78c73085 |
1 | /* |
2 | * HTML backend for Halibut |
3 | */ |
4 | |
5 | /* |
6 | * TODO: |
7 | * |
8 | * - I'm never entirely convinced that having a fragment link to |
9 | * come in at the start of the real text in the file is |
10 | * sensible. Perhaps for the topmost section in the file, no |
11 | * fragment should be used? (Though it should probably still be |
12 | * _there_ even if unused.) |
13 | * |
14 | * - new configurability: |
15 | * * a few new things explicitly labelled as `FIXME: |
16 | * configurable' or similar. |
17 | * * HTML flavour. |
18 | * * Some means of specifying the distinction between |
19 | * restrict-charset and output-charset. It seems to me that |
20 | * `html-charset' is output-charset, and that |
21 | * restrict-charset usually wants to be either output-charset |
22 | * or UTF-8 (the latter indicating that any Unicode character |
23 | * is fair game and it will be specified using &#foo; if it |
24 | * isn't in output-charset). However, since XHTML defaults to |
25 | * UTF-8 and it's fiddly to tell it otherwise, it's just |
26 | * possible that some user may need to set restrict-charset |
27 | * to their charset of choice while leaving _output_-charset |
28 | * at UTF-8. Figure out some configuration, and apply it. |
29 | * |
30 | * - test all HTML flavours and ensure they validate sensibly. Fix |
31 | * remaining confusion issues such as <?xml?> and obsoleteness |
32 | * of <a name>. |
33 | * |
34 | * - proper naming of all fragment IDs. The ones for sections are |
35 | * fine; the ones for numbered list and bibliociteds are utter |
36 | * crap; the ones for indexes _might_ do but it might be worth |
37 | * giving some thought to how to do them better. |
38 | * + also set up a mechanism for ensuring that fragment IDs |
39 | * never clash. |
40 | * |
41 | * - nonbreaking spaces? |
42 | */ |
43 | |
44 | #include <stdio.h> |
45 | #include <stdlib.h> |
46 | #include <assert.h> |
47 | #include <limits.h> |
48 | #include "halibut.h" |
49 | |
50 | #define is_heading_type(type) ( (type) == para_Title || \ |
51 | (type) == para_Chapter || \ |
52 | (type) == para_Appendix || \ |
53 | (type) == para_UnnumberedChapter || \ |
54 | (type) == para_Heading || \ |
55 | (type) == para_Subsect) |
56 | |
57 | #define heading_depth(p) ( (p)->type == para_Subsect ? (p)->aux + 1 : \ |
58 | (p)->type == para_Heading ? 1 : \ |
59 | (p)->type == para_Title ? -1 : 0 ) |
60 | |
61 | typedef struct { |
62 | int just_numbers; |
63 | wchar_t *number_suffix; |
64 | } sectlevel; |
65 | |
66 | typedef struct { |
67 | int nasect; |
68 | sectlevel achapter, *asect; |
69 | int *contents_depths; /* 0=main, 1=chapter, 2=sect etc */ |
70 | int ncdepths; |
71 | int address_section, visible_version_id; |
72 | int leaf_contains_contents, leaf_smallest_contents; |
73 | char *contents_filename; |
74 | char *index_filename; |
75 | char *template_filename; |
76 | char *single_filename; |
77 | char *template_fragment; |
78 | char *head_end, *body_start, *body_end, *addr_start, *addr_end; |
79 | char *body_tag, *nav_attr; |
80 | wchar_t *author, *description; |
81 | int restrict_charset, output_charset; |
82 | enum { |
83 | HTML_3_2, HTML_4, |
84 | XHTML_1_0_TRANSITIONAL, XHTML_1_0_STRICT |
85 | } htmlver; |
86 | wchar_t *lquote, *rquote; |
87 | int leaf_level; |
88 | } htmlconfig; |
89 | |
90 | #define contents_depth(conf, level) \ |
91 | ( (conf).ncdepths > (level) ? (conf).contents_depths[level] : (level)+2 ) |
92 | |
93 | #define is_xhtml(ver) ((ver) >= XHTML_1_0_TRANSITIONAL) |
94 | |
95 | typedef struct htmlfile htmlfile; |
96 | typedef struct htmlsect htmlsect; |
97 | |
98 | struct htmlfile { |
99 | htmlfile *next; |
100 | char *filename; |
101 | int last_fragment_number; |
102 | int min_heading_depth; |
103 | htmlsect *first, *last; /* first/last highest-level sections */ |
104 | }; |
105 | |
106 | struct htmlsect { |
107 | htmlsect *next, *parent; |
108 | htmlfile *file; |
109 | paragraph *title, *text; |
110 | enum { NORMAL, TOP, INDEX } type; |
111 | int contents_depth; |
112 | char *fragment; |
113 | }; |
114 | |
115 | typedef struct { |
116 | htmlfile *head, *tail; |
117 | htmlfile *single, *index; |
118 | } htmlfilelist; |
119 | |
120 | typedef struct { |
121 | htmlsect *head, *tail; |
122 | } htmlsectlist; |
123 | |
124 | typedef struct { |
125 | int nrefs, refsize; |
126 | word **refs; |
127 | } htmlindex; |
128 | |
129 | typedef struct { |
130 | htmlsect *section; |
131 | char *fragment; |
132 | } htmlindexref; |
133 | |
134 | typedef struct { |
135 | /* |
136 | * This level deals with charset conversion, starting and |
137 | * ending tags, and writing to the file. It's the lexical |
138 | * level. |
139 | */ |
140 | FILE *fp; |
141 | int charset; |
142 | charset_state cstate; |
143 | int ver; |
144 | enum { |
145 | HO_NEUTRAL, HO_IN_TAG, HO_IN_EMPTY_TAG, HO_IN_TEXT |
146 | } state; |
147 | /* |
148 | * Stuff beyond here deals with the higher syntactic level: it |
149 | * tracks how many levels of <ul> are currently open when |
150 | * producing a contents list, for example. |
151 | */ |
152 | int contents_level; |
153 | } htmloutput; |
154 | |
155 | static void html_file_section(htmlconfig *cfg, htmlfilelist *files, |
156 | htmlsect *sect, int depth); |
157 | |
158 | static htmlfile *html_new_file(htmlfilelist *list, char *filename); |
159 | static htmlsect *html_new_sect(htmlsectlist *list, paragraph *title); |
160 | |
161 | /* Flags for html_words() flags parameter */ |
162 | #define NOTHING 0x00 |
163 | #define MARKUP 0x01 |
164 | #define LINKS 0x02 |
165 | #define INDEXENTS 0x04 |
166 | #define ALL 0x07 |
167 | static void html_words(htmloutput *ho, word *words, int flags, |
168 | htmlfile *file, keywordlist *keywords, htmlconfig *cfg); |
169 | static void html_codepara(htmloutput *ho, word *words); |
170 | |
171 | static void element_open(htmloutput *ho, char const *name); |
172 | static void element_close(htmloutput *ho, char const *name); |
173 | static void element_empty(htmloutput *ho, char const *name); |
174 | static void element_attr(htmloutput *ho, char const *name, char const *value); |
175 | static void element_attr_w(htmloutput *ho, char const *name, |
176 | wchar_t const *value); |
177 | static void html_text(htmloutput *ho, wchar_t const *str); |
178 | static void html_text_limit(htmloutput *ho, wchar_t const *str, int maxlen); |
179 | static void html_text_limit_internal(htmloutput *ho, wchar_t const *text, |
180 | int maxlen, int quote_quotes); |
181 | static void html_nl(htmloutput *ho); |
182 | static void html_raw(htmloutput *ho, char *text); |
183 | static void html_raw_as_attr(htmloutput *ho, char *text); |
184 | static void cleanup(htmloutput *ho); |
185 | |
186 | static void html_href(htmloutput *ho, htmlfile *thisfile, |
187 | htmlfile *targetfile, char *targetfrag); |
188 | |
189 | static char *html_format(paragraph *p, char *template_string); |
190 | static void html_sanitise_fragment(char *text); |
191 | |
192 | static void html_contents_entry(htmloutput *ho, int depth, htmlsect *s, |
193 | htmlfile *thisfile, keywordlist *keywords, |
194 | htmlconfig *cfg); |
195 | static void html_section_title(htmloutput *ho, htmlsect *s, |
196 | htmlfile *thisfile, keywordlist *keywords, |
197 | htmlconfig *cfg); |
198 | |
199 | static htmlconfig html_configure(paragraph *source) { |
200 | htmlconfig ret; |
201 | paragraph *p; |
