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 | * |
a8070b46 |
15 | * - tune the page breaking algorithm to impose penalties on |
16 | * various things |
17 | * * breaking in the middle of a code paragraph |
18 | * * breaking one line from the start or end of a paragraph |
19 | * * breaking immediately after a heading of any kind (or |
20 | * indeed within one) |
21 | * * we may also need to impose a limit on the amount by which |
22 | * we can _stretch_ a page; after a certain point we may |
23 | * prefer just to unapologetically leave space at the bottom. |
24 | * |
25 | * - implement some simple graphics |
26 | * * I had an underline below chapter headings in the original |
27 | * Perl version, and I thought it looked rather nice |
28 | * * also we need para_Rule. |
29 | * |
43341922 |
30 | * - set up contents section now we know what sections begin on |
31 | * which pages |
32 | * |
43341922 |
33 | * - do PDF outline |
34 | * |
43341922 |
35 | * - index |
36 | * |
a8070b46 |
37 | * - header/footer? Page numbers at least would be handy. Fully |
38 | * configurable footer can wait, though. |
39 | * |
43341922 |
40 | * That should bring us to the same level of functionality that |
41 | * original-Halibut had, and the same in PDF plus the obvious |
42 | * interactive navigation features. After that, in future work: |
43 | * |
44 | * - linearised PDF, perhaps? |
45 | * |
46 | * - I'm uncertain of whether I need to include a ToUnicode CMap |
47 | * in each of my font definitions in PDF. Currently things (by |
48 | * which I mean cut and paste out of acroread) seem to be |
49 | * working fairly happily without it, but I don't know. |
50 | * |
51 | * - configurability |
52 | * |
53 | * - title pages |
54 | */ |
55 | |
56 | #include <assert.h> |
57 | #include <stdio.h> |
58 | |
59 | #include "halibut.h" |
60 | #include "paper.h" |
61 | |
62 | static font_data *make_std_font(font_list *fontlist, char const *name); |
63 | static void wrap_paragraph(para_data *pdata, word *words, |
64 | int w, int i1, int i2); |
65 | static page_data *page_breaks(line_data *first, line_data *last, |
66 | int page_height); |
138d7ffb |
67 | static void render_line(line_data *ldata, int left_x, int top_y, |
68 | xref_dest *dest, keywordlist *keywords); |
515d216b |
69 | static int paper_width_simple(para_data *pdata, word *text); |
70 | static void code_paragraph(para_data *pdata, |
71 | font_data *fn, font_data *fi, font_data *fb, |
72 | int font_size, int indent, word *words); |
43341922 |
73 | |
74 | void *paper_pre_backend(paragraph *sourceform, keywordlist *keywords, |
75 | indexdata *idx) { |
76 | paragraph *p; |
77 | document *doc; |
515d216b |
78 | int indent, extra_indent, firstline_indent, aux_indent; |
43341922 |
79 | para_data *pdata; |
80 | line_data *ldata, *firstline, *lastline; |
515d216b |
81 | font_data *tr, *ti, *hr, *hi, *cr, *co, *cb; |
43341922 |
82 | page_data *pages; |
83 | font_list *fontlist; |
515d216b |
84 | word *aux, *aux2; |
43341922 |
85 | |
86 | /* |
87 | * FIXME: All these things ought to become configurable. |
88 | */ |
89 | int paper_width = 595 * 4096; |
90 | int paper_height = 841 * 4096; |
91 | int left_margin = 72 * 4096; |
92 | int top_margin = 72 * 4096; |
93 | int right_margin = 72 * 4096; |
94 | int bottom_margin = 108 * 4096; |
95 | int indent_list_bullet = 6 * 4096; |
96 | int indent_list = 24 * 4096; |
97 | int indent_quote = 18 * 4096; |
98 | int base_leading = 4096; |
99 | int base_para_spacing = 10 * 4096; |
515d216b |
100 | int chapter_top_space = 72 * 4096; |
101 | int sect_num_left_space = 12 * 4096; |
43341922 |
102 | |
103 | int base_width = paper_width - left_margin - right_margin; |
104 | int page_height = paper_height - top_margin - bottom_margin; |
105 | |
43341922 |
106 | IGNORE(idx); /* FIXME */ |
43341922 |
107 | |
108 | /* |
109 | * First, set up some font structures. |
110 | */ |
111 | fontlist = mknew(font_list); |
112 | fontlist->head = fontlist->tail = NULL; |
113 | tr = make_std_font(fontlist, "Times-Roman"); |
114 | ti = make_std_font(fontlist, "Times-Italic"); |
515d216b |
115 | hr = make_std_font(fontlist, "Helvetica-Bold"); |
116 | hi = make_std_font(fontlist, "Helvetica-BoldOblique"); |
43341922 |
117 | cr = make_std_font(fontlist, "Courier"); |
515d216b |
118 | co = make_std_font(fontlist, "Courier-Oblique"); |
119 | cb = make_std_font(fontlist, "Courier-Bold"); |
43341922 |
120 | |
121 | /* |
122 | * Go through and break up each paragraph into lines. |
123 | */ |
124 | indent = 0; |
125 | firstline = lastline = NULL; |
126 | for (p = sourceform; p; p = p->next) { |
127 | p->private_data = NULL; |
128 | |
129 | switch (p->type) { |
130 | /* |
131 | * These paragraph types are either invisible or don't |
132 | * define text in the normal sense. Either way, they |
