| 1 | /* |
| 2 | * input.c: read the source form |
| 3 | */ |
| 4 | |
| 5 | #include <stdio.h> |
| 6 | #include <assert.h> |
| 7 | #include <time.h> |
| 8 | #include "halibut.h" |
| 9 | |
| 10 | #define TAB_STOP 8 /* for column number tracking */ |
| 11 | |
| 12 | static void setpos(input *in, char *fname) { |
| 13 | in->pos.filename = fname; |
| 14 | in->pos.line = 1; |
| 15 | in->pos.col = (in->reportcols ? 1 : -1); |
| 16 | } |
| 17 | |
| 18 | static void unget(input *in, int c, filepos *pos) { |
| 19 | if (in->npushback >= in->pushbacksize) { |
| 20 | in->pushbacksize = in->npushback + 16; |
| 21 | in->pushback = sresize(in->pushback, in->pushbacksize, pushback); |
| 22 | } |
| 23 | in->pushback[in->npushback].chr = c; |
| 24 | in->pushback[in->npushback].pos = *pos; /* structure copy */ |
| 25 | in->npushback++; |
| 26 | } |
| 27 | |
| 28 | /* ---------------------------------------------------------------------- */ |
| 29 | /* |
| 30 | * Macro subsystem |
| 31 | */ |
| 32 | typedef struct macro_Tag macro; |
| 33 | struct macro_Tag { |
| 34 | wchar_t *name, *text; |
| 35 | }; |
| 36 | struct macrostack_Tag { |
| 37 | macrostack *next; |
| 38 | wchar_t *text; |
| 39 | int ptr, npushback; |
| 40 | filepos pos; |
| 41 | }; |
| 42 | static int macrocmp(void *av, void *bv) { |
| 43 | macro *a = (macro *)av, *b = (macro *)bv; |
| 44 | return ustrcmp(a->name, b->name); |
| 45 | } |
| 46 | static void macrodef(tree234 *macros, wchar_t *name, wchar_t *text, |
| 47 | filepos fpos) { |
| 48 | macro *m = snew(macro); |
| 49 | m->name = name; |
| 50 | m->text = text; |
| 51 | if (add234(macros, m) != m) { |
| 52 | error(err_macroexists, &fpos, name); |
| 53 | sfree(name); |
| 54 | sfree(text); |
| 55 | } |
| 56 | } |
| 57 | static int macrolookup(tree234 *macros, input *in, wchar_t *name, |
| 58 | filepos *pos) { |
| 59 | macro m, *gotit; |
| 60 | m.name = name; |
| 61 | gotit = find234(macros, &m, NULL); |
| 62 | if (gotit) { |
| 63 | macrostack *expansion = snew(macrostack); |
| 64 | expansion->next = in->stack; |
| 65 | expansion->text = gotit->text; |
| 66 | expansion->pos = *pos; /* structure copy */ |
| 67 | expansion->ptr = 0; |
| 68 | expansion->npushback = in->npushback; |
| 69 | in->stack = expansion; |
| 70 | return TRUE; |
| 71 | } else |
| 72 | return FALSE; |
| 73 | } |
| 74 | static void macrocleanup(tree234 *macros) { |
| 75 | int ti; |
| 76 | macro *m; |
| 77 | for (ti = 0; (m = (macro *)index234(macros, ti)) != NULL; ti++) { |
| 78 | sfree(m->name); |
| 79 | sfree(m->text); |
| 80 | sfree(m); |
| 81 | } |
| 82 | freetree234(macros); |
| 83 | } |
| 84 | |
| 85 | static void input_configure(input *in, paragraph *cfg) { |
| 86 | assert(cfg->type == para_Config); |
| 87 | |
| 88 | if (!ustricmp(cfg->keyword, L"input-charset")) { |
| 89 | in->charset = charset_from_ustr(&cfg->fpos, uadv(cfg->keyword)); |
| 90 | } |
| 91 | } |
| 92 | |
| 93 | /* |
| 94 | * Can return EOF |
| 95 | */ |
| 96 | static int get(input *in, filepos *pos, rdstringc *rsc) { |
| 97 | int pushbackpt = in->stack ? in->stack->npushback : 0; |
| 98 | if (in->npushback > pushbackpt) { |
| 99 | --in->npushback; |
| 100 | if (pos) |
| 101 | *pos = in->pushback[in->npushback].pos; /* structure copy */ |
| 102 | return in->pushback[in->npushback].chr; |
| 103 | } |
| 104 | else if (in->stack) { |
| 105 | wchar_t c = in->stack->text[in->stack->ptr]; |
| 106 | if (pos) |
| 107 | *pos = in->stack->pos; |
| 108 | if (in->stack->text[++in->stack->ptr] == L'\0') { |
| 109 | macrostack *tmp = in->stack; |
| 110 | in->stack = tmp->next; |
| 111 | sfree(tmp); |
| 112 | } |
| 113 | return c; |
| 114 | } |
| 115 | else if (in->currfp) { |
| 116 | |
| 117 | while (in->wcpos >= in->nwc) { |
| 118 | |
| 119 | int c = getc(in->currfp); |
| 120 | |
| 121 | if (c == EOF) { |
| 122 | fclose(in->currfp); |
| 123 | in->currfp = NULL; |
| 124 | return EOF; |
| 125 | } |
| 126 | |
| 127 | if (rsc) |
| 128 | rdaddc(rsc, c); |
| 129 | |
| 130 | /* Track line numbers, for error reporting */ |
| 131 | if (pos) |
| 132 | *pos = in->pos; |
| 133 | if (in->reportcols) { |
| 134 | switch (c) { |
| 135 | case '\t': |
| 136 | in->pos.col = 1 + (in->pos.col + TAB_STOP-1) % TAB_STOP; |
| 137 | break; |
| 138 | case '\n': |
| 139 | in->pos.col = 1; |
| 140 | in->pos.line++; |
| 141 | break; |
| 142 | default: |
| 143 | in->pos.col++; |
| 144 | break; |
| 145 | } |
| 146 | } else { |
| 147 | in->pos.col = -1; |
| 148 | if (c == '\n') |
| 149 | in->pos.line++; |
| 150 | } |
| 151 | |
| 152 | /* |
| 153 | * Do input character set translation, so that we return |
| 154 | * Unicode. |
| 155 | */ |
| 156 | { |
| 157 | char buf[1]; |
| 158 | char const *p; |
| 159 | int inlen; |
| 160 | |
| 161 | buf[0] = (char)c; |
| 162 | p = buf; |
| 163 | inlen = 1; |
| 164 | |
| 165 | in->nwc = charset_to_unicode(&p, &inlen, |
| 166 | in->wc, lenof(in->wc), |