202 | |
203 | /* |
204 | * Defaults. |
205 | */ |
206 | ret.leaf_level = 2; |
207 | ret.achapter.just_numbers = FALSE; |
208 | ret.achapter.number_suffix = L": "; |
209 | ret.nasect = 1; |
210 | ret.asect = mknewa(sectlevel, ret.nasect); |
211 | ret.asect[0].just_numbers = TRUE; |
212 | ret.asect[0].number_suffix = L" "; |
213 | ret.ncdepths = 0; |
214 | ret.contents_depths = 0; |
215 | ret.visible_version_id = TRUE; |
216 | ret.address_section = TRUE; |
217 | ret.leaf_contains_contents = FALSE; |
218 | ret.leaf_smallest_contents = 4; |
219 | ret.single_filename = dupstr("Manual.html"); |
220 | ret.contents_filename = dupstr("Contents.html"); |
221 | ret.index_filename = dupstr("IndexPage.html"); |
222 | ret.template_filename = dupstr("%n.html"); |
223 | ret.template_fragment = dupstr("%b"); |
224 | ret.head_end = ret.body_tag = ret.body_start = ret.body_end = |
225 | ret.addr_start = ret.addr_end = ret.nav_attr = NULL; |
226 | ret.author = ret.description = NULL; |
227 | ret.restrict_charset = CS_ASCII; |
228 | ret.output_charset = CS_ASCII; |
229 | ret.htmlver = HTML_4; |
230 | /* |
231 | * Default quote characters are Unicode matched single quotes, |
232 | * falling back to ordinary ASCII ". |
233 | */ |
234 | ret.lquote = L"\x2018\0\x2019\0\"\0\"\0\0"; |
235 | ret.rquote = uadv(ret.lquote); |
236 | |
237 | /* |
238 | * Two-pass configuration so that we can pick up global config |
239 | * (e.g. `quotes') before having it overridden by specific |
240 | * config (`html-quotes'), irrespective of the order in which |
241 | * they occur. |
242 | */ |
243 | for (p = source; p; p = p->next) { |
244 | if (p->type == para_Config) { |
245 | if (!ustricmp(p->keyword, L"quotes")) { |
246 | if (*uadv(p->keyword) && *uadv(uadv(p->keyword))) { |
247 | ret.lquote = uadv(p->keyword); |
248 | ret.rquote = uadv(ret.lquote); |
249 | } |
250 | } |
251 | } |
252 | } |
253 | |
254 | for (p = source; p; p = p->next) { |
255 | if (p->type == para_Config) { |
256 | wchar_t *k = p->keyword; |
257 | |
258 | if (!ustrnicmp(k, L"xhtml-", 6)) |
259 | k++; /* treat `xhtml-' and `html-' the same */ |
260 | |
261 | if (!ustricmp(k, L"html-charset")) { |
262 | char *csname = utoa_dup(uadv(k), CS_ASCII); |
263 | ret.restrict_charset = ret.output_charset = |
264 | charset_from_localenc(csname); |
265 | sfree(csname); |
266 | } else if (!ustricmp(k, L"html-single-filename")) { |
267 | sfree(ret.single_filename); |
268 | ret.single_filename = dupstr(adv(p->origkeyword)); |
269 | } else if (!ustricmp(k, L"html-contents-filename")) { |
270 | sfree(ret.contents_filename); |
271 | ret.contents_filename = dupstr(adv(p->origkeyword)); |
272 | } else if (!ustricmp(k, L"html-index-filename")) { |
273 | sfree(ret.index_filename); |
274 | ret.index_filename = dupstr(adv(p->origkeyword)); |
275 | } else if (!ustricmp(k, L"html-template-filename")) { |
276 | sfree(ret.template_filename); |
277 | ret.template_filename = dupstr(adv(p->origkeyword)); |
278 | } else if (!ustricmp(k, L"html-template-fragment")) { |
279 | sfree(ret.template_fragment); |
280 | ret.template_fragment = dupstr(adv(p->origkeyword)); |
281 | } else if (!ustricmp(k, L"html-chapter-numeric")) { |
282 | ret.achapter.just_numbers = utob(uadv(k)); |
283 | } else if (!ustricmp(k, L"html-chapter-suffix")) { |
284 | ret.achapter.number_suffix = uadv(k); |
285 | } else if (!ustricmp(k, L"html-leaf-level")) { |
286 | ret.leaf_level = utoi(uadv(k)); |
287 | } else if (!ustricmp(k, L"html-section-numeric")) { |
288 | wchar_t *q = uadv(k); |
289 | int n = 0; |
290 | if (uisdigit(*q)) { |
291 | n = utoi(q); |
292 | q = uadv(q); |
293 | } |
294 | if (n >= ret.nasect) { |
295 | int i; |
296 | ret.asect = resize(ret.asect, n+1); |
297 | for (i = ret.nasect; i <= n; i++) |
298 | ret.asect[i] = ret.asect[ret.nasect-1]; |
299 | ret.nasect = n+1; |
300 | } |
301 | ret.asect[n].just_numbers = utob(q); |
302 | } else if (!ustricmp(k, L"html-section-suffix")) { |
303 | wchar_t *q = uadv(k); |
304 | int n = 0; |
305 | if (uisdigit(*q)) { |
306 | n = utoi(q); |
307 | q = uadv(q); |
308 | } |
309 | if (n >= ret.nasect) { |
310 | int i; |
311 | ret.asect = resize(ret.asect, n+1); |
312 | for (i = ret.nasect; i <= n; i++) { |
313 | ret.asect[i] = ret.asect[ret.nasect-1]; |
314 | } |
315 | ret.nasect = n+1; |
316 | } |
317 | ret.asect[n].number_suffix = q; |
318 | } else if (!ustricmp(k, L"html-contents-depth") || |
319 | !ustrnicmp(k, L"html-contents-depth-", 20)) { |
320 | /* |
321 | * Relic of old implementation: this directive used |
322 | * to be written as \cfg{html-contents-depth-3}{2} |
323 | * rather than the usual Halibut convention of |
324 | * \cfg{html-contents-depth}{3}{2}. We therefore |
325 | * support both. |
326 | */ |
327 | wchar_t *q = k[19] ? k+20 : uadv(k); |
328 | int n = 0; |
329 | if (uisdigit(*q)) { |
330 | n = utoi(q); |
331 | q = uadv(q); |
332 | } |
333 | if (n >= ret.ncdepths) { |
334 | int i; |
335 | ret.contents_depths = resize(ret.contents_depths, n+1); |
336 | for (i = ret.ncdepths; i <= n; i++) { |
337 | ret.contents_depths[i] = i+2; |
338 | } |
339 | ret.ncdepths = n+1; |
340 | } |
341 | ret.contents_depths[n] = utoi(q); |
342 | } else if (!ustricmp(k, L"html-head-end")) { |
343 | ret.head_end = adv(p->origkeyword); |
344 | } else if (!ustricmp(k, L"html-body-tag")) { |
345 | ret.body_tag = adv(p->origkeyword); |
346 | } else if (!ustricmp(k, L"html-body-start")) { |
347 | ret.body_start = adv(p->origkeyword); |
348 | } else if (!ustricmp(k, L"html-body-end")) { |
349 | ret.body_end = adv(p->origkeyword); |
350 | } else if (!ustricmp(k, L"html-address-start")) { |
351 | ret.addr_start = adv(p->origkeyword); |
352 | } else if (!ustricmp(k, L"html-address-end")) { |
353 | ret.addr_end = adv(p->origkeyword); |
354 | } else if (!ustricmp(k, L"html-navigation-attributes")) { |
355 | ret.nav_attr = adv(p->origkeyword); |
356 | } else if (!ustricmp(k, L"html-author")) { |
357 | ret.author = uadv(k); |
358 | } else if (!ustricmp(k, L"html-description")) { |
359 | ret.description = uadv(k); |
360 | } else if (!ustricmp(k, L"html-suppress-address")) { |
361 | ret.address_section = !utob(uadv(k)); |
362 | } else if (!ustricmp(k, L"html-versionid")) { |
363 | ret.visible_version_id = utob(uadv(k)); |
364 | } else if (!ustricmp(k, L"html-quotes")) { |
365 | if (*uadv(k) && *uadv(uadv(k))) { |
366 | ret.lquote = uadv(k); |
367 | ret.rquote = uadv(ret.lquote); |
368 | } |
369 | } else if (!ustricmp(k, L"html-leaf-contains-contents")) { |
370 | ret.leaf_contains_contents = utob(uadv(k)); |
371 | } else if (!ustricmp(k, L"html-leaf-smallest-contents")) { |
372 | ret.leaf_smallest_contents = utoi(uadv(k)); |
373 | } |
374 | } |
375 | } |
376 | |
377 | /* |
378 | * Now process fallbacks on quote characters. |
379 | */ |
380 | while (*uadv(ret.rquote) && *uadv(uadv(ret.rquote)) && |
381 | (!cvt_ok(ret.restrict_charset, ret.lquote) || |
382 | !cvt_ok(ret.restrict_charset, ret.rquote))) { |
383 | ret.lquote = uadv(ret.rquote); |
384 | ret.rquote = uadv(ret.lquote); |
385 | } |
386 | |
387 | return ret; |
388 | } |
389 | |
390 | paragraph *html_config_filename(char *filename) |
391 | { |
392 | /* |
393 | * If the user passes in a single filename as a parameter to |
394 | * the `--html' command-line option, then we should assume it |
395 | * to imply _two_ config directives: |
396 | * \cfg{html-single-filename}{whatever} and |