133 | * don't require wrapping. |
134 | */ |
135 | case para_IM: |
136 | case para_BR: |
137 | case para_Rule: |
138 | case para_Biblio: |
139 | case para_NotParaType: |
140 | case para_Config: |
141 | case para_VersionID: |
142 | case para_NoCite: |
143 | break; |
144 | |
145 | /* |
146 | * These paragraph types don't require wrapping, but |
147 | * they do affect the line width to which we wrap the |
148 | * rest of the paragraphs, so we need to pay attention. |
149 | */ |
150 | case para_LcontPush: |
151 | indent += indent_list; break; |
152 | case para_LcontPop: |
153 | indent -= indent_list; assert(indent >= 0); break; |
154 | case para_QuotePush: |
155 | indent += indent_quote; break; |
156 | case para_QuotePop: |
157 | indent -= indent_quote; assert(indent >= 0); break; |
158 | |
159 | /* |
160 | * This paragraph type is special. Process it |
161 | * specially. |
162 | */ |
163 | case para_Code: |
515d216b |
164 | pdata = mknew(para_data); |
165 | code_paragraph(pdata, cr, co, cb, 12, indent, p->words); |
166 | p->private_data = pdata; |
43341922 |
167 | break; |
168 | |
169 | /* |
170 | * All of these paragraph types require wrapping in the |
171 | * ordinary way. So we must supply a set of fonts, a |
172 | * line width and auxiliary information (e.g. bullet |
173 | * text) for each one. |
174 | */ |
175 | case para_Chapter: |
176 | case para_Appendix: |
177 | case para_UnnumberedChapter: |
178 | case para_Heading: |
179 | case para_Subsect: |
180 | case para_Normal: |
181 | case para_BiblioCited: |
182 | case para_Bullet: |
183 | case para_NumberedList: |
184 | case para_DescribedThing: |
185 | case para_Description: |
186 | case para_Copyright: |
187 | case para_Title: |
188 | pdata = mknew(para_data); |
189 | |
190 | /* |
515d216b |
191 | * Choose fonts for this paragraph. |
192 | * |
193 | * FIXME: All of this ought to be completely |
194 | * user-configurable. |
43341922 |
195 | */ |
515d216b |
196 | switch (p->type) { |
197 | case para_Title: |
198 | pdata->fonts[FONT_NORMAL] = hr; |
199 | pdata->sizes[FONT_NORMAL] = 24; |
200 | pdata->fonts[FONT_EMPH] = hi; |
201 | pdata->sizes[FONT_EMPH] = 24; |
202 | pdata->fonts[FONT_CODE] = cb; |
203 | pdata->sizes[FONT_CODE] = 24; |
204 | break; |
205 | |
206 | case para_Chapter: |
207 | case para_Appendix: |
208 | case para_UnnumberedChapter: |
209 | pdata->fonts[FONT_NORMAL] = hr; |
210 | pdata->sizes[FONT_NORMAL] = 20; |
211 | pdata->fonts[FONT_EMPH] = hi; |
212 | pdata->sizes[FONT_EMPH] = 20; |
213 | pdata->fonts[FONT_CODE] = cb; |
214 | pdata->sizes[FONT_CODE] = 20; |
215 | break; |
216 | |
217 | case para_Heading: |
218 | case para_Subsect: |
219 | pdata->fonts[FONT_NORMAL] = hr; |
220 | pdata->fonts[FONT_EMPH] = hi; |
221 | pdata->fonts[FONT_CODE] = cb; |
222 | pdata->sizes[FONT_NORMAL] = |
223 | pdata->sizes[FONT_EMPH] = |
224 | pdata->sizes[FONT_CODE] = |
225 | (p->aux == 0 ? 16 : p->aux == 1 ? 14 : 13); |
226 | break; |
227 | |
228 | case para_Normal: |
229 | case para_BiblioCited: |
230 | case para_Bullet: |
231 | case para_NumberedList: |
232 | case para_DescribedThing: |
233 | case para_Description: |
234 | case para_Copyright: |
235 | pdata->fonts[FONT_NORMAL] = tr; |
236 | pdata->sizes[FONT_NORMAL] = 12; |
237 | pdata->fonts[FONT_EMPH] = ti; |
238 | pdata->sizes[FONT_EMPH] = 12; |
239 | pdata->fonts[FONT_CODE] = cr; |
240 | pdata->sizes[FONT_CODE] = 12; |
241 | break; |
242 | } |
43341922 |
243 | |
244 | /* |
515d216b |
245 | * Also select an indentation level depending on the |
246 | * paragraph type (list paragraphs other than |
43341922 |
247 | * para_DescribedThing need extra indent). |
248 | * |
515d216b |
249 | * (FIXME: Perhaps at some point we might even arrange |
250 | * for the user to be able to request indented first |
251 | * lines in paragraphs.) |
43341922 |
252 | */ |
515d216b |
253 | if (p->type == para_Bullet || |
254 | p->type == para_NumberedList || |
255 | p->type == para_Description) { |
256 | extra_indent = firstline_indent = indent_list; |
257 | } else { |
258 | extra_indent = firstline_indent = 0; |
259 | } |
260 | |
261 | /* |
262 | * Find the auxiliary text for this paragraph. |
263 | */ |
264 | aux = aux2 = NULL; |
265 | aux_indent = 0; |
266 | |
267 | switch (p->type) { |
268 | case para_Chapter: |
269 | case para_Appendix: |
270 | case para_Heading: |
271 | case para_Subsect: |
272 | /* |
273 | * For some heading styles (FIXME: be able to |
274 | * configure which), the auxiliary text contains |
275 | * the chapter number and is arranged to be |
276 | * right-aligned a few points left of the primary |
277 | * margin. For other styles, the auxiliary text is |
278 | * the full chapter _name_ and takes up space |
279 | * within the (wrapped) chapter title, meaning that |
280 | * we must move the first line indent over to make |
281 | * space for it. |