| 167 | in->charset, &in->csstate, |
| 168 | NULL, 0); |
| 169 | assert(p == buf+1 && inlen == 0); |
| 170 | |
| 171 | in->wcpos = 0; |
| 172 | } |
| 173 | } |
| 174 | |
| 175 | return in->wc[in->wcpos++]; |
| 176 | |
| 177 | } else |
| 178 | return EOF; |
| 179 | } |
| 180 | |
| 181 | /* |
| 182 | * Lexical analysis of source files. |
| 183 | */ |
| 184 | typedef struct token_Tag token; |
| 185 | struct token_Tag { |
| 186 | int type; |
| 187 | int cmd, aux; |
| 188 | wchar_t *text; |
| 189 | char *origtext; |
| 190 | filepos pos; |
| 191 | }; |
| 192 | enum { |
| 193 | tok_eof, /* end of file */ |
| 194 | tok_eop, /* end of paragraph */ |
| 195 | tok_white, /* whitespace */ |
| 196 | tok_word, /* a word or word fragment */ |
| 197 | tok_cmd, /* \command */ |
| 198 | tok_lbrace, /* { */ |
| 199 | tok_rbrace /* } */ |
| 200 | }; |
| 201 | |
| 202 | /* Halibut command keywords. */ |
| 203 | enum { |
| 204 | c__invalid, /* invalid command */ |
| 205 | c__comment, /* comment command (\#) */ |
| 206 | c__escaped, /* escaped character */ |
| 207 | c__nop, /* no-op */ |
| 208 | c__nbsp, /* nonbreaking space */ |
| 209 | c_A, /* appendix heading */ |
| 210 | c_B, /* bibliography entry */ |
| 211 | c_BR, /* bibliography rewrite */ |
| 212 | c_C, /* chapter heading */ |
| 213 | c_H, /* heading */ |
| 214 | c_I, /* invisible index mark */ |
| 215 | c_IM, /* index merge/rewrite */ |
| 216 | c_K, /* capitalised cross-reference */ |
| 217 | c_S, /* aux field is 0, 1, 2, ... */ |
| 218 | c_U, /* unnumbered-chapter heading */ |
| 219 | c_W, /* Web hyperlink */ |
| 220 | c_b, /* bulletted list */ |
| 221 | c_c, /* code */ |
| 222 | c_cfg, /* configuration directive */ |
| 223 | c_copyright, /* copyright statement */ |
| 224 | c_cq, /* quoted code (sugar for \q{\cw{x}}) */ |
| 225 | c_cw, /* weak code */ |
| 226 | c_date, /* document processing date */ |
| 227 | c_dd, /* description list: description */ |
| 228 | c_define, /* macro definition */ |
| 229 | c_dt, /* description list: described thing */ |
| 230 | c_e, /* emphasis */ |
| 231 | c_i, /* visible index mark */ |
| 232 | c_ii, /* uncapitalised visible index mark */ |
| 233 | c_k, /* uncapitalised cross-reference */ |
| 234 | c_lcont, /* continuation para(s) for list item */ |
| 235 | c_n, /* numbered list */ |
| 236 | c_nocite, /* bibliography trickery */ |
| 237 | c_preamble, /* (obsolete) preamble text */ |
| 238 | c_q, /* quote marks */ |
| 239 | c_quote, /* block-quoted paragraphs */ |
| 240 | c_rule, /* horizontal rule */ |
| 241 | c_title, /* document title */ |
| 242 | c_u, /* aux field is char code */ |
| 243 | c_versionid /* document RCS id */ |
| 244 | }; |
| 245 | |
| 246 | /* Perhaps whitespace should be defined in a more Unicode-friendly way? */ |
| 247 | #define iswhite(c) ( (c)==32 || (c)==9 || (c)==13 || (c)==10 ) |
| 248 | #define isnl(c) ( (c)==10 ) |
| 249 | #define isdec(c) ( ((c)>='0'&&(c)<='9') ) |
| 250 | #define fromdec(c) ( (c)-'0' ) |
| 251 | #define ishex(c) ( ((c)>='0'&&(c)<='9') || ((c)>='A'&&(c)<='F') || ((c)>='a'&&(c)<='f')) |
| 252 | #define fromhex(c) ( (c)<='9' ? (c)-'0' : ((c)&0xDF) - ('A'-10) ) |
| 253 | #define iscmd(c) ( ((c)>='0'&&(c)<='9') || ((c)>='A'&&(c)<='Z') || ((c)>='a'&&(c)<='z')) |
| 254 | |
| 255 | /* |
| 256 | * Keyword comparison function. Like strcmp, but between a wchar_t * |
| 257 | * and a char *. |
| 258 | */ |
| 259 | static int kwcmp(wchar_t const *p, char const *q) { |
| 260 | int i; |
| 261 | do { |
| 262 | i = *p - *q; |
| 263 | } while (*p++ && *q++ && !i); |
| 264 | return i; |
| 265 | } |
| 266 | |
| 267 | /* |
| 268 | * Match a keyword. |
| 269 | */ |
| 270 | static void match_kw(token *tok) { |
| 271 | /* |
| 272 | * FIXME. The ids are explicit in here so as to allow long-name |
| 273 | * equivalents to the various very short keywords. |
| 274 | */ |
| 275 | static const struct { char const *name; int id; } keywords[] = { |
| 276 | {"#", c__comment}, /* comment command (\#) */ |
| 277 | {"-", c__escaped}, /* nonbreaking hyphen */ |
| 278 | {".", c__nop}, /* no-op */ |
| 279 | {"A", c_A}, /* appendix heading */ |
| 280 | {"B", c_B}, /* bibliography entry */ |
| 281 | {"BR", c_BR}, /* bibliography rewrite */ |
| 282 | {"C", c_C}, /* chapter heading */ |
| 283 | {"H", c_H}, /* heading */ |
| 284 | {"I", c_I}, /* invisible index mark */ |
| 285 | {"IM", c_IM}, /* index merge/rewrite */ |
| 286 | {"K", c_K}, /* capitalised cross-reference */ |
| 287 | {"U", c_U}, /* unnumbered-chapter heading */ |
| 288 | {"W", c_W}, /* Web hyperlink */ |
| 289 | {"\\", c__escaped}, /* escaped backslash (\\) */ |
| 290 | {"_", c__nbsp}, /* nonbreaking space (\_) */ |
| 291 | {"b", c_b}, /* bulletted list */ |
| 292 | {"c", c_c}, /* code */ |
| 293 | {"cfg", c_cfg}, /* configuration directive */ |
| 294 | {"copyright", c_copyright}, /* copyright statement */ |
| 295 | {"cq", c_cq}, /* quoted code (sugar for \q{\cw{x}}) */ |