397 | * \cfg{html-leaf-level}{0}; the rationale being that the user |
398 | * wants their output _in that file_. |
399 | */ |
400 | paragraph *p, *q; |
401 | |
402 | p = cmdline_cfg_simple("html-single-filename", filename, NULL); |
403 | q = cmdline_cfg_simple("html-leaf-level", "0", NULL); |
404 | p->next = q; |
405 | return p; |
406 | } |
407 | |
408 | void html_backend(paragraph *sourceform, keywordlist *keywords, |
409 | indexdata *idx, void *unused) { |
410 | paragraph *p; |
411 | htmlconfig conf; |
412 | htmlfilelist files = { NULL, NULL, NULL, NULL }; |
413 | htmlsectlist sects = { NULL, NULL }, nonsects = { NULL, NULL }; |
414 | |
415 | IGNORE(unused); |
416 | |
417 | conf = html_configure(sourceform); |
418 | |
419 | /* |
420 | * We're going to make heavy use of paragraphs' private data |
421 | * fields in the forthcoming code. Clear them first, so we can |
422 | * reliably tell whether we have auxiliary data for a |
423 | * particular paragraph. |
424 | */ |
425 | for (p = sourceform; p; p = p->next) |
426 | p->private_data = NULL; |
427 | |
428 | /* |
429 | * Start by figuring out into which file each piece of the |
430 | * document should be put. We'll do this by inventing an |
431 | * `htmlsect' structure and stashing it in the private_data |
432 | * field of each section paragraph; we also need one additional |
433 | * htmlsect for the document index, which won't show up in the |
434 | * source form but needs to be consistently mentioned in |
435 | * contents links. |
436 | * |
437 | * While we're here, we'll also invent the HTML fragment name |
438 | * for each section. |
439 | */ |
440 | { |
441 | htmlsect *topsect, *sect; |
442 | int d; |
443 | |
444 | topsect = html_new_sect(§s, p); |
445 | topsect->type = TOP; |
446 | topsect->title = NULL; |
447 | topsect->text = sourceform; |
448 | topsect->contents_depth = contents_depth(conf, 0); |
449 | html_file_section(&conf, &files, topsect, -1); |
450 | topsect->fragment = NULL; |
451 | |
452 | for (p = sourceform; p; p = p->next) |
453 | if (is_heading_type(p->type)) { |
454 | d = heading_depth(p); |
455 | |
456 | if (p->type == para_Title) { |
457 | topsect->title = p; |
458 | continue; |
459 | } |
460 | |
461 | sect = html_new_sect(§s, p); |
462 | sect->text = p->next; |
463 | |
464 | sect->contents_depth = contents_depth(conf, d+1) - (d+1); |
465 | |
466 | if (p->parent) { |
467 | sect->parent = (htmlsect *)p->parent->private_data; |
468 | assert(sect->parent != NULL); |
469 | } else |
470 | sect->parent = topsect; |
471 | p->private_data = sect; |
472 | |
473 | html_file_section(&conf, &files, sect, d); |
474 | |
475 | sect->fragment = html_format(p, conf.template_fragment); |
476 | html_sanitise_fragment(sect->fragment); |
477 | /* FIXME: clash checking? add to a tree of (file,frag)? */ |
478 | } |
479 | |
480 | /* And the index. */ |
481 | sect = html_new_sect(§s, NULL); |
482 | sect->fragment = dupstr("Index"); /* FIXME: this _can't_ be right */ |
483 | sect->text = NULL; |
484 | sect->type = INDEX; |
485 | sect->parent = topsect; |
486 | html_file_section(&conf, &files, sect, 0); /* peer of chapters */ |
487 | files.index = sect->file; |
488 | } |
489 | |
490 | /* |
491 | * Go through the keyword list and sort out fragment IDs for |
492 | * all the potentially referenced paragraphs which _aren't_ |
493 | * headings. |
494 | */ |
495 | { |
496 | int i; |
497 | keyword *kw; |
498 | htmlsect *sect; |
499 | |
500 | for (i = 0; (kw = index234(keywords->keys, i)) != NULL; i++) { |
501 | paragraph *q, *p = kw->para; |
502 | |
503 | if (!is_heading_type(p->type)) { |
504 | htmlsect *parent; |
505 | |
506 | /* |
507 | * Find the paragraph's parent htmlsect, to |
508 | * determine which file it will end up in. |
509 | */ |
510 | q = p->parent; |
511 | if (!q) { |
512 | /* |
513 | * Preamble paragraphs have no parent. So if we |
514 | * have a non-heading with no parent, it must |
515 | * be preamble, and therefore its parent |
516 | * htmlsect must be the preamble one. |
517 | */ |
518 | assert(sects.head && |
519 | sects.head->type == TOP); |
520 | parent = sects.head; |
521 | } else |
522 | parent = (htmlsect *)q->private_data; |
523 | |
524 | /* |
525 | * Now we can construct an htmlsect for this |
526 | * paragraph itself, taking care to put it in the |
527 | * list of non-sections rather than the list of |
528 | * sections (so that traverses of the `sects' list |
529 | * won't attempt to add it to the contents or |
530 | * anything weird like that). |
531 | */ |
532 | sect = html_new_sect(&nonsects, p); |
533 | sect->file = parent->file; |
534 | sect->parent = parent; |
535 | p->private_data = sect; |
536 | |
537 | /* |
538 | * FIXME: We need a much better means of naming |
539 | * these, possibly involving an additional |
540 | * configuration template. For the moment I'll just |
541 | * invent something completely stupid. |
542 | */ |
543 | sect->fragment = mknewa(char, 40); |
544 | sprintf(sect->fragment, "frag%p", sect); |
545 | } |
546 | } |
547 | } |
548 | |
549 | /* |
550 | * Now sort out the index. This involves: |
551 | * |
552 | * - For each index term, we set up an htmlindex structure to |
553 | * store all the references to that term. |
554 | * |
555 | * - Then we make a pass over the actual document, finding |
556 | * every word_IndexRef; for each one, we actually figure out |
557 | * the HTML filename/fragment pair we will use to reference |
558 | * it, store that information in the private data field of |
559 | * the word_IndexRef itself (so we can recreate it when the |
560 | * time comes to output our HTML), and add a reference to it |
561 | * to the index term in question. |
562 | */ |
563 | { |
564 | int i; |
565 | indexentry *entry; |
566 | htmlsect *lastsect; |
567 | word *w; |
568 | |
569 | /* |
570 | * Set up the htmlindex structures. |
571 | */ |
572 | |
573 | for (i = 0; (entry = index234(idx->entries, i)) != NULL; i++) { |
574 | htmlindex *hi = mknew(htmlindex); |
575 | |
576 | hi->nrefs = hi->refsize = 0; |
577 | hi->refs = NULL; |
578 | |
579 | entry->backend_data = hi; |
580 | } |
581 | |
582 | /* |
583 | * Run over the document inventing fragments. Each fragment |
584 | * is of the form `i' followed by an integer. |
585 | * |
586 | * FIXME: Probably in the file-organisation pass we should |
587 | * work out the fragment names of every section, so that we |
588 | * could load them all into a tree and hence ensure these |
589 | * index fragments don't clash with them. |
590 | */ |
591 | lastsect = NULL; |
592 | for (p = sourceform; p; p = p->next) { |
593 | if (is_heading_type(p->type)) |
594 | lastsect = (htmlsect *)p->private_data; |
595 | |
596 | for (w = p->words; w; w = w->next) |
597 | if (w->type == word_IndexRef) { |
598 | htmlindexref *hr = mknew(htmlindexref); |
599 | indextag *tag; |
600 | int i; |
601 | |
602 | hr->section = lastsect; |
603 | /* FIXME: clash checking */ |
604 | { |
605 | char buf[40]; |
606 | sprintf(buf, "i%d", |
607 | lastsect->file->last_fragment_number++); |
608 | hr->fragment = dupstr(buf); |
609 | } |
610 | w->private_data = hr; |
611 | |
612 | tag = index_findtag(idx, w->text); |
613 | if (!tag) |
614 | break; |
615 | |
616 | for (i = 0; i < tag->nrefs; i++) { |
617 | indexentry *entry = tag->refs[i]; |
618 | htmlindex *hi = (htmlindex *)entry->backend_data; |
619 | |
620 | if (hi->nrefs >= hi->refsize) { |
621 | hi->refsize += 32; |