282 | */ |
283 | if (p->type == para_Heading || p->type == para_Subsect) { |
284 | int len; |
285 | |
286 | aux = p->kwtext2; |
287 | len = paper_width_simple(pdata, p->kwtext2); |
288 | aux_indent = -len - sect_num_left_space; |
289 | } else { |
290 | aux = p->kwtext; |
291 | aux2 = mknew(word); |
292 | aux2->next = NULL; |
293 | aux2->alt = NULL; |
294 | aux2->type = word_Normal; |
295 | aux2->text = ustrdup(L": "); |
296 | aux2->breaks = FALSE; |
297 | aux2->aux = 0; |
298 | aux_indent = 0; |
299 | |
300 | firstline_indent += paper_width_simple(pdata, aux); |
301 | firstline_indent += paper_width_simple(pdata, aux2); |
302 | } |
303 | break; |
304 | |
305 | case para_Bullet: |
306 | /* |
307 | * Auxiliary text consisting of a bullet. (FIXME: |
308 | * configurable bullet.) |
309 | */ |
310 | aux = mknew(word); |
311 | aux->next = NULL; |
312 | aux->alt = NULL; |
313 | aux->type = word_Normal; |
314 | aux->text = ustrdup(L"\x2022"); |
315 | aux->breaks = FALSE; |
316 | aux->aux = 0; |
317 | aux_indent = indent + indent_list_bullet; |
318 | break; |
319 | |
320 | case para_NumberedList: |
321 | /* |
322 | * Auxiliary text consisting of the number followed |
323 | * by a (FIXME: configurable) full stop. |
324 | */ |
325 | aux = p->kwtext; |
326 | aux2 = mknew(word); |
327 | aux2->next = NULL; |
328 | aux2->alt = NULL; |
329 | aux2->type = word_Normal; |
330 | aux2->text = ustrdup(L"."); |
331 | aux2->breaks = FALSE; |
332 | aux2->aux = 0; |
333 | aux_indent = indent + indent_list_bullet; |
334 | break; |
335 | |
336 | case para_BiblioCited: |
337 | /* |
338 | * Auxiliary text consisting of the bibliography |
339 | * reference text, and a trailing space. |
340 | */ |
341 | aux = p->kwtext; |
342 | aux2 = mknew(word); |
343 | aux2->next = NULL; |
344 | aux2->alt = NULL; |
345 | aux2->type = word_Normal; |
346 | aux2->text = ustrdup(L" "); |
347 | aux2->breaks = FALSE; |
348 | aux2->aux = 0; |
349 | aux_indent = indent; |
350 | firstline_indent += paper_width_simple(pdata, aux); |
351 | firstline_indent += paper_width_simple(pdata, aux2); |
352 | break; |
353 | } |
43341922 |
354 | |
355 | wrap_paragraph(pdata, p->words, base_width, |
356 | indent + firstline_indent, |
357 | indent + extra_indent); |
358 | |
43341922 |
359 | p->private_data = pdata; |
360 | |
515d216b |
361 | pdata->first->aux_text = aux; |
362 | pdata->first->aux_text_2 = aux2; |
363 | pdata->first->aux_left_indent = aux_indent; |
364 | |
365 | break; |
366 | } |
367 | |
368 | if (p->private_data) { |
369 | pdata = (para_data *)p->private_data; |
370 | |
43341922 |
371 | /* |
372 | * Set the line spacing for each line in this paragraph. |
373 | */ |
374 | for (ldata = pdata->first; ldata; ldata = ldata->next) { |
375 | if (ldata == pdata->first) |
376 | ldata->space_before = base_para_spacing / 2; |
377 | else |
378 | ldata->space_before = base_leading / 2; |
379 | if (ldata == pdata->last) |
380 | ldata->space_after = base_para_spacing / 2; |
381 | else |
382 | ldata->space_after = base_leading / 2; |
383 | ldata->page_break = FALSE; |
384 | } |
385 | |
386 | /* |
515d216b |
387 | * Some kinds of section heading do require a page |
388 | * break before them. |
43341922 |
389 | */ |
515d216b |
390 | if (p->type == para_Title || |
391 | p->type == para_Chapter || |
392 | p->type == para_Appendix || |
393 | p->type == para_UnnumberedChapter) { |
394 | pdata->first->page_break = TRUE; |
395 | pdata->first->space_before = chapter_top_space; |
396 | } |
43341922 |
397 | |
515d216b |
398 | /* |
399 | * Link all line structures together into a big list. |
400 | */ |
43341922 |
401 | if (pdata->first) { |
402 | if (lastline) { |
403 | lastline->next = pdata->first; |
404 | pdata->first->prev = lastline; |
405 | } else { |
406 | firstline = pdata->first; |
407 | pdata->first->prev = NULL; |
408 | } |
409 | lastline = pdata->last; |
410 | } |
411 | } |
412 | } |
413 | |
414 | /* |
415 | * Now we have an enormous linked list of every line of text in |
416 | * the document. Break it up into pages. |
417 | */ |
418 | pages = page_breaks(firstline, lastline, page_height); |
419 | |
420 | /* |
421 | * Now we're ready to actually lay out the pages. We do this by |
422 | * looping over _paragraphs_, since we may need to track cross- |
423 | * references between lines and even across pages. |
424 | */ |
425 | for (p = sourceform; p; p = p->next) { |
426 | pdata = (para_data *)p->private_data; |
427 | |
428 | if (pdata) { |
138d7ffb |
429 | xref_dest dest; |
430 | dest.type = NONE; |
43341922 |
431 | for (ldata = pdata->first; ldata; ldata = ldata->next) { |
138d7ffb |
432 | render_line(ldata, left_margin, paper_height - top_margin, |
433 | &dest, keywords); |
43341922 |
434 | if (ldata == pdata->last) |
435 | break; |
436 | } |
437 | } |
438 | } |
439 | |
440 | doc = mknew(document); |