| 296 | {"cw", c_cw}, /* weak code */ |
| 297 | {"date", c_date}, /* document processing date */ |
| 298 | {"dd", c_dd}, /* description list: description */ |
| 299 | {"define", c_define}, /* macro definition */ |
| 300 | {"dt", c_dt}, /* description list: described thing */ |
| 301 | {"e", c_e}, /* emphasis */ |
| 302 | {"i", c_i}, /* visible index mark */ |
| 303 | {"ii", c_ii}, /* uncapitalised visible index mark */ |
| 304 | {"k", c_k}, /* uncapitalised cross-reference */ |
| 305 | {"lcont", c_lcont}, /* continuation para(s) for list item */ |
| 306 | {"n", c_n}, /* numbered list */ |
| 307 | {"nocite", c_nocite}, /* bibliography trickery */ |
| 308 | {"preamble", c_preamble}, /* (obsolete) preamble text */ |
| 309 | {"q", c_q}, /* quote marks */ |
| 310 | {"quote", c_quote}, /* block-quoted paragraphs */ |
| 311 | {"rule", c_rule}, /* horizontal rule */ |
| 312 | {"title", c_title}, /* document title */ |
| 313 | {"versionid", c_versionid}, /* document RCS id */ |
| 314 | {"{", c__escaped}, /* escaped lbrace (\{) */ |
| 315 | {"}", c__escaped}, /* escaped rbrace (\}) */ |
| 316 | }; |
| 317 | int i, j, k, c; |
| 318 | |
| 319 | /* |
| 320 | * Special cases: \S{0,1,2,...} and \uABCD. If the syntax |
| 321 | * doesn't match correctly, we just fall through to the |
| 322 | * binary-search phase. |
| 323 | */ |
| 324 | if (tok->text[0] == 'S') { |
| 325 | /* We expect numeric characters thereafter. */ |
| 326 | wchar_t *p = tok->text+1; |
| 327 | int n; |
| 328 | if (!*p) |
| 329 | n = 1; |
| 330 | else { |
| 331 | n = 0; |
| 332 | while (*p && isdec(*p)) { |
| 333 | n = 10 * n + fromdec(*p); |
| 334 | p++; |
| 335 | } |
| 336 | } |
| 337 | if (!*p) { |
| 338 | tok->cmd = c_S; |
| 339 | tok->aux = n; |
| 340 | return; |
| 341 | } |
| 342 | } else if (tok->text[0] == 'u') { |
| 343 | /* We expect hex characters thereafter. */ |
| 344 | wchar_t *p = tok->text+1; |
| 345 | int n = 0; |
| 346 | while (*p && ishex(*p)) { |
| 347 | n = 16 * n + fromhex(*p); |
| 348 | p++; |
| 349 | } |
| 350 | if (!*p) { |
| 351 | tok->cmd = c_u; |
| 352 | tok->aux = n; |
| 353 | return; |
| 354 | } |
| 355 | } |
| 356 | |
| 357 | i = -1; |
| 358 | j = sizeof(keywords)/sizeof(*keywords); |
| 359 | while (j-i > 1) { |
| 360 | k = (i+j)/2; |
| 361 | c = kwcmp(tok->text, keywords[k].name); |
| 362 | if (c < 0) |
| 363 | j = k; |
| 364 | else if (c > 0) |
| 365 | i = k; |
| 366 | else /* c == 0 */ { |
| 367 | tok->cmd = keywords[k].id; |
| 368 | return; |
| 369 | } |
| 370 | } |
| 371 | |
| 372 | tok->cmd = c__invalid; |
| 373 | } |
| 374 | |
| 375 | |
| 376 | /* |
| 377 | * Read a token from the input file, in the normal way (`normal' in |
| 378 | * the sense that code paragraphs work a different way). |
| 379 | */ |
| 380 | token get_token(input *in) { |
| 381 | int c; |
| 382 | int nls; |
| 383 | int prevpos; |
| 384 | token ret; |
| 385 | rdstring rs = { 0, 0, NULL }; |
| 386 | rdstringc rsc = { 0, 0, NULL }; |
| 387 | filepos cpos; |
| 388 | |
| 389 | ret.text = NULL; /* default */ |
| 390 | ret.origtext = NULL; /* default */ |
| 391 | if (in->pushback_chars) { |
| 392 | rdaddsc(&rsc, in->pushback_chars); |
| 393 | sfree(in->pushback_chars); |
| 394 | in->pushback_chars = NULL; |
| 395 | } |
| 396 | c = get(in, &cpos, &rsc); |
| 397 | ret.pos = cpos; |
| 398 | if (iswhite(c)) { /* tok_white or tok_eop */ |
| 399 | nls = 0; |
| 400 | prevpos = 0; |
| 401 | do { |
| 402 | if (isnl(c)) |
| 403 | nls++; |
| 404 | prevpos = rsc.pos; |
| 405 | } while ((c = get(in, &cpos, &rsc)) != EOF && iswhite(c)); |
| 406 | if (c == EOF) { |
| 407 | ret.type = tok_eof; |
| 408 | sfree(rsc.text); |
| 409 | return ret; |
| 410 | } |
| 411 | if (rsc.text) { |
| 412 | in->pushback_chars = dupstr(rsc.text + prevpos); |
| 413 | sfree(rsc.text); |
| 414 | } |
| 415 | unget(in, c, &cpos); |
| 416 | ret.type = (nls > 1 ? tok_eop : tok_white); |
| 417 | return ret; |
| 418 | } else if (c == EOF) { /* tok_eof */ |
| 419 | ret.type = tok_eof; |
| 420 | sfree(rsc.text); |
| 421 | return ret; |
| 422 | } else if (c == '\\') { /* tok_cmd */ |
| 423 | rsc.pos = prevpos = 0; |
| 424 | c = get(in, &cpos, &rsc); |
| 425 | if (c == '-' || c == '\\' || c == '_' || |
| 426 | c == '#' || c == '{' || c == '}' || c == '.') { |
| 427 | /* single-char command */ |
| 428 | rdadd(&rs, c); |
| 429 | prevpos = rsc.pos; |
| 430 | } else if (c == 'u') { |
| 431 | int len = 0; |
| 432 | do { |
| 433 | rdadd(&rs, c); |
| 434 | len++; |
| 435 | prevpos = rsc.pos; |
| 436 | c = get(in, &cpos, &rsc); |
| 437 | } while (ishex(c) && len < 5); |
| 438 | unget(in, c, &cpos); |
| 439 | } else if (iscmd(c)) { |
| 440 | do { |
| 441 | rdadd(&rs, c); |
| 442 | prevpos = rsc.pos; |
| 443 | c = get(in, &cpos, &rsc); |
| 444 | } while (iscmd(c)); |
| 445 | unget(in, c, &cpos); |
| 446 | } |
| 447 | /* |
| 448 | * Now match the command against the list of available |