622 | hi->refs = resize(hi->refs, hi->refsize); |
623 | } |
624 | |
625 | hi->refs[hi->nrefs++] = w; |
626 | } |
627 | } |
628 | } |
629 | } |
630 | |
631 | /* |
632 | * Now we're ready to write out the actual HTML files. |
633 | * |
634 | * For each file: |
635 | * |
636 | * - we open that file and write its header |
637 | * - we run down the list of sections |
638 | * - for each section directly contained within that file, we |
639 | * output the section text |
640 | * - for each section which is not in the file but which has a |
641 | * parent that is, we output a contents entry for the |
642 | * section if appropriate |
643 | * - finally, we output the file trailer and close the file. |
644 | */ |
645 | { |
646 | htmlfile *f, *prevf; |
647 | htmlsect *s; |
648 | paragraph *p; |
649 | |
650 | prevf = NULL; |
651 | |
652 | for (f = files.head; f; f = f->next) { |
653 | htmloutput ho; |
654 | int displaying; |
655 | enum LISTTYPE { NOLIST, UL, OL, DL }; |
656 | enum ITEMTYPE { NOITEM, LI, DT, DD }; |
657 | struct stackelement { |
658 | struct stackelement *next; |
659 | enum LISTTYPE listtype; |
660 | enum ITEMTYPE itemtype; |
661 | } *stackhead; |
662 | |
663 | #define listname(lt) ( (lt)==UL ? "ul" : (lt)==OL ? "ol" : "dl" ) |
664 | #define itemname(lt) ( (lt)==LI ? "li" : (lt)==DT ? "dt" : "dd" ) |
665 | |
666 | ho.fp = fopen(f->filename, "w"); |
667 | ho.charset = conf.output_charset; |
668 | ho.cstate = charset_init_state; |
669 | ho.ver = conf.htmlver; |
670 | ho.state = HO_NEUTRAL; |
671 | ho.contents_level = 0; |
672 | |
673 | /* <!DOCTYPE>. */ |
674 | switch (conf.htmlver) { |
675 | case HTML_3_2: |
676 | fprintf(ho.fp, "<!DOCTYPE HTML PUBLIC \"-//W3C//DTD " |
677 | "HTML 3.2 Final//EN\">\n"); |
678 | break; |
679 | case HTML_4: |
680 | fprintf(ho.fp, "<!DOCTYPE HTML PUBLIC \"-//W3C//DTD HTML" |
681 | " 4.01//EN\"\n\"http://www.w3.org/TR/html4/" |
682 | "strict.dtd\">\n"); |
683 | break; |
684 | case XHTML_1_0_TRANSITIONAL: |
685 | /* FIXME: <?xml?> to specify character encoding. |
686 | * This breaks HTML backwards compat, so perhaps avoid, or |
687 | * perhaps only emit when not using the default UTF-8? */ |
688 | fprintf(ho.fp, "<!DOCTYPE html PUBLIC \"-//W3C//DTD XHTML" |
689 | " 1.0 Transitional//EN\"\n\"http://www.w3.org/TR/" |
690 | "xhtml1/DTD/xhtml1-transitional.dtd\">\n"); |
691 | break; |
692 | case XHTML_1_0_STRICT: |
693 | /* FIXME: <?xml?> to specify character encoding. */ |
694 | fprintf(ho.fp, "<!DOCTYPE html PUBLIC \"-//W3C//DTD XHTML" |
695 | " 1.0 Strict//EN\"\n\"http://www.w3.org/TR/xhtml1/" |
696 | "DTD/xhtml1-strict.dtd\">\n"); |
697 | break; |
698 | } |
699 | |
700 | element_open(&ho, "html"); |
701 | if (is_xhtml(conf.htmlver)) { |
702 | element_attr(&ho, "xmlns", "http://www.w3.org/1999/xhtml"); |
703 | } |
704 | html_nl(&ho); |
705 | |
706 | element_open(&ho, "head"); |
707 | html_nl(&ho); |
708 | |
709 | element_empty(&ho, "meta"); |
710 | element_attr(&ho, "http-equiv", "content-type"); |
711 | { |
712 | char buf[200]; |
713 | sprintf(buf, "text/html; charset=%.150s", |
714 | charset_to_mimeenc(conf.output_charset)); |
715 | element_attr(&ho, "content", buf); |
716 | } |
717 | html_nl(&ho); |
718 | |
719 | if (conf.author) { |
720 | element_empty(&ho, "meta"); |
721 | element_attr(&ho, "name", "author"); |
722 | element_attr_w(&ho, "content", conf.author); |
723 | html_nl(&ho); |
724 | } |
725 | |
726 | if (conf.description) { |
727 | element_empty(&ho, "meta"); |
728 | element_attr(&ho, "name", "description"); |
729 | element_attr_w(&ho, "content", conf.description); |
730 | html_nl(&ho); |
731 | } |
732 | |
733 | element_open(&ho, "title"); |
734 | if (f->first && f->first->title) { |
735 | html_words(&ho, f->first->title->words, NOTHING, |
736 | f, keywords, &conf); |
737 | |
738 | assert(f->last); |
739 | if (f->last != f->first && f->last->title) { |
740 | html_text(&ho, L" - "); /* FIXME: configurable? */ |
741 | html_words(&ho, f->last->title->words, NOTHING, |
742 | f, keywords, &conf); |
743 | } |
744 | } |
745 | element_close(&ho, "title"); |
746 | html_nl(&ho); |
747 | |
748 | if (conf.head_end) |
749 | html_raw(&ho, conf.head_end); |
750 | |
751 | element_close(&ho, "head"); |
752 | html_nl(&ho); |
753 | |
754 | /* FIXME: need to be able to specify replacement for this */ |
755 | if (conf.body_tag) |
756 | html_raw(&ho, conf.body_tag); |
757 | else |
758 | element_open(&ho, "body"); |
759 | html_nl(&ho); |
760 | |
761 | if (conf.body_start) |
762 | html_raw(&ho, conf.body_start); |
763 | |
764 | /* |
765 | * Write out a nav bar. Special case: we don't do this |
766 | * if there is only one file. |
767 | */ |
768 | if (files.head != files.tail) { |
769 | element_open(&ho, "p"); |
770 | if (conf.nav_attr) |
771 | html_raw_as_attr(&ho, conf.nav_attr); |
772 | |
773 | if (prevf) { |
774 | element_open(&ho, "a"); |
775 | element_attr(&ho, "href", prevf->filename); |
776 | } |
777 | html_text(&ho, L"Previous");/* FIXME: conf? */ |
778 | if (prevf) |
779 | element_close(&ho, "a"); |
780 | |
781 | html_text(&ho, L" | "); /* FIXME: conf? */ |
782 | |
783 | if (f != files.head) { |
784 | element_open(&ho, "a"); |
785 | element_attr(&ho, "href", files.head->filename); |
786 | } |
787 | html_text(&ho, L"Contents");/* FIXME: conf? */ |
788 | if (f != files.head) |
789 | element_close(&ho, "a"); |
790 | |
791 | html_text(&ho, L" | "); /* FIXME: conf? */ |
792 | |
793 | if (f != files.index) { |
794 | element_open(&ho, "a"); |
795 | element_attr(&ho, "href", files.index->filename); |
796 | } |
797 | html_text(&ho, L"Index");/* FIXME: conf? */ |
798 | if (f != files.index) |
799 | element_close(&ho, "a"); |
800 | |
801 | html_text(&ho, L" | "); /* FIXME: conf? */ |
802 | |
803 | if (f->next) { |
804 | element_open(&ho, "a"); |
805 | element_attr(&ho, "href", f->next->filename); |
806 | } |
807 | html_text(&ho, L"Next"); /* FIXME: conf? */ |
808 | if (f->next) |
809 | element_close(&ho, "a"); |
810 | |
811 | element_close(&ho, "p"); |
812 | html_nl(&ho); |
813 | } |
814 | prevf = f; |
815 | |
816 | /* |
817 | * Write out a prefix TOC for the file. |
818 | * |
819 | * We start by going through the section list and |
820 | * collecting the sections which need to be added to |
821 | * the contents. On the way, we also test to see if |
822 | * this file is a leaf file (defined as one which |
823 | * contains all descendants of any section it |
824 | * contains), because this will play a part in our |
825 | * decision on whether or not to _output_ the TOC. |
826 | * |
827 | * Special case: we absolutely do not do this if we're |
828 | * in single-file mode. |
829 | */ |
830 | if (files.head != files.tail) { |
831 | int ntoc = 0, tocsize = 0; |
832 | htmlsect **toc = NULL; |
833 | int leaf = TRUE; |
834 | |
835 | for (s = sects.head; s; s = s->next) { |
836 | htmlsect *a, *ac; |
837 | int depth, adepth; |
838 | |
839 | /* |
840 | * Search up from this section until we find |
841 | * the highest-level one which belongs in this |
842 | * file. |
843 | */ |
844 | depth = adepth = 0; |
845 | a = NULL; |
846 | for (ac = s; ac; ac = ac->parent) { |
847 | if (ac->file == f) { |
848 | a = ac; |
849 | adepth = depth; |
850 | } |
851 | depth++; |
852 | } |
853 | |
854 | if (s->file != f && a != NULL) |
855 | leaf = FALSE; |
856 | |
857 | if (a) { |
858 | if (adepth <= a->contents_depth) { |
859 | if (ntoc >= tocsize) { |
