | text[textpos] = '\0'; |
929 | add_string_to_page(page, x, y, subfont, fontsize, text); |
930 | x += textwid; |
931 | } else { |
932 | assert(textpos == 0); |
933 | } |
934 | textpos = 0; |
935 | subfont = sf; |
936 | } |
937 | |
938 | text[textpos++] = font->subfont_map[glyph].position; |
939 | textwid += font->widths[glyph] * fontsize; |
940 | |
941 | str++; |
942 | } |
943 | |
944 | if (textpos > 0) { |
945 | text[textpos] = '\0'; |
946 | add_string_to_page(page, x, y, subfont, fontsize, text); |
947 | x += textwid; |
948 | } |
949 | |
950 | return x; |
951 | } |
952 | |
953 | /* |
954 | * Returns the updated x coordinate. |
955 | */ |
138d7ffb |
956 | static int render_text(page_data *page, para_data *pdata, line_data *ldata, |
957 | int x, int y, word *text, word *text_end, xref **xr, |
958 | int shortfall, int nspaces, int *nspace, |
959 | keywordlist *keywords) |
43341922 |
960 | { |
faad4952 |
961 | while (text && text != text_end) { |
43341922 |
962 | int style, type, findex, errs; |
963 | wchar_t *str; |
138d7ffb |
964 | xref_dest dest; |
43341922 |
965 | |
966 | switch (text->type) { |
138d7ffb |
967 | /* |
968 | * Start a cross-reference. |
969 | */ |
43341922 |
970 | case word_HyperLink: |
43341922 |
971 | case word_UpperXref: |
972 | case word_LowerXref: |
138d7ffb |
973 | |
974 | if (text->type == word_HyperLink) { |
975 | dest.type = URL; |
976 | dest.url = utoa_dup(text->text); |
977 | dest.page = NULL; |
978 | } else { |
979 | keyword *kwl = kw_lookup(keywords, text->text); |
980 | para_data *pdata; |
981 | |
982 | if (kwl) { |
983 | assert(kwl->para->private_data); |
984 | pdata = (para_data *) kwl->para->private_data; |
985 | dest.type = PAGE; |
986 | dest.page = pdata->first->page; |
987 | dest.url = NULL; |
988 | } else { |
989 | /* |
990 | * Shouldn't happen, but *shrug* |
991 | */ |
992 | dest.type = NONE; |
993 | dest.page = NULL; |
994 | dest.url = NULL; |
995 | } |
996 | } |
997 | if (dest.type != NONE) { |
998 | *xr = mknew(xref); |
999 | (*xr)->dest = dest; /* structure copy */ |
1000 | if (page->last_xref) |
1001 | page->last_xref->next = *xr; |
1002 | else |
1003 | page->first_xref = *xr; |
1004 | page->last_xref = *xr; |
1005 | |
1006 | /* |
1007 | * FIXME: Ideally we should have, and use, some |
1008 | * vertical font metric information here so that |
1009 | * our cross-ref rectangle can take account of |
1010 | * descenders and the font's cap height. This will |
1011 | * do for the moment, but it isn't ideal. |
1012 | */ |
1013 | (*xr)->lx = (*xr)->rx = x; |
1014 | (*xr)->by = y; |
1015 | (*xr)->ty = y + ldata->line_height; |
1016 | } |
1017 | goto nextword; |
1018 | |
1019 | /* |
1020 | * Finish extending a cross-reference box. |
1021 | */ |
1022 | case word_HyperEnd: |
43341922 |
1023 | case word_XrefEnd: |
138d7ffb |
1024 | *xr = NULL; |
1025 | goto nextword; |
1026 | |
43341922 |
1027 | case word_IndexRef: |
1028 | goto nextword; |
1029 | /* |
1030 | * FIXME: we should do something with all of these! |
1031 | * Hyperlinks and xrefs have meaning in PDF, and this |
1032 | * is probably the right place to nail down the index |