441 | doc->fonts = fontlist; |
442 | doc->pages = pages; |
443 | doc->paper_width = paper_width; |
444 | doc->paper_height = paper_height; |
445 | return doc; |
446 | } |
447 | |
448 | static font_encoding *new_font_encoding(font_data *font) |
449 | { |
450 | font_encoding *fe; |
451 | int i; |
452 | |
453 | fe = mknew(font_encoding); |
454 | fe->next = NULL; |
455 | |
456 | if (font->list->tail) |
457 | font->list->tail->next = fe; |
458 | else |
459 | font->list->head = fe; |
460 | font->list->tail = fe; |
461 | |
462 | fe->font = font; |
463 | fe->free_pos = 0x21; |
464 | |
465 | for (i = 0; i < 256; i++) { |
466 | fe->vector[i] = NULL; |
467 | fe->indices[i] = -1; |
468 | fe->to_unicode[i] = 0xFFFF; |
469 | } |
470 | |
471 | return fe; |
472 | } |
473 | |
474 | static font_data *make_std_font(font_list *fontlist, char const *name) |
475 | { |
476 | const int *widths; |
477 | int nglyphs; |
478 | font_data *f; |
479 | font_encoding *fe; |
480 | int i; |
481 | |
482 | widths = ps_std_font_widths(name); |
483 | if (!widths) |
484 | return NULL; |
485 | |
486 | for (nglyphs = 0; ps_std_glyphs[nglyphs] != NULL; nglyphs++); |
487 | |
488 | f = mknew(font_data); |
489 | |
490 | f->list = fontlist; |
491 | f->name = name; |
492 | f->nglyphs = nglyphs; |
493 | f->glyphs = ps_std_glyphs; |
494 | f->widths = widths; |
495 | f->subfont_map = mknewa(subfont_map_entry, nglyphs); |
496 | |
497 | /* |
498 | * Our first subfont will contain all of US-ASCII. This isn't |
499 | * really necessary - we could just create custom subfonts |
500 | * precisely as the whim of render_string dictated - but |
501 | * instinct suggests that it might be nice to have the text in |
502 | * the output files look _marginally_ recognisable. |
503 | */ |
504 | fe = new_font_encoding(f); |
505 | fe->free_pos = 0xA1; /* only the top half is free */ |
506 | f->latest_subfont = fe; |
507 | |
508 | for (i = 0; i < (int)lenof(f->bmp); i++) |
509 | f->bmp[i] = 0xFFFF; |
510 | |
511 | for (i = 0; i < nglyphs; i++) { |
512 | wchar_t ucs; |
513 | ucs = ps_glyph_to_unicode(f->glyphs[i]); |
514 | assert(ucs != 0xFFFF); |
515 | f->bmp[ucs] = i; |
516 | if (ucs >= 0x20 && ucs <= 0x7E) { |
517 | fe->vector[ucs] = f->glyphs[i]; |
518 | fe->indices[ucs] = i; |
519 | fe->to_unicode[ucs] = ucs; |
520 | f->subfont_map[i].subfont = fe; |
521 | f->subfont_map[i].position = ucs; |
522 | } else { |
523 | /* |
524 | * This character is not yet assigned to a subfont. |
525 | */ |
526 | f->subfont_map[i].subfont = NULL; |
527 | f->subfont_map[i].position = 0; |
528 | } |
529 | } |
530 | |
531 | return f; |
532 | } |
533 | |
534 | static int string_width(font_data *font, wchar_t const *string, int *errs) |
535 | { |
536 | int width = 0; |
537 | |
538 | if (errs) |
539 | *errs = 0; |
540 | |
541 | for (; *string; string++) { |
542 | int index; |
543 | |
544 | index = font->bmp[(unsigned short)*string]; |
545 | if (index == 0xFFFF) { |
546 | if (errs) |
547 | *errs = 1; |
548 | } else { |
549 | width += font->widths[index]; |
550 | } |
551 | } |
552 | |
553 | return width; |
554 | } |
555 | |
faad4952 |
556 | static int paper_width_internal(void *vctx, word *word, int *nspaces); |
43341922 |
557 | |
558 | struct paper_width_ctx { |
559 | int minspacewidth; |
560 | para_data *pdata; |
561 | }; |
562 | |
faad4952 |
563 | static int paper_width_list(void *vctx, word *text, word *end, int *nspaces) { |
43341922 |
564 | int w = 0; |
faad4952 |
565 | while (text && text != end) { |
566 | w += paper_width_internal(vctx, text, nspaces); |
43341922 |
567 | text = text->next; |
568 | } |
569 | return w; |
570 | } |
571 | |
faad4952 |
572 | static int paper_width_internal(void *vctx, word *word, int *nspaces) |
43341922 |
573 | { |
574 | struct paper_width_ctx *ctx = (struct paper_width_ctx *)vctx; |
575 | int style, type, findex, width, errs; |
576 | wchar_t *str; |
577 | |
578 | switch (word->type) { |
579 | case word_HyperLink: |
580 | case word_HyperEnd: |
581 | case word_UpperXref: |
582 | case word_LowerXref: |
583 | case word_XrefEnd: |
584 | case word_IndexRef: |
585 | return 0; |
586 | } |
587 | |
588 | style = towordstyle(word->type); |
589 | type = removeattr(word->type); |
590 | |
591 | findex = (style == word_Normal ? FONT_NORMAL : |
592 | style == word_Emph ? FONT_EMPH : |
593 | FONT_CODE); |
594 | |
595 | if (type == word_Normal) { |
596 | str = word->text; |
597 | } else if (type == word_WhiteSpace) { |
faad4952 |
598 | if (findex != FONT_CODE) { |
599 | if (nspaces) |
600 | (*nspaces)++; |
43341922 |
601 | return ctx->minspacewidth; |
faad4952 |
602 | } else |
43341922 |
603 | str = L" "; |
604 | } else /* if (type == word_Quote) */ { |
605 | if (word->aux == quote_Open) |
606 | str = L"\x2018"; /* FIXME: configurability! */ |
607 | else |
608 | str = L"\x2019"; /* FIXME: configurability! */ |