860 | tocsize += 64; |
861 | toc = resize(toc, tocsize); |
862 | } |
863 | toc[ntoc++] = s; |
864 | } |
865 | } |
866 | } |
867 | |
868 | if (leaf && conf.leaf_contains_contents && |
869 | ntoc >= conf.leaf_smallest_contents) { |
870 | int i; |
871 | |
872 | for (i = 0; i < ntoc; i++) { |
873 | htmlsect *s = toc[i]; |
874 | int hlevel = (s->type == TOP ? -1 : |
875 | s->type == INDEX ? 0 : |
876 | heading_depth(s->title)) |
877 | - f->min_heading_depth + 1; |
878 | |
879 | assert(hlevel >= 1); |
880 | html_contents_entry(&ho, hlevel, s, |
881 | f, keywords, &conf); |
882 | } |
883 | html_contents_entry(&ho, 0, NULL, f, keywords, &conf); |
884 | } |
885 | } |
886 | |
887 | /* |
888 | * Now go through the document and output some real |
889 | * text. |
890 | */ |
891 | displaying = FALSE; |
892 | for (s = sects.head; s; s = s->next) { |
893 | if (s->file == f) { |
894 | /* |
895 | * This section belongs in this file. |
896 | * Display it. |
897 | */ |
898 | displaying = TRUE; |
899 | } else { |
900 | htmlsect *a, *ac; |
901 | int depth, adepth; |
902 | |
903 | displaying = FALSE; |
904 | |
905 | /* |
906 | * Search up from this section until we find |
907 | * the highest-level one which belongs in this |
908 | * file. |
909 | */ |
910 | depth = adepth = 0; |
911 | a = NULL; |
912 | for (ac = s; ac; ac = ac->parent) { |
913 | if (ac->file == f) { |
914 | a = ac; |
915 | adepth = depth; |
916 | } |
917 | depth++; |
918 | } |
919 | |
920 | if (a != NULL) { |
921 | /* |
922 | * This section does not belong in this |
923 | * file, but an ancestor of it does. Write |
924 | * out a contents table entry, if the depth |
925 | * doesn't exceed the maximum contents |
926 | * depth for the ancestor section. |
927 | */ |
928 | if (adepth <= a->contents_depth) { |
929 | html_contents_entry(&ho, adepth, s, |
930 | f, keywords, &conf); |
931 | } |
932 | } |
933 | } |
934 | |
935 | if (displaying) { |
936 | int hlevel; |
937 | char htag[3]; |
938 | |
939 | html_contents_entry(&ho, 0, NULL, f, keywords, &conf); |
940 | |
941 | /* |
942 | * Display the section heading. |
943 | */ |
944 | |
945 | hlevel = (s->type == TOP ? -1 : |
946 | s->type == INDEX ? 0 : |
947 | heading_depth(s->title)) |
948 | - f->min_heading_depth + 1; |
949 | assert(hlevel >= 1); |
950 | /* HTML headings only go up to <h6> */ |
951 | if (hlevel > 6) |
952 | hlevel = 6; |
953 | htag[0] = 'h'; |
954 | htag[1] = '0' + hlevel; |
955 | htag[2] = '\0'; |
956 | element_open(&ho, htag); |
957 | |
958 | /* |
959 | * Provide anchor for cross-links to target. |
960 | * |
961 | * FIXME: AIcurrentlyUI, this needs to be done |
962 | * differently in XHTML because <a name> is |
963 | * deprecated or obsolete. |
964 | * |
965 | * (Also we'll have to do this separately in |
966 | * other paragraph types - NumberedList and |
967 | * BiblioCited.) |
968 | */ |
969 | element_open(&ho, "a"); |
970 | element_attr(&ho, "name", s->fragment); |
971 | element_close(&ho, "a"); |
972 | |
973 | html_section_title(&ho, s, f, keywords, &conf); |
974 | |
975 | element_close(&ho, htag); |
976 | |
977 | /* |
978 | * Now display the section text. |
979 | */ |
980 | if (s->text) { |
981 | stackhead = mknew(struct stackelement); |
982 | stackhead->next = NULL; |
983 | stackhead->listtype = NOLIST; |
984 | stackhead->itemtype = NOITEM; |
985 | |
986 | for (p = s->text;; p = p->next) { |
987 | enum LISTTYPE listtype; |
988 | struct stackelement *se; |
989 | |
990 | /* |
991 | * Preliminary switch to figure out what |
992 | * sort of list we expect to be inside at |
993 | * this stage. |
994 | * |
995 | * Since p may still be NULL at this point, |
996 | * I invent a harmless paragraph type for |
997 | * it if it is. |
998 | */ |
999 | switch (p ? p->type : para_Normal) { |
1000 | case para_Rule: |
1001 | case para_Normal: |
1002 | case para_Copyright: |
1003 | case para_BiblioCited: |
1004 | case para_Code: |
1005 | case para_QuotePush: |
1006 | case para_QuotePop: |
1007 | case para_Chapter: |
1008 | case para_Appendix: |
1009 | case para_UnnumberedChapter: |
1010 | case para_Heading: |
1011 | case para_Subsect: |
1012 | case para_LcontPop: |
1013 | listtype = NOLIST; |
1014 | break; |
1015 | |
1016 | case para_Bullet: |
1017 | listtype = UL; |
1018 | break; |
1019 | |
1020 | case para_NumberedList: |
1021 | listtype = OL; |
1022 | break; |
1023 | |
1024 | case para_DescribedThing: |
1025 | case para_Description: |
1026 | listtype = DL; |
1027 | break; |
1028 | |
1029 | case para_LcontPush: |
1030 | se = mknew(struct stackelement); |
1031 | se->next = stackhead; |
1032 | se->listtype = NOLIST; |
1033 | se->itemtype = NOITEM; |
1034 | stackhead = se; |
1035 | continue; |
1036 | |
1037 | default: /* some totally non-printing para */ |
1038 | continue; |
1039 | } |
1040 | |
1041 | html_nl(&ho); |
1042 | |
1043 | /* |
1044 | * Terminate the most recent list item, if |
1045 | * any. (We left this until after |
1046 | * processing LcontPush, since in that case |
1047 | * the list item won't want to be |
1048 | * terminated until after the corresponding |
1049 | * LcontPop.) |
1050 | */ |
1051 | if (stackhead->itemtype != NOITEM) { |
1052 | element_close(&ho, itemname(stackhead->itemtype)); |
1053 | html_nl(&ho); |
1054 | } |
1055 | stackhead->itemtype = NOITEM; |
1056 | |
1057 | /* |
1058 | * Terminate the current list, if it's not |
1059 | * the one we want to be in. |
1060 | */ |
1061 | if (listtype != stackhead->listtype && |
1062 | stackhead->listtype != NOLIST) { |
1063 | element_close(&ho, listname(stackhead->listtype)); |
1064 | html_nl(&ho); |
1065 | } |
1066 | |
1067 | /* |
1068 | * Leave the loop if our time has come. |
1069 | */ |
1070 | if (!p || (is_heading_type(p->type) && |
1071 | p->type != para_Title)) |
1072 | break; /* end of section text */ |
1073 | |
1074 | /* |
1075 | * Start a fresh list if necessary. |
1076 | */ |
1077 | if (listtype != stackhead->listtype && |
1078 | listtype != NOLIST) |
1079 | element_open(&ho, listname(listtype)); |
1080 | |
1081 | stackhead->listtype = listtype; |
1082 | |
1083 | switch (p->type) { |
1084 | case para_Rule: |
1085 | element_empty(&ho, "hr"); |
1086 | break; |
1087 | case para_Code: |
1088 | html_codepara(&ho, p->words); |
1089 | break; |
1090 | case para_Normal: |
1091 | case para_Copyright: |
1092 | element_open(&ho, "p"); |
1093 | html_nl(&ho); |
1094 | html_words(&ho, p->words, ALL, |
1095 | f, keywords, &conf); |
1096 | html_nl(&ho); |
1097 | element_close(&ho, "p"); |
1098 | break; |
1099 | case para_BiblioCited: |
1100 | element_open(&ho, "p"); |
1101 | if (p->private_data) { |
1102 | htmlsect *s = (htmlsect *)p->private_data; |
1103 | element_open(&ho, "a"); |
1104 | element_attr(&ho, "name", s->fragment); |
1105 | element_close(&ho, "a"); |
1106 | } |
1107 | html_nl(&ho); |
1108 | html_words(&ho, p->kwtext, ALL, |
1109 | f, keywords, &conf); |
1110 | html_text(&ho, L" "); |
1111 | html_words(&ho, p->words, ALL, |
1112 | f, keywords, &conf); |
1113 | html_nl(&ho); |
1114 | element_close(&ho, "p"); |
1115 | break; |
1116 | case para_Bullet: |
1117 | case para_NumberedList: |
1118 | element_open(&ho, "li"); |
1119 | if (p->private_data) { |
1120 | htmlsect *s = (htmlsect *)p->private_data; |
1121 | element_open(&ho, "a"); |
1122 | element_attr(&ho, "name", s->fragment); |
1123 | element_close(&ho, "a"); |
1124 | } |
1125 | html_nl(&ho); |
1126 | stackhead->itemtype = LI; |