1033 | * references too. |
1034 | */ |
1035 | } |
1036 | |
1037 | style = towordstyle(text->type); |
1038 | type = removeattr(text->type); |
1039 | |
1040 | findex = (style == word_Normal ? FONT_NORMAL : |
1041 | style == word_Emph ? FONT_EMPH : |
1042 | FONT_CODE); |
1043 | |
1044 | if (type == word_Normal) { |
1045 | str = text->text; |
1046 | } else if (type == word_WhiteSpace) { |
1047 | x += pdata->sizes[findex] * |
1048 | string_width(pdata->fonts[findex], L" ", NULL); |
faad4952 |
1049 | if (nspaces && findex != FONT_CODE) { |
1050 | x += (*nspace+1) * shortfall / nspaces; |
1051 | x -= *nspace * shortfall / nspaces; |
1052 | (*nspace)++; |
1053 | } |
43341922 |
1054 | goto nextword; |
1055 | } else /* if (type == word_Quote) */ { |
1056 | if (text->aux == quote_Open) |
1057 | str = L"\x2018"; /* FIXME: configurability! */ |
1058 | else |
1059 | str = L"\x2019"; /* FIXME: configurability! */ |
1060 | } |
1061 | |
1062 | (void) string_width(pdata->fonts[findex], str, &errs); |
1063 | |
1064 | if (errs && text->alt) |
138d7ffb |
1065 | x = render_text(page, pdata, ldata, x, y, text->alt, NULL, |
1066 | xr, shortfall, nspaces, nspace, keywords); |
43341922 |
1067 | else |
1068 | x = render_string(page, pdata->fonts[findex], |
1069 | pdata->sizes[findex], x, y, str); |
1070 | |
138d7ffb |
1071 | if (*xr) |
1072 | (*xr)->rx = x; |
1073 | |
43341922 |
1074 | nextword: |
43341922 |
1075 | text = text->next; |
1076 | } |
1077 | |
1078 | return x; |
1079 | } |
1080 | |
138d7ffb |
1081 | static void render_line(line_data *ldata, int left_x, int top_y, |
1082 | xref_dest *dest, keywordlist *keywords) |
43341922 |
1083 | { |
faad4952 |
1084 | int nspace; |
138d7ffb |
1085 | xref *xr; |
1086 | |
faad4952 |
1087 | if (ldata->aux_text) { |
515d216b |
1088 | int x; |
138d7ffb |
1089 | xr = NULL; |
faad4952 |
1090 | nspace = 0; |
515d216b |
1091 | x = render_text(ldata->page, ldata->pdata, ldata, |
1092 | left_x + ldata->aux_left_indent, |
1093 | top_y - ldata->ypos, |
1094 | ldata->aux_text, NULL, &xr, 0, 0, &nspace, keywords); |
1095 | if (ldata->aux_text_2) |
1096 | render_text(ldata->page, ldata->pdata, ldata, |
1097 | x, top_y - ldata->ypos, |
1098 | ldata->aux_text_2, NULL, &xr, 0, 0, &nspace, keywords); |
faad4952 |
1099 | } |
1100 | nspace = 0; |
138d7ffb |
1101 | |
1102 | /* |
1103 | * There might be a cross-reference carried over from a |
1104 | * previous line. |
1105 | */ |
1106 | if (dest->type != NONE) { |
1107 | xr = mknew(xref); |
1108 | xr->dest = *dest; /* structure copy */ |
1109 | if (ldata->page->last_xref) |
1110 | ldata->page->last_xref->next = xr; |
1111 | else |
1112 | ldata->page->first_xref = xr; |
1113 | ldata->page->last_xref = xr; |
1114 | xr->lx = xr->rx = left_x + ldata->xpos; |
1115 | xr->by = top_y - ldata->ypos; |
1116 | xr->ty = top_y - ldata->ypos + ldata->line_height; |
1117 | } else |
1118 | xr = NULL; |
1119 | |
1120 | render_text(ldata->page, ldata->pdata, ldata, left_x + ldata->xpos, |
1121 | top_y - ldata->ypos, ldata->first, ldata->end, &xr, |
1122 | ldata->hshortfall, ldata->nspaces, &nspace, keywords); |