| 449 | * ones. |
| 450 | */ |
| 451 | ret.type = tok_cmd; |
| 452 | ret.text = ustrdup(rs.text); |
| 453 | if (rsc.text) { |
| 454 | in->pushback_chars = dupstr(rsc.text + prevpos); |
| 455 | rsc.text[prevpos] = '\0'; |
| 456 | ret.origtext = dupstr(rsc.text); |
| 457 | } else { |
| 458 | ret.origtext = dupstr(""); |
| 459 | } |
| 460 | match_kw(&ret); |
| 461 | sfree(rs.text); |
| 462 | sfree(rsc.text); |
| 463 | return ret; |
| 464 | } else if (c == '{') { /* tok_lbrace */ |
| 465 | ret.type = tok_lbrace; |
| 466 | sfree(rsc.text); |
| 467 | return ret; |
| 468 | } else if (c == '}') { /* tok_rbrace */ |
| 469 | ret.type = tok_rbrace; |
| 470 | sfree(rsc.text); |
| 471 | return ret; |
| 472 | } else { /* tok_word */ |
| 473 | /* |
| 474 | * Read a word: the longest possible contiguous sequence of |
| 475 | * things other than whitespace, backslash, braces and |
| 476 | * hyphen. A hyphen terminates the word but is returned as |
| 477 | * part of it; everything else is pushed back for the next |
| 478 | * token. The `aux' field contains TRUE if the word ends in |
| 479 | * a hyphen. |
| 480 | */ |
| 481 | ret.aux = FALSE; /* assumed for now */ |
| 482 | prevpos = 0; |
| 483 | while (1) { |
| 484 | if (iswhite(c) || c=='{' || c=='}' || c=='\\' || c==EOF) { |
| 485 | /* Put back the character that caused termination */ |
| 486 | unget(in, c, &cpos); |
| 487 | break; |
| 488 | } else { |
| 489 | rdadd(&rs, c); |
| 490 | if (c == '-') { |
| 491 | prevpos = rsc.pos; |
| 492 | ret.aux = TRUE; |
| 493 | break; /* hyphen terminates word */ |
| 494 | } |
| 495 | } |
| 496 | prevpos = rsc.pos; |
| 497 | c = get(in, &cpos, &rsc); |
| 498 | } |
| 499 | ret.type = tok_word; |
| 500 | ret.text = ustrdup(rs.text); |
| 501 | if (rsc.text) { |
| 502 | in->pushback_chars = dupstr(rsc.text + prevpos); |
| 503 | rsc.text[prevpos] = '\0'; |
| 504 | ret.origtext = dupstr(rsc.text); |
| 505 | } else { |
| 506 | ret.origtext = dupstr(""); |
| 507 | } |
| 508 | sfree(rs.text); |
| 509 | sfree(rsc.text); |
| 510 | return ret; |
| 511 | } |
| 512 | } |
| 513 | |
| 514 | /* |
| 515 | * Determine whether the next input character is an open brace (for |
| 516 | * telling code paragraphs from paragraphs which merely start with |
| 517 | * code). |
| 518 | */ |
| 519 | int isbrace(input *in) { |
| 520 | int c; |
| 521 | filepos cpos; |
| 522 | |
| 523 | c = get(in, &cpos, NULL); |
| 524 | unget(in, c, &cpos); |
| 525 | return (c == '{'); |
| 526 | } |
| 527 | |
| 528 | /* |
| 529 | * Read the rest of a line that starts `\c'. Including nothing at |
| 530 | * all (tok_word with empty text). |
| 531 | */ |
| 532 | token get_codepar_token(input *in) { |
| 533 | int c; |
| 534 | token ret; |
| 535 | rdstring rs = { 0, 0, NULL }; |
| 536 | filepos cpos; |
| 537 | |
| 538 | ret.type = tok_word; |
| 539 | ret.origtext = NULL; |
| 540 | c = get(in, &cpos, NULL); /* expect (and discard) one space */ |
| 541 | ret.pos = cpos; |
| 542 | if (c == ' ') { |
| 543 | c = get(in, &cpos, NULL); |
| 544 | ret.pos = cpos; |
| 545 | } |
| 546 | while (!isnl(c) && c != EOF) { |
| 547 | int c2 = c; |
| 548 | c = get(in, &cpos, NULL); |
| 549 | /* Discard \r just before \n. */ |
| 550 | if (c2 != 13 || !isnl(c)) |
| 551 | rdadd(&rs, c2); |
| 552 | } |
| 553 | unget(in, c, &cpos); |
| 554 | ret.text = ustrdup(rs.text); |
| 555 | sfree(rs.text); |
| 556 | return ret; |
| 557 | } |
| 558 | |
| 559 | /* |
| 560 | * Adds a new word to a linked list |
| 561 | */ |
| 562 | static word *addword(word newword, word ***hptrptr) { |
| 563 | word *mnewword; |
| 564 | if (!hptrptr) |
| 565 | return NULL; |
| 566 | mnewword = snew(word); |
| 567 | *mnewword = newword; /* structure copy */ |
| 568 | mnewword->next = NULL; |
| 569 | **hptrptr = mnewword; |
| 570 | *hptrptr = &mnewword->next; |
| 571 | return mnewword; |
| 572 | } |
| 573 | |
| 574 | /* |
| 575 | * Adds a new paragraph to a linked list |
| 576 | */ |
| 577 | static paragraph *addpara(paragraph newpara, paragraph ***hptrptr) { |
| 578 | paragraph *mnewpara = snew(paragraph); |
| 579 | *mnewpara = newpara; /* structure copy */ |
| 580 | mnewpara->next = NULL; |
| 581 | **hptrptr = mnewpara; |
| 582 | *hptrptr = &mnewpara->next; |
| 583 | return mnewpara; |
| 584 | } |
| 585 | |
| 586 | /* |
| 587 | * Destructor before token is reassigned; should catch most memory |
| 588 | * leaks |
| 589 | */ |
| 590 | #define dtor(t) ( sfree(t.text), sfree(t.origtext) ) |
| 591 | |
| 592 | /* |
| 593 | * Reads a single file (ie until get() returns EOF) |
| 594 | */ |
| 595 | static void read_file(paragraph ***ret, input *in, indexdata *idx, |
| 596 | tree234 *macros) { |
| 597 | token t; |
| 598 | paragraph par; |
| 599 | word wd, **whptr, **idximplicit; |
| 600 | wchar_t utext[2], *wdtext; |
| 601 | int style, spcstyle; |
| 602 | int already; |
| 603 | int iswhite, seenwhite; |
| 604 | int type; |