1127 | html_words(&ho, p->words, ALL, |
1128 | f, keywords, &conf); |
1129 | break; |
1130 | case para_DescribedThing: |
1131 | element_open(&ho, "dt"); |
1132 | html_nl(&ho); |
1133 | stackhead->itemtype = DT; |
1134 | html_words(&ho, p->words, ALL, |
1135 | f, keywords, &conf); |
1136 | break; |
1137 | case para_Description: |
1138 | element_open(&ho, "dd"); |
1139 | html_nl(&ho); |
1140 | stackhead->itemtype = DD; |
1141 | html_words(&ho, p->words, ALL, |
1142 | f, keywords, &conf); |
1143 | break; |
1144 | |
1145 | case para_QuotePush: |
1146 | element_open(&ho, "blockquote"); |
1147 | break; |
1148 | case para_QuotePop: |
1149 | element_close(&ho, "blockquote"); |
1150 | break; |
1151 | |
1152 | case para_LcontPop: |
1153 | se = stackhead; |
1154 | stackhead = stackhead->next; |
1155 | assert(stackhead); |
1156 | sfree(se); |
1157 | break; |
1158 | } |
1159 | } |
1160 | |
1161 | assert(stackhead && !stackhead->next); |
1162 | sfree(stackhead); |
1163 | } |
1164 | |
1165 | if (s->type == INDEX) { |
1166 | indexentry *entry; |
1167 | int i; |
1168 | |
1169 | /* |
1170 | * This section is the index. I'll just |
1171 | * render it as a single paragraph, with a |
1172 | * colon between the index term and the |
1173 | * references, and <br> in between each |
1174 | * entry. |
1175 | */ |
1176 | element_open(&ho, "p"); |
1177 | |
1178 | for (i = 0; (entry = |
1179 | index234(idx->entries, i)) != NULL; i++) { |
1180 | htmlindex *hi = (htmlindex *)entry->backend_data; |
1181 | int j; |
1182 | |
1183 | if (i > 0) |
1184 | element_empty(&ho, "br"); |
1185 | html_nl(&ho); |
1186 | |
1187 | html_words(&ho, entry->text, MARKUP|LINKS, |
1188 | f, keywords, &conf); |
1189 | |
1190 | html_text(&ho, L": ");/* FIXME: configurable */ |
1191 | |
1192 | for (j = 0; j < hi->nrefs; j++) { |
1193 | htmlindexref *hr = |
1194 | (htmlindexref *)hi->refs[j]->private_data; |
1195 | paragraph *p = hr->section->title; |
1196 | |
1197 | if (j > 0) |
1198 | html_text(&ho, L", "); /* FIXME: conf */ |
1199 | |
1200 | html_href(&ho, f, hr->section->file, |
1201 | hr->fragment); |
1202 | if (p && p->kwtext) |
1203 | html_words(&ho, p->kwtext, MARKUP|LINKS, |
1204 | f, keywords, &conf); |
1205 | else if (p && p->words) |
1206 | html_words(&ho, p->words, MARKUP|LINKS, |
1207 | f, keywords, &conf); |
1208 | else |
1209 | html_text(&ho, L"FIXME"); |
1210 | element_close(&ho, "a"); |
1211 | } |
1212 | } |
1213 | element_close(&ho, "p"); |
1214 | } |
1215 | } |
1216 | } |
1217 | |
1218 | html_contents_entry(&ho, 0, NULL, f, keywords, &conf); |
1219 | html_nl(&ho); |
1220 | |
1221 | { |
1222 | /* |
1223 | * Footer. |
1224 | */ |
1225 | int done_version_ids = FALSE; |
1226 | |
1227 | element_empty(&ho, "hr"); |
1228 | |
1229 | if (conf.body_end) |
1230 | html_raw(&ho, conf.body_end); |
1231 | |
1232 | if (conf.address_section) { |
1233 | element_open(&ho, "address"); |
1234 | if (conf.addr_start) { |
1235 | html_raw(&ho, conf.addr_start); |
1236 | html_nl(&ho); |
1237 | } |
1238 | if (conf.visible_version_id) { |
1239 | int started = FALSE; |
1240 | for (p = sourceform; p; p = p->next) |
1241 | if (p->type == para_VersionID) { |
1242 | if (!started) |
1243 | element_open(&ho, "p"); |
1244 | else |
1245 | element_empty(&ho, "br"); |
1246 | html_nl(&ho); |
1247 | html_text(&ho, L"["); /* FIXME: conf? */ |
1248 | html_words(&ho, p->words, NOTHING, |
1249 | f, keywords, &conf); |
1250 | html_text(&ho, L"]"); /* FIXME: conf? */ |
1251 | started = TRUE; |
1252 | } |
1253 | if (started) |
1254 | element_close(&ho, "p"); |
1255 | done_version_ids = TRUE; |
1256 | } |
1257 | if (conf.addr_end) |
1258 | html_raw(&ho, conf.addr_end); |
1259 | element_close(&ho, "address"); |
1260 | } |
1261 | |
1262 | if (!done_version_ids) { |
1263 | /* |
1264 | * If the user didn't want the version IDs |
1265 | * visible, I think we still have a duty to put |
1266 | * them in an HTML comment. |
1267 | */ |
1268 | int started = FALSE; |
1269 | for (p = sourceform; p; p = p->next) |
1270 | if (p->type == para_VersionID) { |
1271 | if (!started) { |
1272 | html_raw(&ho, "<!-- version IDs:\n"); |
1273 | started = TRUE; |
1274 | } |
1275 | html_words(&ho, p->words, NOTHING, |
1276 | f, keywords, &conf); |
1277 | html_nl(&ho); |
1278 | } |
1279 | if (started) |
1280 | html_raw(&ho, "-->\n"); |
1281 | } |
1282 | } |
1283 | |
1284 | element_close(&ho, "body"); |
1285 | html_nl(&ho); |
1286 | element_close(&ho, "html"); |
1287 | html_nl(&ho); |
1288 | cleanup(&ho); |
1289 | } |
1290 | } |
1291 | |
1292 | /* |
1293 | * FIXME: Figure out a way to free the htmlindex and |
1294 | * htmlindexref structures. |
1295 | */ |
1296 | } |
1297 | |
1298 | static void html_file_section(htmlconfig *cfg, htmlfilelist *files, |
1299 | htmlsect *sect, int depth) |
1300 | { |
1301 | htmlfile *file; |
1302 | int ldepth; |
1303 | |
1304 | /* |
1305 | * `depth' is derived from the heading_depth() macro at the top |
1306 | * of this file, which counts title as -1, chapter as 0, |
1307 | * heading as 1 and subsection as 2. However, the semantics of |
1308 | * cfg->leaf_level are defined to count chapter as 1, heading |
1309 | * as 2 etc. So first I increment depth :-( |
1310 | */ |
1311 | ldepth = depth + 1; |
1312 | |
1313 | if (cfg->leaf_level == 0) { |
1314 | /* |
1315 | * leaf_level==0 is a special case, in which everything is |
1316 | * put into a single file. |
1317 | */ |
1318 | if (!files->single) |
1319 | files->single = html_new_file(files, cfg->single_filename); |
1320 | |
1321 | file = files->single; |
1322 | } else { |
1323 | /* |
1324 | * If the depth of this section is at or above leaf_level, |
1325 | * we invent a fresh file and put this section at its head. |
1326 | * Otherwise, we put it in the same file as its parent |
1327 | * section. |
1328 | */ |
1329 | if (ldepth > cfg->leaf_level) { |
1330 | /* |
1331 | * We know that sect->parent cannot be NULL. The only |
1332 | * circumstance in which it can be is if sect is at |
1333 | * chapter or appendix level, i.e. ldepth==1; and if |
1334 | * that's the case, then we cannot have entered this |
1335 | * branch unless cfg->leaf_level==0, in which case we |
1336 | * would be in the single-file case above and not here |
1337 | * at all. |
1338 | */ |
1339 | assert(sect->parent); |
1340 | |
1341 | file = sect->parent->file; |
1342 | } else { |
1343 | if (sect->type == TOP) { |
1344 | file = html_new_file(files, cfg->contents_filename); |
1345 | } else if (sect->type == INDEX) { |
1346 | file = html_new_file(files, cfg->index_filename); |
1347 | } else { |
1348 | char *title; |
1349 | |
1350 | assert(ldepth > 0 && sect->title); |
1351 | title = html_format(sect->title, cfg->template_filename); |
1352 | file = html_new_file(files, title); |
1353 | sfree(title); |
1354 | } |
1355 | } |
1356 | } |
1357 | |
1358 | sect->file = file; |
1359 | |
1360 | if (file->min_heading_depth > depth) { |
1361 | /* |
1362 | * This heading is at a higher level than any heading we |
1363 | * have so far placed in this file; so we set the `first' |
1364 | * pointer. |
1365 | */ |
1366 | file->min_heading_depth = depth; |
1367 | file->first = sect; |
1368 | } |
1369 | |
1370 | if (file->min_heading_depth == depth) |
1371 | file->last = sect; |
1372 | } |
1373 | |
1374 | static htmlfile *html_new_file(htmlfilelist *list, char *filename) |
1375 | { |
1376 | htmlfile *ret = mknew(htmlfile); |
1377 | |