609 | } |
610 | |
611 | width = string_width(ctx->pdata->fonts[findex], str, &errs); |
612 | |
613 | if (errs && word->alt) |
faad4952 |
614 | return paper_width_list(vctx, word->alt, NULL, nspaces); |
43341922 |
615 | else |
616 | return ctx->pdata->sizes[findex] * width; |
617 | } |
618 | |
faad4952 |
619 | static int paper_width(void *vctx, word *word) |
620 | { |
621 | return paper_width_internal(vctx, word, NULL); |
622 | } |
623 | |
515d216b |
624 | static int paper_width_simple(para_data *pdata, word *text) |
625 | { |
626 | struct paper_width_ctx ctx; |
627 | |
628 | ctx.pdata = pdata; |
629 | ctx.minspacewidth = |
630 | (pdata->sizes[FONT_NORMAL] * |
631 | string_width(pdata->fonts[FONT_NORMAL], L" ", NULL)); |
632 | |
633 | return paper_width_list(&ctx, text, NULL, NULL); |
634 | } |
635 | |
43341922 |
636 | static void wrap_paragraph(para_data *pdata, word *words, |
637 | int w, int i1, int i2) |
638 | { |
639 | wrappedline *wrapping, *p; |
640 | int spacewidth; |
641 | struct paper_width_ctx ctx; |
642 | int line_height; |
643 | |
644 | /* |
645 | * We're going to need to store the line height in every line |
646 | * structure we generate. |
647 | */ |
648 | { |
649 | int i; |
650 | line_height = 0; |
651 | for (i = 0; i < NFONTS; i++) |
652 | if (line_height < pdata->sizes[i]) |
653 | line_height = pdata->sizes[i]; |
654 | line_height *= 4096; |
655 | } |
656 | |
657 | spacewidth = (pdata->sizes[FONT_NORMAL] * |
658 | string_width(pdata->fonts[FONT_NORMAL], L" ", NULL)); |
659 | if (spacewidth == 0) { |
660 | /* |
661 | * A font without a space?! Disturbing. I hope this never |
662 | * comes up, but I'll make a random guess anyway and set my |
663 | * space width to half the point size. |
664 | */ |
665 | spacewidth = pdata->sizes[FONT_NORMAL] * 4096 / 2; |
666 | } |
667 | |
668 | /* |
669 | * I'm going to set the _minimum_ space width to 3/5 of the |
670 | * standard one, and use the standard one as the optimum. |
671 | */ |
672 | ctx.minspacewidth = spacewidth * 3 / 5; |
673 | ctx.pdata = pdata; |
674 | |
675 | wrapping = wrap_para(words, w - i1, w - i2, paper_width, &ctx, spacewidth); |
676 | |
677 | /* |
678 | * Having done the wrapping, we now concoct a set of line_data |
679 | * structures. |
680 | */ |
681 | pdata->first = pdata->last = NULL; |
682 | |
683 | for (p = wrapping; p; p = p->next) { |
684 | line_data *ldata; |
685 | word *wd; |
686 | int len, wid, spaces; |
687 | |
688 | ldata = mknew(line_data); |
689 | |
690 | ldata->pdata = pdata; |
691 | ldata->first = p->begin; |
faad4952 |
692 | ldata->end = p->end; |
43341922 |
693 | ldata->line_height = line_height; |
694 | |
695 | ldata->xpos = (p == wrapping ? i1 : i2); |
696 | |
697 | if (pdata->last) { |
698 | pdata->last->next = ldata; |
699 | ldata->prev = pdata->last; |
700 | } else { |
701 | pdata->first = ldata; |
702 | ldata->prev = NULL; |
703 | } |
704 | ldata->next = NULL; |
705 | pdata->last = ldata; |
706 | |
43341922 |
707 | spaces = 0; |
faad4952 |
708 | len = paper_width_list(&ctx, ldata->first, ldata->end, &spaces); |
709 | wid = (p == wrapping ? w - i1 : w - i2); |
43341922 |
710 | wd = ldata->first; |
43341922 |
711 | |
faad4952 |
712 | ldata->hshortfall = wid - len; |
713 | ldata->nspaces = spaces; |
714 | /* |
715 | * This tells us how much the space width needs to |
716 | * change from _min_spacewidth. But we want to store |
717 | * its difference from the _natural_ space width, to |
718 | * make the text rendering easier. |
719 | */ |
720 | ldata->hshortfall += ctx.minspacewidth * spaces; |
721 | ldata->hshortfall -= spacewidth * spaces; |
722 | /* |
723 | * Special case: on the last line of a paragraph, we |
724 | * never stretch spaces. |
725 | */ |
726 | if (ldata->hshortfall > 0 && !p->next) |
727 | ldata->hshortfall = 0; |
43341922 |
728 | |
729 | ldata->aux_text = NULL; |
515d216b |
730 | ldata->aux_text_2 = NULL; |
43341922 |
731 | ldata->aux_left_indent = 0; |
732 | } |
733 | |
734 | } |
735 | |
736 | static page_data *page_breaks(line_data *first, line_data *last, |
737 | int page_height) |
738 | { |
739 | line_data *l, *m; |
740 | page_data *ph, *pt; |
741 | |
742 | /* |
743 | * Page breaking is done by a close analogue of the optimal |
744 | * paragraph wrapping algorithm used by wrap_para(). We work |
745 | * backwards from the end of the document line by line; for |
746 | * each line, we contemplate every possible number of lines we |
747 | * could put on a page starting with that line, determine a |
748 | * cost function for each one, add it to the pre-computed cost |
749 | * function for optimally page-breaking everything after that |
750 | * page, and pick the best option. |
751 | * |
752 | * Since my line_data structures are only used for this |
753 | * purpose, I might as well just store the algorithm data |