| 605 | int prev_para_type; |
| 606 | struct stack_item { |
| 607 | enum { |
| 608 | stack_nop = 0, /* do nothing (for error recovery) */ |
| 609 | stack_ualt = 1, /* \u alternative */ |
| 610 | stack_style = 2, /* \e, \c, \cw */ |
| 611 | stack_idx = 4, /* \I, \i, \ii */ |
| 612 | stack_hyper = 8, /* \W */ |
| 613 | stack_quote = 16 /* \q */ |
| 614 | } type; |
| 615 | word **whptr; /* to restore from \u alternatives */ |
| 616 | word **idximplicit; /* to restore from \u alternatives */ |
| 617 | filepos fpos; |
| 618 | int in_code; |
| 619 | } *sitem; |
| 620 | stack parsestk; |
| 621 | struct crossparaitem { |
| 622 | int type; /* currently c_lcont, c_quote or -1 */ |
| 623 | int seen_lcont, seen_quote; |
| 624 | }; |
| 625 | stack crossparastk; |
| 626 | word *indexword, *uword, *iword; |
| 627 | word *idxwordlist; |
| 628 | rdstring indexstr; |
| 629 | int index_downcase, index_visible, indexing; |
| 630 | const rdstring nullrs = { 0, 0, NULL }; |
| 631 | wchar_t uchr; |
| 632 | |
| 633 | t.text = NULL; |
| 634 | t.origtext = NULL; |
| 635 | already = FALSE; |
| 636 | |
| 637 | crossparastk = stk_new(); |
| 638 | |
| 639 | /* |
| 640 | * Loop on each paragraph. |
| 641 | */ |
| 642 | while (1) { |
| 643 | int start_cmd = c__invalid; |
| 644 | par.words = NULL; |
| 645 | par.keyword = NULL; |
| 646 | par.origkeyword = NULL; |
| 647 | whptr = &par.words; |
| 648 | |
| 649 | /* |
| 650 | * Get a token. |
| 651 | */ |
| 652 | do { |
| 653 | if (!already) { |
| 654 | dtor(t), t = get_token(in); |
| 655 | } |
| 656 | already = FALSE; |
| 657 | } while (t.type == tok_eop); |
| 658 | if (t.type == tok_eof) |
| 659 | break; |
| 660 | |
| 661 | /* |
| 662 | * Parse code paragraphs separately. |
| 663 | */ |
| 664 | if (t.type == tok_cmd && t.cmd == c_c && !isbrace(in)) { |
| 665 | int wtype = word_WeakCode; |
| 666 | |
| 667 | par.type = para_Code; |
| 668 | par.fpos = t.pos; |
| 669 | while (1) { |
| 670 | dtor(t), t = get_codepar_token(in); |
| 671 | wd.type = wtype; |
| 672 | wd.breaks = FALSE; /* shouldn't need this... */ |
| 673 | wd.text = ustrdup(t.text); |
| 674 | wd.alt = NULL; |
| 675 | wd.fpos = t.pos; |
| 676 | addword(wd, &whptr); |
| 677 | dtor(t), t = get_token(in); |
| 678 | if (t.type == tok_white) { |
| 679 | /* |
| 680 | * The newline after a code-paragraph line |
| 681 | */ |
| 682 | dtor(t), t = get_token(in); |
| 683 | } |
| 684 | if (t.type == tok_eop || t.type == tok_eof || |
| 685 | t.type == tok_rbrace) { /* might be } terminating \lcont */ |
| 686 | if (t.type == tok_rbrace) |
| 687 | already = TRUE; |
| 688 | break; |
| 689 | } else if (t.type == tok_cmd && t.cmd == c_c) { |
| 690 | wtype = word_WeakCode; |
| 691 | } else if (t.type == tok_cmd && t.cmd == c_e && |
| 692 | wtype == word_WeakCode) { |
| 693 | wtype = word_Emph; |
| 694 | } else { |
| 695 | error(err_brokencodepara, &t.pos); |
| 696 | prev_para_type = par.type; |
| 697 | addpara(par, ret); |
| 698 | while (t.type != tok_eop) /* error recovery: */ |
| 699 | dtor(t), t = get_token(in); /* eat rest of paragraph */ |
| 700 | goto codeparabroken; /* ick, but such is life */ |
| 701 | } |
| 702 | } |
| 703 | prev_para_type = par.type; |
| 704 | addpara(par, ret); |
| 705 | codeparabroken: |
| 706 | continue; |
| 707 | } |
| 708 | |
| 709 | /* |
| 710 | * Spot the special commands that define a grouping of more |
| 711 | * than one paragraph, and also the closing braces that |
| 712 | * finish them. |
| 713 | */ |
| 714 | if (t.type == tok_cmd && |
| 715 | (t.cmd == c_lcont || t.cmd == c_quote)) { |
| 716 | struct crossparaitem *sitem, *stop; |
| 717 | int cmd = t.cmd; |
| 718 | |
| 719 | /* |
| 720 | * Expect, and swallow, an open brace. |
| 721 | */ |
| 722 | dtor(t), t = get_token(in); |
| 723 | if (t.type != tok_lbrace) { |
| 724 | error(err_explbr, &t.pos); |
| 725 | continue; |
| 726 | } |
| 727 | |
| 728 | /* |
| 729 | * Also expect, and swallow, any whitespace after that |
| 730 | * (a newline before a code paragraph wouldn't be |
| 731 | * surprising). |
| 732 | */ |
| 733 | do { |
| 734 | dtor(t), t = get_token(in); |
| 735 | } while (t.type == tok_white); |
| 736 | already = TRUE; |
| 737 | |
| 738 | if (cmd == c_lcont) { |
| 739 | /* |
| 740 | * \lcont causes a continuation of a list item into |
| 741 | * multiple paragraphs (which may in turn contain |
| 742 | * nested lists, code paras etc). Hence, the previous |
| 743 | * paragraph must be of a list type. |
| 744 | */ |
| 745 | sitem = snew(struct crossparaitem); |
| 746 | stop = (struct crossparaitem *)stk_top(crossparastk); |
| 747 | if (stop) |
| 748 | *sitem = *stop; |
| 749 | else |
| 750 | sitem->seen_quote = sitem->seen_lcont = 0; |
| 751 | |
| 752 | if (prev_para_type == para_Bullet || |
| 753 | prev_para_type == para_NumberedList || |
| 754 | prev_para_type == para_Description) { |
| 755 | sitem->type = c_lcont; |