1378 | ret->next = NULL; |
1379 | if (list->tail) |
1380 | list->tail->next = ret; |
1381 | else |
1382 | list->head = ret; |
1383 | list->tail = ret; |
1384 | |
1385 | ret->filename = dupstr(filename); |
1386 | ret->last_fragment_number = 0; |
1387 | ret->min_heading_depth = INT_MAX; |
1388 | ret->first = ret->last = NULL; |
1389 | |
1390 | return ret; |
1391 | } |
1392 | |
1393 | static htmlsect *html_new_sect(htmlsectlist *list, paragraph *title) |
1394 | { |
1395 | htmlsect *ret = mknew(htmlsect); |
1396 | |
1397 | ret->next = NULL; |
1398 | if (list->tail) |
1399 | list->tail->next = ret; |
1400 | else |
1401 | list->head = ret; |
1402 | list->tail = ret; |
1403 | |
1404 | ret->title = title; |
1405 | ret->file = NULL; |
1406 | ret->parent = NULL; |
1407 | ret->type = NORMAL; |
1408 | |
1409 | return ret; |
1410 | } |
1411 | |
1412 | static void html_words(htmloutput *ho, word *words, int flags, |
1413 | htmlfile *file, keywordlist *keywords, htmlconfig *cfg) |
1414 | { |
1415 | word *w; |
1416 | char *c; |
1417 | int style, type; |
1418 | |
1419 | for (w = words; w; w = w->next) switch (w->type) { |
1420 | case word_HyperLink: |
1421 | if (flags & LINKS) { |
1422 | element_open(ho, "a"); |
1423 | c = utoa_dup(w->text, CS_ASCII); |
1424 | element_attr(ho, "href", c); |
1425 | sfree(c); |
1426 | } |
1427 | break; |
1428 | case word_UpperXref: |
1429 | case word_LowerXref: |
1430 | if (flags & LINKS) { |
1431 | keyword *kwl = kw_lookup(keywords, w->text); |
1432 | paragraph *p = kwl->para; |
1433 | htmlsect *s = (htmlsect *)p->private_data; |
1434 | |
1435 | assert(s); |
1436 | |
1437 | html_href(ho, file, s->file, s->fragment); |
1438 | } |
1439 | break; |
1440 | case word_HyperEnd: |
1441 | case word_XrefEnd: |
1442 | if (flags & LINKS) |
1443 | element_close(ho, "a"); |
1444 | break; |
1445 | case word_IndexRef: |
1446 | if (flags & INDEXENTS) { |
1447 | htmlindexref *hr = (htmlindexref *)w->private_data; |
1448 | element_open(ho, "a"); |
1449 | element_attr(ho, "name", hr->fragment); |
1450 | element_close(ho, "a"); |
1451 | } |
1452 | break; |
1453 | case word_Normal: |
1454 | case word_Emph: |
1455 | case word_Code: |
1456 | case word_WeakCode: |
1457 | case word_WhiteSpace: |
1458 | case word_EmphSpace: |
1459 | case word_CodeSpace: |
1460 | case word_WkCodeSpace: |
1461 | case word_Quote: |
1462 | case word_EmphQuote: |
1463 | case word_CodeQuote: |
1464 | case word_WkCodeQuote: |
1465 | style = towordstyle(w->type); |
1466 | type = removeattr(w->type); |
1467 | if (style == word_Emph && |
1468 | (attraux(w->aux) == attr_First || |
1469 | attraux(w->aux) == attr_Only) && |
1470 | (flags & MARKUP)) |
1471 | element_open(ho, "em"); |
1472 | else if ((style == word_Code || style == word_WeakCode) && |
1473 | (attraux(w->aux) == attr_First || |
1474 | attraux(w->aux) == attr_Only) && |
1475 | (flags & MARKUP)) |
1476 | element_open(ho, "code"); |
1477 | |
1478 | if (type == word_WhiteSpace) |
1479 | html_text(ho, L" "); |
1480 | else if (type == word_Quote) { |
1481 | if (quoteaux(w->aux) == quote_Open) |
1482 | html_text(ho, cfg->lquote); |
1483 | else |
1484 | html_text(ho, cfg->rquote); |
1485 | } else { |
1486 | if (cvt_ok(ho->charset, w->text) || !w->alt) |
1487 | html_text(ho, w->text); |
1488 | else |
1489 | html_words(ho, w->alt, flags, file, keywords, cfg); |
1490 | } |
1491 | |
1492 | if (style == word_Emph && |
1493 | (attraux(w->aux) == attr_Last || |
1494 | attraux(w->aux) == attr_Only) && |
1495 | (flags & MARKUP)) |
1496 | element_close(ho, "em"); |
1497 | else if ((style == word_Code || style == word_WeakCode) && |
1498 | (attraux(w->aux) == attr_Last || |
1499 | attraux(w->aux) == attr_Only) && |
1500 | (flags & MARKUP)) |
1501 | element_close(ho, "code"); |
1502 | |
1503 | break; |
1504 | } |
1505 | } |
1506 | |
1507 | static void html_codepara(htmloutput *ho, word *words) |
1508 | { |
1509 | element_open(ho, "pre"); |
1510 | element_open(ho, "code"); |
1511 | for (; words; words = words->next) if (words->type == word_WeakCode) { |
1512 | char *open_tag; |
1513 | wchar_t *t, *e; |
1514 | |
1515 | t = words->text; |
1516 | if (words->next && words->next->type == word_Emph) { |
1517 | e = words->next->text; |
1518 | words = words->next; |
1519 | } else |
1520 | e = NULL; |
1521 | |
1522 | while (e && *e && *t) { |
1523 | int n; |
1524 | int ec = *e; |
1525 | |
1526 | for (n = 0; t[n] && e[n] && e[n] == ec; n++); |
1527 | |
1528 | open_tag = NULL; |
1529 | if (ec == 'i') |
1530 | open_tag = "em"; |
1531 | else if (ec == 'b') |
1532 | open_tag = "b"; |
1533 | if (open_tag) |
1534 | element_open(ho, open_tag); |
1535 | |
1536 | html_text_limit(ho, t, n); |
1537 | |
1538 | if (open_tag) |
1539 | element_close(ho, open_tag); |
1540 | |
1541 | t += n; |
1542 | e += n; |
1543 | } |
1544 | html_text(ho, t); |
1545 | html_nl(ho); |
1546 | } |
1547 | element_close(ho, "code"); |
1548 | element_close(ho, "pre"); |
1549 | } |
1550 | |
1551 | static void html_charset_cleanup(htmloutput *ho) |
1552 | { |
1553 | char outbuf[256]; |
1554 | int bytes; |
1555 | |
1556 | bytes = charset_from_unicode(NULL, NULL, outbuf, lenof(outbuf), |
1557 | ho->charset, &ho->cstate, NULL); |
1558 | if (bytes > 0) |
1559 | fwrite(outbuf, 1, bytes, ho->fp); |
1560 | } |
1561 | |
1562 | static void return_to_neutral(htmloutput *ho) |
1563 | { |
1564 | if (ho->state == HO_IN_TEXT) { |
1565 | html_charset_cleanup(ho); |
1566 | } else if (ho->state == HO_IN_EMPTY_TAG && is_xhtml(ho->ver)) { |
1567 | fprintf(ho->fp, " />"); |
1568 | } else if (ho->state == HO_IN_EMPTY_TAG || ho->state == HO_IN_TAG) { |
1569 | fprintf(ho->fp, ">"); |
1570 | } |
1571 | |
1572 | ho->state = HO_NEUTRAL; |
1573 | } |
1574 | |
1575 | static void element_open(htmloutput *ho, char const *name) |
1576 | { |
1577 | return_to_neutral(ho); |
1578 | fprintf(ho->fp, "<%s", name); |
1579 | ho->state = HO_IN_TAG; |
1580 | } |
1581 | |
1582 | static void element_close(htmloutput *ho, char const *name) |
1583 | { |
1584 | return_to_neutral(ho); |
1585 | fprintf(ho->fp, "</%s>", name); |
1586 | ho->state = HO_NEUTRAL; |
1587 | } |
1588 | |
1589 | static void element_empty(htmloutput *ho, char const *name) |
1590 | { |
1591 | return_to_neutral(ho); |
1592 | fprintf(ho->fp, "<%s", name); |
1593 | ho->state = HO_IN_EMPTY_TAG; |
1594 | } |
1595 | |
1596 | static void html_nl(htmloutput *ho) |
1597 | { |
1598 | return_to_neutral(ho); |
1599 | fputc('\n', ho->fp); |
1600 | } |
1601 | |
1602 | static void html_raw(htmloutput *ho, char *text) |
1603 | { |
1604 | return_to_neutral(ho); |
1605 | fputs(text, ho->fp); |
1606 | } |
1607 | |
1608 | static void html_raw_as_attr(htmloutput *ho, char *text) |
1609 | { |
1610 | assert(ho->state == HO_IN_TAG || ho->state == HO_IN_EMPTY_TAG); |
1611 | fputc(' ', ho->fp); |
1612 | fputs(text, ho->fp); |
1613 | } |
1614 | |
1615 | static void element_attr(htmloutput *ho, char const *name, char const *value) |
1616 | { |
1617 | html_charset_cleanup(ho); |
1618 | assert(ho->state == HO_IN_TAG || ho->state == HO_IN_EMPTY_TAG); |
1619 | fprintf(ho->fp, " %s=\"%s\"", name, value); |
1620 | } |
1621 | |
1622 | static void element_attr_w(htmloutput *ho, char const *name, |
1623 | wchar_t const *value) |
1624 | { |
1625 | html_charset_cleanup(ho); |
1626 | fprintf(ho->fp, " %s=\"", name); |
1627 | html_text_limit_internal(ho, value, 0, TRUE); |
1628 | html_charset_cleanup(ho); |