754 | * directly in them. |
755 | */ |
756 | |
757 | for (l = last; l; l = l->prev) { |
758 | int minheight, text = 0, space = 0; |
759 | int cost; |
760 | |
761 | l->bestcost = -1; |
762 | for (m = l; m; m = m->next) { |
763 | if (m != l && m->page_break) |
764 | break; /* we've gone as far as we can */ |
765 | |
766 | if (m != l) |
767 | space += m->prev->space_after; |
768 | if (m != l || m->page_break) |
769 | space += m->space_before; |
770 | text += m->line_height; |
771 | minheight = text + space; |
772 | |
773 | if (m != l && minheight > page_height) |
774 | break; |
775 | |
776 | /* |
777 | * Compute the cost of this arrangement, as the square |
778 | * of the amount of wasted space on the page. |
779 | * Exception: if this is the last page before a |
780 | * mandatory break or the document end, we don't |
781 | * penalise a large blank area. |
782 | */ |
783 | if (m->next && !m->next->page_break) |
784 | { |
785 | int x = page_height - minheight; |
786 | int xf; |
787 | |
788 | xf = x & 0xFF; |
789 | x >>= 8; |
790 | |
791 | cost = x*x; |
792 | cost += (x * xf) >> 8; |
793 | } else |
794 | cost = 0; |
795 | |
796 | /* |
797 | * FIXME: here I should introduce penalties for |
798 | * breaking in mid-paragraph, particularly very close |
799 | * to one end of a paragraph and particularly in code |
800 | * paragraphs. |
801 | */ |
802 | |
803 | if (m->next && !m->next->page_break) |
804 | cost += m->next->bestcost; |
805 | |
806 | if (l->bestcost == -1 || l->bestcost > cost) { |
807 | /* |
808 | * This is the best option yet for this starting |
809 | * point. |
810 | */ |
811 | l->bestcost = cost; |
812 | if (m->next && !m->next->page_break) |
faad4952 |
813 | l->vshortfall = page_height - minheight; |
43341922 |
814 | else |
faad4952 |
815 | l->vshortfall = 0; |
43341922 |
816 | l->text = text; |
817 | l->space = space; |
818 | l->page_last = m; |
819 | } |
820 | } |
821 | } |
822 | |
823 | /* |
824 | * Now go through the line list forwards and assemble the |
825 | * actual pages. |
826 | */ |
827 | ph = pt = NULL; |
828 | |
829 | l = first; |
830 | while (l) { |
831 | page_data *page; |
832 | int text, space; |
833 | |
834 | page = mknew(page_data); |
835 | page->next = NULL; |
836 | page->prev = pt; |
837 | if (pt) |
838 | pt->next = page; |
839 | else |
840 | ph = page; |
841 | pt = page; |
842 | |
843 | page->first_line = l; |
844 | page->last_line = l->page_last; |
845 | |
846 | page->first_text = page->last_text = NULL; |
847 | |
138d7ffb |
848 | page->first_xref = page->last_xref = NULL; |
849 | |
43341922 |
850 | /* |
851 | * Now assign a y-coordinate to each line on the page. |
852 | */ |
853 | text = space = 0; |
854 | for (l = page->first_line; l; l = l->next) { |
855 | if (l != page->first_line) |
856 | space += l->prev->space_after; |
857 | if (l != page->first_line || l->page_break) |
858 | space += l->space_before; |
859 | text += l->line_height; |
860 | |
861 | l->page = page; |
862 | l->ypos = text + space + |
faad4952 |
863 | space * (float)page->first_line->vshortfall / |
43341922 |
864 | page->first_line->space; |
865 | |
866 | if (l == page->last_line) |
867 | break; |
868 | } |
869 | |
870 | l = page->last_line->next; |
871 | } |
872 | |
873 | return ph; |
874 | } |
875 | |
876 | static void add_string_to_page(page_data *page, int x, int y, |
877 | font_encoding *fe, int size, char *text) |
878 | { |
879 | text_fragment *frag; |
880 | |
881 | frag = mknew(text_fragment); |
882 | frag->next = NULL; |
883 | |
884 | if (page->last_text) |
885 | page->last_text->next = frag; |
886 | else |
887 | page->first_text = frag; |
888 | page->last_text = frag; |
889 | |
890 | frag->x = x; |
891 | frag->y = y; |
892 | frag->fe = fe; |
893 | frag->fontsize = size; |
894 | frag->text = dupstr(text); |
895 | } |
896 | |
897 | /* |
898 | * Returns the updated x coordinate. |
899 | */ |
900 | static int render_string(page_data *page, font_data *font, int fontsize, |
901 | int x, int y, wchar_t *str) |
902 | { |
903 | char *text; |
904 | int textpos, textwid, glyph; |
905 | font_encoding *subfont = NULL, *sf; |
906 | |
907 | text = mknewa(char, 1 + ustrlen(str)); |
908 | textpos = textwid = 0; |
909 | |
910 | while (*str) { |
911 | glyph = font->bmp[*str]; |
912 | |
913 | if (glyph == 0xFFFF) |
914 | continue; /* nothing more we can do here */ |
915 | |
916 | /* |
917 | * Find which subfont this character is going in. |
918 | */ |
919 | sf = font->subfont_map[glyph].subfont; |
920 | |
921 | if (!sf) { |
922 | int c; |
923 | |
924 | /* |
925 | * This character is not yet in a subfont. Assign one. |
926 | */ |
927 | if (font->latest_subfont->free_pos >= 0x100) |
928 | font->latest_subfont = new_font_encoding(font); |
929 | |
930 | c = font->latest_subfont->free_pos++; |
931 | if (font->latest_subfont->free_pos == 0x7F) |