| 756 | sitem->seen_lcont = 1; |
| 757 | par.type = para_LcontPush; |
| 758 | prev_para_type = par.type; |
| 759 | addpara(par, ret); |
| 760 | } else { |
| 761 | /* |
| 762 | * Push a null item on the cross-para stack so that |
| 763 | * when we see the corresponding closing brace we |
| 764 | * don't give a cascade error. |
| 765 | */ |
| 766 | sitem->type = -1; |
| 767 | error(err_misplacedlcont, &t.pos); |
| 768 | } |
| 769 | } else { |
| 770 | /* |
| 771 | * \quote causes a group of paragraphs to be |
| 772 | * block-quoted (typically they will be indented a |
| 773 | * bit). |
| 774 | */ |
| 775 | sitem = snew(struct crossparaitem); |
| 776 | stop = (struct crossparaitem *)stk_top(crossparastk); |
| 777 | if (stop) |
| 778 | *sitem = *stop; |
| 779 | else |
| 780 | sitem->seen_quote = sitem->seen_lcont = 0; |
| 781 | sitem->type = c_quote; |
| 782 | sitem->seen_quote = 1; |
| 783 | par.type = para_QuotePush; |
| 784 | prev_para_type = par.type; |
| 785 | addpara(par, ret); |
| 786 | } |
| 787 | stk_push(crossparastk, sitem); |
| 788 | continue; |
| 789 | } else if (t.type == tok_rbrace) { |
| 790 | struct crossparaitem *sitem = stk_pop(crossparastk); |
| 791 | if (!sitem) |
| 792 | error(err_unexbrace, &t.pos); |
| 793 | else { |
| 794 | switch (sitem->type) { |
| 795 | case c_lcont: |
| 796 | par.type = para_LcontPop; |
| 797 | prev_para_type = par.type; |
| 798 | addpara(par, ret); |
| 799 | break; |
| 800 | case c_quote: |
| 801 | par.type = para_QuotePop; |
| 802 | prev_para_type = par.type; |
| 803 | addpara(par, ret); |
| 804 | break; |
| 805 | } |
| 806 | sfree(sitem); |
| 807 | } |
| 808 | continue; |
| 809 | } |
| 810 | |
| 811 | while (t.type == tok_cmd && |
| 812 | macrolookup(macros, in, t.text, &t.pos)) { |
| 813 | dtor(t), t = get_token(in); |
| 814 | } |
| 815 | |
| 816 | /* |
| 817 | * This token begins a paragraph. See if it's one of the |
| 818 | * special commands that define a paragraph type. |
| 819 | * |
| 820 | * (note that \# is special in a way, and \nocite takes no |
| 821 | * text) |
| 822 | */ |
| 823 | par.type = para_Normal; |
| 824 | if (t.type == tok_cmd) { |
| 825 | int needkw; |
| 826 | int is_macro = FALSE; |
| 827 | |
| 828 | par.fpos = t.pos; |
| 829 | switch (t.cmd) { |
| 830 | default: |
| 831 | needkw = -1; |
| 832 | break; |
| 833 | case c__invalid: |
| 834 | error(err_badparatype, t.text, &t.pos); |
| 835 | needkw = 4; |
| 836 | break; |
| 837 | case c__comment: |
| 838 | if (isbrace(in)) |
| 839 | break; /* `\#{': isn't a comment para */ |
| 840 | do { |
| 841 | dtor(t), t = get_token(in); |
| 842 | } while (t.type != tok_eop && t.type != tok_eof); |
| 843 | continue; /* next paragraph */ |
| 844 | /* |
| 845 | * `needkw' values: |
| 846 | * |
| 847 | * 1 -- exactly one keyword |
| 848 | * 2 -- at least one keyword |
| 849 | * 4 -- any number of keywords including zero |
| 850 | * 8 -- at least one keyword and then nothing else |
| 851 | * 16 -- nothing at all! no keywords, no body |
| 852 | * 32 -- no keywords at all |
| 853 | */ |
| 854 | case c_A: needkw = 2; par.type = para_Appendix; break; |
| 855 | case c_B: needkw = 2; par.type = para_Biblio; break; |
| 856 | case c_BR: needkw = 1; par.type = para_BR; |
| 857 | start_cmd = c_BR; break; |
| 858 | case c_C: needkw = 2; par.type = para_Chapter; break; |
| 859 | case c_H: needkw = 2; par.type = para_Heading; |
| 860 | par.aux = 0; |
| 861 | break; |
| 862 | case c_IM: needkw = 2; par.type = para_IM; |
| 863 | start_cmd = c_IM; break; |
| 864 | case c_S: needkw = 2; par.type = para_Subsect; |
| 865 | par.aux = t.aux; break; |
| 866 | case c_U: needkw = 32; par.type = para_UnnumberedChapter; break; |
| 867 | /* For \b and \n the keyword is optional */ |
| 868 | case c_b: needkw = 4; par.type = para_Bullet; break; |
| 869 | case c_dt: needkw = 4; par.type = para_DescribedThing; break; |
| 870 | case c_dd: needkw = 4; par.type = para_Description; break; |
| 871 | case c_n: needkw = 4; par.type = para_NumberedList; break; |
| 872 | case c_cfg: needkw = 8; par.type = para_Config; |
| 873 | start_cmd = c_cfg; break; |
| 874 | case c_copyright: needkw = 32; par.type = para_Copyright; break; |
| 875 | case c_define: is_macro = TRUE; needkw = 1; break; |
| 876 | /* For \nocite the keyword is _everything_ */ |
| 877 | case c_nocite: needkw = 8; par.type = para_NoCite; break; |
| 878 | case c_preamble: needkw = 32; par.type = para_Normal; break; |
| 879 | case c_rule: needkw = 16; par.type = para_Rule; break; |
| 880 | case c_title: needkw = 32; par.type = para_Title; break; |
| 881 | case c_versionid: needkw = 32; par.type = para_VersionID; break; |
| 882 | } |
| 883 | |
| 884 | if (par.type == para_Chapter || |
| 885 | par.type == para_Heading || |
| 886 | par.type == para_Subsect || |
| 887 | par.type == para_Appendix || |
| 888 | par.type == para_UnnumberedChapter) { |
| 889 | struct crossparaitem *sitem = stk_top(crossparastk); |
