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