932 | font->latest_subfont->free_pos = 0xA1; |
933 | |
934 | font->subfont_map[glyph].subfont = font->latest_subfont; |
935 | font->subfont_map[glyph].position = c; |
936 | font->latest_subfont->vector[c] = font->glyphs[glyph]; |
937 | font->latest_subfont->indices[c] = glyph; |
938 | font->latest_subfont->to_unicode[c] = *str; |
939 | |
940 | sf = font->latest_subfont; |
941 | } |
942 | |
943 | if (!subfont || sf != subfont) { |
944 | if (subfont) { |
945 | text[textpos] = '\0'; |
946 | add_string_to_page(page, x, y, subfont, fontsize, text); |
947 | x += textwid; |
948 | } else { |
949 | assert(textpos == 0); |
950 | } |
951 | textpos = 0; |
952 | subfont = sf; |
953 | } |
954 | |
955 | text[textpos++] = font->subfont_map[glyph].position; |
956 | textwid += font->widths[glyph] * fontsize; |
957 | |
958 | str++; |
959 | } |
960 | |
961 | if (textpos > 0) { |
962 | text[textpos] = '\0'; |
963 | add_string_to_page(page, x, y, subfont, fontsize, text); |
964 | x += textwid; |
965 | } |
966 | |
967 | return x; |
968 | } |
969 | |
970 | /* |
971 | * Returns the updated x coordinate. |
972 | */ |
138d7ffb |
973 | static int render_text(page_data *page, para_data *pdata, line_data *ldata, |
974 | int x, int y, word *text, word *text_end, xref **xr, |
975 | int shortfall, int nspaces, int *nspace, |
976 | keywordlist *keywords) |
43341922 |
977 | { |
faad4952 |
978 | while (text && text != text_end) { |
43341922 |
979 | int style, type, findex, errs; |
980 | wchar_t *str; |
138d7ffb |
981 | xref_dest dest; |
43341922 |
982 | |
983 | switch (text->type) { |
138d7ffb |
984 | /* |
985 | * Start a cross-reference. |
986 | */ |
43341922 |
987 | case word_HyperLink: |
43341922 |
988 | case word_UpperXref: |
989 | case word_LowerXref: |
138d7ffb |
990 | |
991 | if (text->type == word_HyperLink) { |
992 | dest.type = URL; |
993 | dest.url = utoa_dup(text->text); |
994 | dest.page = NULL; |
995 | } else { |
996 | keyword *kwl = kw_lookup(keywords, text->text); |
997 | para_data *pdata; |
998 | |
999 | if (kwl) { |
1000 | assert(kwl->para->private_data); |
1001 | pdata = (para_data *) kwl->para->private_data; |
1002 | dest.type = PAGE; |
1003 | dest.page = pdata->first->page; |
1004 | dest.url = NULL; |
1005 | } else { |
1006 | /* |
1007 | * Shouldn't happen, but *shrug* |
1008 | */ |
1009 | dest.type = NONE; |
1010 | dest.page = NULL; |
1011 | dest.url = NULL; |
1012 | } |
1013 | } |
1014 | if (dest.type != NONE) { |
1015 | *xr = mknew(xref); |
1016 | (*xr)->dest = dest; /* structure copy */ |
1017 | if (page->last_xref) |
1018 | page->last_xref->next = *xr; |
1019 | else |
1020 | page->first_xref = *xr; |
1021 | page->last_xref = *xr; |
1022 | |
1023 | /* |
1024 | * FIXME: Ideally we should have, and use, some |
1025 | * vertical font metric information here so that |
1026 | * our cross-ref rectangle can take account of |
1027 | * descenders and the font's cap height. This will |
1028 | * do for the moment, but it isn't ideal. |
1029 | */ |
1030 | (*xr)->lx = (*xr)->rx = x; |
1031 | (*xr)->by = y; |
1032 | (*xr)->ty = y + ldata->line_height; |
1033 | } |
1034 | goto nextword; |
1035 | |
1036 | /* |
1037 | * Finish extending a cross-reference box. |
1038 | */ |
1039 | case word_HyperEnd: |
43341922 |
1040 | case word_XrefEnd: |
138d7ffb |
1041 | *xr = NULL; |
1042 | goto nextword; |
1043 | |
43341922 |
1044 | case word_IndexRef: |
1045 | goto nextword; |
1046 | /* |
a8070b46 |
1047 | * FIXME: we should do something with this. |
43341922 |
1048 | */ |
1049 | } |
1050 | |
1051 | style = towordstyle(text->type); |
1052 | type = removeattr(text->type); |
1053 | |
1054 | findex = (style == word_Normal ? FONT_NORMAL : |
1055 | style == word_Emph ? FONT_EMPH : |
1056 | FONT_CODE); |
1057 | |
1058 | if (type == word_Normal) { |
1059 | str = text->text; |
1060 | } else if (type == word_WhiteSpace) { |
1061 | x += pdata->sizes[findex] * |
1062 | string_width(pdata->fonts[findex], L" ", NULL); |
faad4952 |
1063 | if (nspaces && findex != FONT_CODE) { |
1064 | x += (*nspace+1) * shortfall / nspaces; |
1065 | x -= *nspace * shortfall / nspaces; |
1066 | (*nspace)++; |
1067 | } |
43341922 |
1068 | goto nextword; |
1069 | } else /* if (type == word_Quote) */ { |
1070 | if (text->aux == quote_Open) |
1071 | str = L"\x2018"; /* FIXME: configurability! */ |
1072 | else |
1073 | str = L"\x2019"; /* FIXME: configurability! */ |
1074 | } |
1075 | |
1076 | (void) string_width(pdata->fonts[findex], str, &errs); |
1077 | |
1078 | if (errs && text->alt) |
138d7ffb |
1079 | x = render_text(page, pdata, ldata, x, y, text->alt, NULL, |
1080 | xr, shortfall, nspaces, nspace, keywords); |
43341922 |
1081 | else |
1082 | x = render_string(page, pdata->fonts[findex], |
1083 | pdata->sizes[findex], x, y, str); |
1084 | |
138d7ffb |
1085 | if (*xr) |
1086 | (*xr)->rx = x; |