| 890 | if (sitem && (sitem->seen_lcont || sitem->seen_quote)) { |
| 891 | error(err_sectmarkerinblock, |
| 892 | &t.pos, |
| 893 | (sitem->seen_lcont ? "lcont" : "quote")); |
| 894 | } |
| 895 | } |
| 896 | |
| 897 | if (needkw > 0) { |
| 898 | rdstring rs = { 0, 0, NULL }; |
| 899 | rdstringc rsc = { 0, 0, NULL }; |
| 900 | int nkeys = 0; |
| 901 | filepos fp; |
| 902 | |
| 903 | /* Get keywords. */ |
| 904 | dtor(t), t = get_token(in); |
| 905 | fp = t.pos; |
| 906 | while (t.type == tok_lbrace || |
| 907 | (t.type == tok_white && (needkw & 24))) { |
| 908 | /* |
| 909 | * In paragraph types which can't accept any |
| 910 | * body text (such as \cfg), we are lenient |
| 911 | * about whitespace between keywords. This is |
| 912 | * important for \cfg in particular since it |
| 913 | * can often have many keywords which are long |
| 914 | * pieces of text, so it's useful to permit the |
| 915 | * user to wrap the line between them. |
| 916 | */ |
| 917 | if (t.type == tok_white) { |
| 918 | dtor(t), t = get_token(in); /* eat the space */ |
| 919 | continue; |
| 920 | } |
| 921 | /* This is a keyword. */ |
| 922 | nkeys++; |
| 923 | /* FIXME: there will be bugs if anyone specifies an |
| 924 | * empty keyword (\foo{}), so trap this case. */ |
| 925 | while (dtor(t), t = get_token(in), |
| 926 | t.type == tok_word || |
| 927 | t.type == tok_white || |
| 928 | (t.type == tok_cmd && t.cmd == c__nbsp) || |
| 929 | (t.type == tok_cmd && t.cmd == c__escaped) || |
| 930 | (t.type == tok_cmd && t.cmd == c_u)) { |
| 931 | if (t.type == tok_white || |
| 932 | (t.type == tok_cmd && t.cmd == c__nbsp)) { |
| 933 | rdadd(&rs, ' '); |
| 934 | rdaddc(&rsc, ' '); |
| 935 | } else if (t.type == tok_cmd && t.cmd == c_u) { |
| 936 | rdadd(&rs, t.aux); |
| 937 | rdaddc(&rsc, '\\'); |
| 938 | rdaddsc(&rsc, t.origtext); |
| 939 | } else { |
| 940 | rdadds(&rs, t.text); |
| 941 | rdaddsc(&rsc, t.origtext); |
| 942 | } |
| 943 | } |
| 944 | if (t.type != tok_rbrace) { |
| 945 | error(err_kwunclosed, &t.pos); |
| 946 | continue; |
| 947 | } |
| 948 | rdadd(&rs, 0); /* add string terminator */ |
| 949 | rdaddc(&rsc, 0); /* add string terminator */ |
| 950 | dtor(t), t = get_token(in); /* eat right brace */ |
| 951 | } |
| 952 | |
| 953 | rdadd(&rs, 0); /* add string terminator */ |
| 954 | rdaddc(&rsc, 0); /* add string terminator */ |
| 955 | |
| 956 | /* See whether we have the right number of keywords. */ |
| 957 | if ((needkw & 48) && nkeys > 0) |
| 958 | error(err_kwillegal, &fp); |
| 959 | if ((needkw & 11) && nkeys == 0) |
| 960 | error(err_kwexpected, &fp); |
| 961 | if ((needkw & 5) && nkeys > 1) |
| 962 | error(err_kwtoomany, &fp); |
| 963 | |
| 964 | if (is_macro) { |
| 965 | /* |
| 966 | * Macro definition. Get the rest of the line |
| 967 | * as a code-paragraph token, repeatedly until |
| 968 | * there's nothing more left of it. Separate |
| 969 | * with newlines. |
| 970 | */ |
| 971 | rdstring macrotext = { 0, 0, NULL }; |
| 972 | while (1) { |
| 973 | dtor(t), t = get_codepar_token(in); |
| 974 | if (macrotext.pos > 0) |
| 975 | rdadd(¯otext, L'\n'); |
| 976 | rdadds(¯otext, t.text); |
| 977 | dtor(t), t = get_token(in); |
| 978 | if (t.type == tok_eop) break; |
| 979 | } |
| 980 | macrodef(macros, rs.text, macrotext.text, fp); |
| 981 | continue; /* next paragraph */ |
| 982 | } |
| 983 | |
| 984 | par.keyword = rdtrim(&rs); |
| 985 | par.origkeyword = rdtrimc(&rsc); |
| 986 | |
| 987 | /* Move to EOP in case of needkw==8 or 16 (no body) */ |
| 988 | if (needkw & 24) { |
| 989 | /* We allow whitespace even when we expect no para body */ |
| 990 | while (t.type == tok_white) |
| 991 | dtor(t), t = get_token(in); |
| 992 | if (t.type != tok_eop && t.type != tok_eof && |
| 993 | (start_cmd == c__invalid || |
| 994 | t.type != tok_cmd || t.cmd != start_cmd)) { |
| 995 | error(err_bodyillegal, &t.pos); |
| 996 | /* Error recovery: eat the rest of the paragraph */ |
| 997 | while (t.type != tok_eop && t.type != tok_eof && |
| 998 | (start_cmd == c__invalid || |
| 999 | t.type != tok_cmd || t.cmd != start_cmd)) |
| 1000 | dtor(t), t = get_token(in); |
| 1001 | } |
| 1002 | if (t.type == tok_cmd) |
| 1003 | already = TRUE;/* inhibit get_token at top of loop */ |
| 1004 | prev_para_type = par.type; |
| 1005 | addpara(par, ret); |
| 1006 | |
| 1007 | if (par.type == para_Config) { |
| 1008 | input_configure(in, &par); |
| 1009 | } |
| 1010 | continue; /* next paragraph */ |
| 1011 | } |
| 1012 | } |
| 1013 | } |
| 1014 | |
| 1015 | /* |
| 1016 | * Now read the actual paragraph, word by word, adding to |
| 1017 | * the paragraph list. |
| 1018 | * |
| 1019 | * Mid-paragraph commands: |
| 1020 | * |
| 1021 | * \K \k |
| 1022 | * \c \cw \cq |
| 1023 | * \e |
| 1024 | * \i \ii |
| 1025 | * \I |
| 1026 | * \q |
| 1027 | * \u |
| 1028 | * \W |
| 1029 | * \date |
| 1030 | * \\ \{ \} |