1629 | fputc('"', ho->fp); |
1630 | } |
1631 | |
1632 | static void html_text(htmloutput *ho, wchar_t const *text) |
1633 | { |
1634 | html_text_limit(ho, text, 0); |
1635 | } |
1636 | |
1637 | static void html_text_limit(htmloutput *ho, wchar_t const *text, int maxlen) |
1638 | { |
1639 | return_to_neutral(ho); |
1640 | html_text_limit_internal(ho, text, maxlen, FALSE); |
1641 | } |
1642 | |
1643 | static void html_text_limit_internal(htmloutput *ho, wchar_t const *text, |
1644 | int maxlen, int quote_quotes) |
1645 | { |
1646 | int textlen = ustrlen(text); |
1647 | char outbuf[256]; |
1648 | int bytes, err; |
1649 | |
1650 | if (maxlen > 0 && textlen > maxlen) |
1651 | textlen = maxlen; |
1652 | |
1653 | while (textlen > 0) { |
1654 | /* Scan ahead for characters we really can't display in HTML. */ |
1655 | int lenbefore, lenafter; |
1656 | for (lenbefore = 0; lenbefore < textlen; lenbefore++) |
1657 | if (text[lenbefore] == L'<' || |
1658 | text[lenbefore] == L'>' || |
1659 | text[lenbefore] == L'&' || |
1660 | (text[lenbefore] == L'"' && quote_quotes)) |
1661 | break; |
1662 | lenafter = lenbefore; |
1663 | bytes = charset_from_unicode(&text, &lenafter, outbuf, lenof(outbuf), |
1664 | ho->charset, &ho->cstate, &err); |
1665 | textlen -= (lenbefore - lenafter); |
1666 | if (bytes > 0) |
1667 | fwrite(outbuf, 1, bytes, ho->fp); |
1668 | if (err) { |
1669 | /* |
1670 | * We have encountered a character that cannot be |
1671 | * displayed in the selected output charset. Therefore, |
1672 | * we use an HTML numeric entity reference. |
1673 | */ |
1674 | assert(textlen > 0); |
1675 | fprintf(ho->fp, "&#%ld;", (long int)*text); |
1676 | text++, textlen--; |
1677 | } else if (lenafter == 0 && textlen > 0) { |
1678 | /* |
1679 | * We have encountered a character which is special to |
1680 | * HTML. |
1681 | */ |
1682 | if (*text == L'<') |
1683 | fprintf(ho->fp, "<"); |
1684 | else if (*text == L'>') |
1685 | fprintf(ho->fp, ">"); |
1686 | else if (*text == L'&') |
1687 | fprintf(ho->fp, "&"); |
1688 | else if (*text == L'"') |
1689 | fprintf(ho->fp, """); |
1690 | else |
1691 | assert(!"Can't happen"); |
1692 | text++, textlen--; |
1693 | } |
1694 | } |
1695 | } |
1696 | |
1697 | static void cleanup(htmloutput *ho) |
1698 | { |
1699 | return_to_neutral(ho); |
1700 | fclose(ho->fp); |
1701 | } |
1702 | |
1703 | static void html_href(htmloutput *ho, htmlfile *thisfile, |
1704 | htmlfile *targetfile, char *targetfrag) |
1705 | { |
1706 | rdstringc rs = { 0, 0, NULL }; |
1707 | char *url; |
1708 | |
1709 | if (targetfile != thisfile) |
1710 | rdaddsc(&rs, targetfile->filename); |
1711 | if (targetfrag) { |
1712 | rdaddc(&rs, '#'); |
1713 | rdaddsc(&rs, targetfrag); |
1714 | } |
1715 | url = rs.text; |
1716 | |
1717 | element_open(ho, "a"); |
1718 | element_attr(ho, "href", url); |
1719 | sfree(url); |
1720 | } |
1721 | |
1722 | static char *html_format(paragraph *p, char *template_string) |
1723 | { |
1724 | char *c, *t; |
1725 | word *w; |
1726 | wchar_t *ws, wsbuf[2]; |
1727 | rdstringc rs = { 0, 0, NULL }; |
1728 | |
1729 | t = template_string; |
1730 | while (*t) { |
1731 | if (*t == '%' && t[1]) { |
1732 | int fmt; |
1733 | |
1734 | t++; |
1735 | fmt = *t++; |
1736 | |
1737 | if (fmt == '%') { |
1738 | rdaddc(&rs, fmt); |
1739 | continue; |
1740 | } |
1741 | |
1742 | w = NULL; |
1743 | ws = NULL; |
1744 | |
1745 | if (p->kwtext && fmt == 'n') |
1746 | w = p->kwtext; |
1747 | else if (p->kwtext2 && fmt == 'b') { |
1748 | /* |
1749 | * HTML fragment names must start with a letter, so |
1750 | * simply `1.2.3' is not adequate. In this case I'm |
1751 | * going to cheat slightly by prepending the first |
1752 | * character of the first word of kwtext, so that |
1753 | * we get `C1' for chapter 1, `S2.3' for section |
1754 | * 2.3 etc. |
1755 | */ |
1756 | if (p->kwtext && p->kwtext->text[0]) { |
1757 | ws = wsbuf; |
1758 | wsbuf[1] = '\0'; |
1759 | wsbuf[0] = p->kwtext->text[0]; |
1760 | } |
1761 | w = p->kwtext2; |
1762 | } else if (p->keyword && *p->keyword && fmt == 'k') |
1763 | ws = p->keyword; |
1764 | else |
1765 | w = p->words; |
1766 | |
1767 | if (ws) { |
1768 | c = utoa_dup(ws, CS_ASCII); |
1769 | rdaddsc(&rs,c); |
1770 | sfree(c); |
1771 | } |
1772 | |
1773 | while (w) { |
1774 | if (removeattr(w->type) == word_Normal) { |
1775 | c = utoa_dup(w->text, CS_ASCII); |
1776 | rdaddsc(&rs,c); |
1777 | sfree(c); |
1778 | } |
1779 | w = w->next; |
1780 | } |
1781 | } else { |
1782 | rdaddc(&rs, *t++); |
1783 | } |
1784 | } |
1785 | |
1786 | return rdtrimc(&rs); |
1787 | } |
1788 | |
1789 | static void html_sanitise_fragment(char *text) |
1790 | { |
1791 | /* |
1792 | * The HTML 4 spec's strictest definition of fragment names (<a |
1793 | * name> and "id" attributes) says that they `must begin with a |
1794 | * letter and may be followed by any number of letters, digits, |
1795 | * hyphens, underscores, colons, and periods'. |
1796 | * |
1797 | * So here we unceremoniously rip out any characters not |
1798 | * conforming to this limitation. |
1799 | */ |
1800 | char *p = text, *q = text; |
1801 | |
1802 | while (*p && !((*p>='A' && *p<='Z') || (*p>='a' && *p<='z'))) |
1803 | p++; |
1804 | if (!(*q++ = *p++)) |
1805 | return; |
1806 | while (*p) { |
1807 | if ((*p>='A' && *p<='Z') || |
1808 | (*p>='a' && *p<='z') || |
1809 | (*p>='0' && *p<='9') || |
1810 | *p=='-' || *p=='_' || *p==':' || *p=='.') |
1811 | *q++ = *p; |
1812 | p++; |
1813 | } |
1814 | |
1815 | *q = '\0'; |
1816 | } |
1817 | |
1818 | static void html_contents_entry(htmloutput *ho, int depth, htmlsect *s, |
1819 | htmlfile *thisfile, keywordlist *keywords, |
1820 | htmlconfig *cfg) |
1821 | { |
1822 | while (ho->contents_level > depth) { |
1823 | element_close(ho, "ul"); |
1824 | ho->contents_level--; |
1825 | } |
1826 | |
1827 | while (ho->contents_level < depth) { |
1828 | element_open(ho, "ul"); |
1829 | ho->contents_level++; |
1830 | } |
1831 | |
1832 | if (!s) |
1833 | return; |
1834 | |
1835 | element_open(ho, "li"); |
1836 | html_href(ho, thisfile, s->file, s->fragment); |
1837 | html_section_title(ho, s, thisfile, keywords, cfg); |
1838 | element_close(ho, "a"); |
1839 | element_close(ho, "li"); |
1840 | } |
1841 | |
1842 | static void html_section_title(htmloutput *ho, htmlsect *s, htmlfile *thisfile, |
1843 | keywordlist *keywords, htmlconfig *cfg) |
1844 | { |
1845 | if (s->title) { |
1846 | sectlevel *sl; |
1847 | word *number; |
1848 | int depth = heading_depth(s->title); |
1849 | |
1850 | if (depth < 0) |
1851 | sl = NULL; |
1852 | else if (depth == 0) |
1853 | sl = &cfg->achapter; |
1854 | else if (depth <= cfg->nasect) |
1855 | sl = &cfg->asect[depth-1]; |
1856 | else |
1857 | sl = &cfg->asect[cfg->nasect-1]; |
1858 | |
1859 | if (!sl) |
1860 | number = NULL; |
1861 | else if (sl->just_numbers) |
1862 | number = s->title->kwtext2; |
1863 | else |
1864 | number = s->title->kwtext; |
1865 | |
1866 | if (number) { |
1867 | html_words(ho, number, MARKUP, |
1868 | thisfile, keywords, cfg); |
1869 | html_text(ho, sl->number_suffix); |
1870 | } |
1871 | |
1872 | html_words(ho, s->title->words, MARKUP, |
1873 | thisfile, keywords, cfg); |
1874 | } else { |
1875 | assert(s->type != NORMAL); |
1876 | if (s->type == TOP) |
1877 | html_text(ho, L"Preamble");/* FIXME: configure */ |
1878 | else if (s->type == INDEX) |
1879 | html_text(ho, L"Index");/* FIXME: configure */ |
1880 | } |
1881 | } |