1087 | |
43341922 |
1088 | nextword: |
43341922 |
1089 | text = text->next; |
1090 | } |
1091 | |
1092 | return x; |
1093 | } |
1094 | |
138d7ffb |
1095 | static void render_line(line_data *ldata, int left_x, int top_y, |
1096 | xref_dest *dest, keywordlist *keywords) |
43341922 |
1097 | { |
faad4952 |
1098 | int nspace; |
138d7ffb |
1099 | xref *xr; |
1100 | |
faad4952 |
1101 | if (ldata->aux_text) { |
515d216b |
1102 | int x; |
138d7ffb |
1103 | xr = NULL; |
faad4952 |
1104 | nspace = 0; |
515d216b |
1105 | x = render_text(ldata->page, ldata->pdata, ldata, |
1106 | left_x + ldata->aux_left_indent, |
1107 | top_y - ldata->ypos, |
1108 | ldata->aux_text, NULL, &xr, 0, 0, &nspace, keywords); |
1109 | if (ldata->aux_text_2) |
1110 | render_text(ldata->page, ldata->pdata, ldata, |
1111 | x, top_y - ldata->ypos, |
1112 | ldata->aux_text_2, NULL, &xr, 0, 0, &nspace, keywords); |
faad4952 |
1113 | } |
1114 | nspace = 0; |
138d7ffb |
1115 | |
1116 | /* |
1117 | * There might be a cross-reference carried over from a |
1118 | * previous line. |
1119 | */ |
1120 | if (dest->type != NONE) { |
1121 | xr = mknew(xref); |
1122 | xr->dest = *dest; /* structure copy */ |
1123 | if (ldata->page->last_xref) |
1124 | ldata->page->last_xref->next = xr; |
1125 | else |
1126 | ldata->page->first_xref = xr; |
1127 | ldata->page->last_xref = xr; |
1128 | xr->lx = xr->rx = left_x + ldata->xpos; |
1129 | xr->by = top_y - ldata->ypos; |
1130 | xr->ty = top_y - ldata->ypos + ldata->line_height; |
1131 | } else |
1132 | xr = NULL; |
1133 | |
1134 | render_text(ldata->page, ldata->pdata, ldata, left_x + ldata->xpos, |
1135 | top_y - ldata->ypos, ldata->first, ldata->end, &xr, |
1136 | ldata->hshortfall, ldata->nspaces, &nspace, keywords); |
1137 | |
1138 | if (xr) { |
1139 | /* |
1140 | * There's a cross-reference continued on to the next line. |
1141 | */ |
1142 | *dest = xr->dest; |
1143 | } else |
1144 | dest->type = NONE; |
43341922 |
1145 | } |
515d216b |
1146 | |
1147 | static void code_paragraph(para_data *pdata, |
1148 | font_data *fn, font_data *fi, font_data *fb, |
1149 | int font_size, int indent, word *words) |
1150 | { |
1151 | /* |
1152 | * For code paragraphs, I'm going to hack grievously and |
1153 | * pretend the three normal fonts are the three code paragraph |
1154 | * fonts. |
1155 | */ |
1156 | pdata->fonts[FONT_NORMAL] = fb; |
1157 | pdata->fonts[FONT_EMPH] = fi; |
1158 | pdata->fonts[FONT_CODE] = fn; |
1159 | pdata->sizes[FONT_NORMAL] = |
1160 | pdata->sizes[FONT_EMPH] = |
1161 | pdata->sizes[FONT_CODE] = font_size; |
1162 | |
1163 | pdata->first = pdata->last = NULL; |
1164 | |
1165 | for (; words; words = words->next) { |
1166 | wchar_t *t, *e, *start; |
1167 | word *lhead = NULL, *ltail = NULL, *w; |
1168 | line_data *ldata; |
1169 | int prev = -1, curr; |
1170 | |
1171 | t = words->text; |
1172 | if (words->next && words->next->type == word_Emph) { |
1173 | e = words->next->text; |
1174 | words = words->next; |
1175 | } else |
1176 | e = NULL; |
1177 | |
1178 | start = t; |
1179 | |
1180 | while (*start) { |
1181 | while (*t) { |
1182 | if (!e || !*e) |
1183 | curr = 0; |
1184 | else if (*e == L'i') |
1185 | curr = 1; |
1186 | else if (*e == L'b') |
1187 | curr = 2; |
1188 | else |
1189 | curr = 0; |
1190 | |
1191 | if (prev < 0) |
1192 | prev = curr; |
1193 | |
1194 | if (curr != prev) |
1195 | break; |
1196 | |
1197 | t++; |
1198 | if (e && *e) |
1199 | e++; |
1200 | } |
1201 | |
1202 | /* |
1203 | * We've isolated a maximal subsequence of the line |
1204 | * which has the same emphasis. Form it into a word |
1205 | * structure. |
1206 | */ |
1207 | w = mknew(word); |
1208 | w->next = NULL; |
1209 | w->alt = NULL; |
1210 | w->type = (prev == 0 ? word_WeakCode : |
1211 | prev == 1 ? word_Emph : word_Normal); |
1212 | w->text = mknewa(wchar_t, t-start+1); |
1213 | memcpy(w->text, start, (t-start) * sizeof(wchar_t)); |
1214 | w->text[t-start] = '\0'; |
1215 | w->breaks = FALSE; |
1216 | |
1217 | if (ltail) |
1218 | ltail->next = w; |
1219 | else |
1220 | lhead = w; |
1221 | ltail = w; |
1222 | |
1223 | start = t; |
1224 | prev = -1; |
1225 | } |
1226 | |
1227 | ldata = mknew(line_data); |
1228 | |
1229 | ldata->pdata = pdata; |
1230 | ldata->first = lhead; |
1231 | ldata->end = NULL; |
1232 | ldata->line_height = font_size * 4096; |
1233 | |
1234 | ldata->xpos = indent; |
1235 | |
1236 | if (pdata->last) { |
1237 | pdata->last->next = ldata; |
1238 | ldata->prev = pdata->last; |
1239 | } else { |
1240 | pdata->first = ldata; |
1241 | ldata->prev = NULL; |
1242 | } |
1243 | ldata->next = NULL; |
1244 | pdata->last = ldata; |
1245 | |
1246 | ldata->hshortfall = 0; |
1247 | ldata->nspaces = 0; |
1248 | ldata->aux_text = NULL; |
1249 | ldata->aux_text_2 = NULL; |
1250 | ldata->aux_left_indent = 0; |
1251 | |
1252 | } |
1253 | } |