1123 | |
1124 | if (xr) { |
1125 | /* |
1126 | * There's a cross-reference continued on to the next line. |
1127 | */ |
1128 | *dest = xr->dest; |
1129 | } else |
1130 | dest->type = NONE; |
43341922 |
1131 | } |
515d216b |
1132 | |
1133 | static void code_paragraph(para_data *pdata, |
1134 | font_data *fn, font_data *fi, font_data *fb, |
1135 | int font_size, int indent, word *words) |
1136 | { |
1137 | /* |
1138 | * For code paragraphs, I'm going to hack grievously and |
1139 | * pretend the three normal fonts are the three code paragraph |
1140 | * fonts. |
1141 | */ |
1142 | pdata->fonts[FONT_NORMAL] = fb; |
1143 | pdata->fonts[FONT_EMPH] = fi; |
1144 | pdata->fonts[FONT_CODE] = fn; |
1145 | pdata->sizes[FONT_NORMAL] = |
1146 | pdata->sizes[FONT_EMPH] = |
1147 | pdata->sizes[FONT_CODE] = font_size; |
1148 | |
1149 | pdata->first = pdata->last = NULL; |
1150 | |
1151 | for (; words; words = words->next) { |
1152 | wchar_t *t, *e, *start; |
1153 | word *lhead = NULL, *ltail = NULL, *w; |
1154 | line_data *ldata; |
1155 | int prev = -1, curr; |
1156 | |
1157 | t = words->text; |
1158 | if (words->next && words->next->type == word_Emph) { |
1159 | e = words->next->text; |
1160 | words = words->next; |
1161 | } else |
1162 | e = NULL; |
1163 | |
1164 | start = t; |
1165 | |
1166 | while (*start) { |
1167 | while (*t) { |
1168 | if (!e || !*e) |
1169 | curr = 0; |
1170 | else if (*e == L'i') |
1171 | curr = 1; |
1172 | else if (*e == L'b') |
1173 | curr = 2; |
1174 | else |
1175 | curr = 0; |
1176 | |
1177 | if (prev < 0) |
1178 | prev = curr; |
1179 | |
1180 | if (curr != prev) |
1181 | break; |
1182 | |
1183 | t++; |
1184 | if (e && *e) |
1185 | e++; |
1186 | } |
1187 | |
1188 | /* |
1189 | * We've isolated a maximal subsequence of the line |
1190 | * which has the same emphasis. Form it into a word |
1191 | * structure. |
1192 | */ |
1193 | w = mknew(word); |
1194 | w->next = NULL; |
1195 | w->alt = NULL; |
1196 | w->type = (prev == 0 ? word_WeakCode : |
1197 | prev == 1 ? word_Emph : word_Normal); |
1198 | w->text = mknewa(wchar_t, t-start+1); |
1199 | memcpy(w->text, start, (t-start) * sizeof(wchar_t)); |
1200 | w->text[t-start] = '\0'; |
1201 | w->breaks = FALSE; |
1202 | |
1203 | if (ltail) |
1204 | ltail->next = w; |
1205 | else |
1206 | lhead = w; |
1207 | ltail = w; |
1208 | |
1209 | start = t; |
1210 | prev = -1; |
1211 | } |
1212 | |
1213 | ldata = mknew(line_data); |
1214 | |
1215 | ldata->pdata = pdata; |
1216 | ldata->first = lhead; |
1217 | ldata->end = NULL; |
1218 | ldata->line_height = font_size * 4096; |
1219 | |
1220 | ldata->xpos = indent; |
1221 | |
1222 | if (pdata->last) { |
1223 | pdata->last->next = ldata; |
1224 | ldata->prev = pdata->last; |
1225 | } else { |
1226 | pdata->first = ldata; |
1227 | ldata->prev = NULL; |
1228 | } |
1229 | ldata->next = NULL; |
1230 | pdata->last = ldata; |
1231 | |
1232 | ldata->hshortfall = 0; |
1233 | ldata->nspaces = 0; |
1234 | ldata->aux_text = NULL; |
1235 | ldata->aux_text_2 = NULL; |
1236 | ldata->aux_left_indent = 0; |
1237 | |
1238 | } |
1239 | } |