| 1031 | */ |
| 1032 | parsestk = stk_new(); |
| 1033 | style = word_Normal; |
| 1034 | spcstyle = word_WhiteSpace; |
| 1035 | indexing = FALSE; |
| 1036 | seenwhite = TRUE; |
| 1037 | while (t.type != tok_eop && t.type != tok_eof) { |
| 1038 | iswhite = FALSE; |
| 1039 | already = FALSE; |
| 1040 | |
| 1041 | /* Handle implicit paragraph breaks after \IM, \BR etc */ |
| 1042 | if (start_cmd != c__invalid && |
| 1043 | t.type == tok_cmd && t.cmd == start_cmd) { |
| 1044 | already = TRUE; /* inhibit get_token at top of loop */ |
| 1045 | break; |
| 1046 | } |
| 1047 | |
| 1048 | if (t.type == tok_cmd && t.cmd == c__nop) { |
| 1049 | dtor(t), t = get_token(in); |
| 1050 | continue; /* do nothing! */ |
| 1051 | } |
| 1052 | |
| 1053 | if (t.type == tok_cmd && t.cmd == c__escaped) { |
| 1054 | t.type = tok_word; /* nice and simple */ |
| 1055 | t.aux = 0; /* even if `\-' - nonbreaking! */ |
| 1056 | } |
| 1057 | if (t.type == tok_cmd && t.cmd == c__nbsp) { |
| 1058 | t.type = tok_word; /* nice and simple */ |
| 1059 | sfree(t.text); |
| 1060 | t.text = ustrdup(L" "); /* text is ` ' not `_' */ |
| 1061 | t.aux = 0; /* (nonbreaking) */ |
| 1062 | } |
| 1063 | switch (t.type) { |
| 1064 | case tok_white: |
| 1065 | if (whptr == &par.words) |
| 1066 | break; /* strip whitespace at start of para */ |
| 1067 | wd.text = NULL; |
| 1068 | wd.type = spcstyle; |
| 1069 | wd.alt = NULL; |
| 1070 | wd.aux = 0; |
| 1071 | wd.fpos = t.pos; |
| 1072 | wd.breaks = FALSE; |
| 1073 | |
| 1074 | /* |
| 1075 | * Inhibit use of whitespace if it's (probably the |
| 1076 | * newline) before a repeat \IM / \BR type |
| 1077 | * directive. |
| 1078 | */ |
| 1079 | if (start_cmd != c__invalid) { |
| 1080 | dtor(t), t = get_token(in); |
| 1081 | already = TRUE; |
| 1082 | if (t.type == tok_cmd && t.cmd == start_cmd) |
| 1083 | break; |
| 1084 | } |
| 1085 | |
| 1086 | if (indexing) |
| 1087 | rdadd(&indexstr, ' '); |
| 1088 | if (!indexing || index_visible) |
| 1089 | addword(wd, &whptr); |
| 1090 | if (indexing) |
| 1091 | addword(wd, &idximplicit); |
| 1092 | iswhite = TRUE; |
| 1093 | break; |
| 1094 | case tok_word: |
| 1095 | if (indexing) |
| 1096 | rdadds(&indexstr, t.text); |
| 1097 | wd.type = style; |
| 1098 | wd.alt = NULL; |
| 1099 | wd.aux = 0; |
| 1100 | wd.fpos = t.pos; |
| 1101 | wd.breaks = t.aux; |
| 1102 | if (!indexing || index_visible) { |
| 1103 | wd.text = ustrdup(t.text); |
| 1104 | addword(wd, &whptr); |
| 1105 | } |
| 1106 | if (indexing) { |
| 1107 | wd.text = ustrdup(t.text); |
| 1108 | addword(wd, &idximplicit); |
| 1109 | } |
| 1110 | break; |
| 1111 | case tok_lbrace: |
| 1112 | error(err_unexbrace, &t.pos); |
| 1113 | /* Error recovery: push nop */ |
| 1114 | sitem = snew(struct stack_item); |
| 1115 | sitem->type = stack_nop; |
| 1116 | sitem->fpos = t.pos; |
| 1117 | stk_push(parsestk, sitem); |
| 1118 | break; |
| 1119 | case tok_rbrace: |
| 1120 | sitem = stk_pop(parsestk); |
| 1121 | if (!sitem) { |
| 1122 | /* |
| 1123 | * This closing brace could have been an |
| 1124 | * indication that the cross-paragraph stack |
| 1125 | * wants popping. Accordingly, we treat it here |
| 1126 | * as an indication that the paragraph is over. |
| 1127 | */ |
| 1128 | already = TRUE; |
| 1129 | goto finished_para; |
| 1130 | } else { |
| 1131 | if (sitem->type & stack_ualt) { |
| 1132 | whptr = sitem->whptr; |
| 1133 | idximplicit = sitem->idximplicit; |
| 1134 | } |
| 1135 | if (sitem->type & stack_style) { |
| 1136 | style = word_Normal; |
| 1137 | spcstyle = word_WhiteSpace; |
| 1138 | } |
| 1139 | if (sitem->type & stack_idx) { |
| 1140 | indexword->text = ustrdup(indexstr.text); |
| 1141 | if (index_downcase) { |
| 1142 | word *w; |
| 1143 | |
| 1144 | ustrlow(indexword->text); |
| 1145 | ustrlow(indexstr.text); |
| 1146 | |
| 1147 | for (w = idxwordlist; w; w = w->next) |
| 1148 | if (w->text) |
| 1149 | ustrlow(w->text); |
| 1150 | } |
| 1151 | indexing = FALSE; |
| 1152 | rdadd(&indexstr, L'\0'); |
| 1153 | index_merge(idx, FALSE, indexstr.text, |
| 1154 | idxwordlist, &sitem->fpos); |
| 1155 | sfree(indexstr.text); |
| 1156 | } |
| 1157 | if (sitem->type & stack_hyper) { |
| 1158 | wd.text = NULL; |
| 1159 | wd.type = word_HyperEnd; |
| 1160 | wd.alt = NULL; |
| 1161 | wd.aux = 0; |
| 1162 | wd.fpos = t.pos; |
| 1163 | wd.breaks = FALSE; |
| 1164 | if (!indexing || index_visible) |
| 1165 | addword(wd, &whptr); |
| 1166 | if (indexing) |
| 1167 | addword(wd, &idximplicit); |
| 1168 | } |
| 1169 | if (sitem->type & stack_quote) { |
| 1170 | wd.text = NULL; |
| 1171 | wd.type = toquotestyle(style); |
| 1172 | wd.alt = NULL; |
| 1173 | wd.aux = quote_Close; |
| 1174 | wd.fpos = t.pos; |
| 1175 | wd.breaks = FALSE; |
| 1176 | if (!indexing || index_visible) |
| 1177 | addword(wd, &whptr); |
| 1178 | if (indexing) { |
| 1179 | rdadd(&indexstr, L'"'); |
| 1180 | addword(wd, &idximplicit); |
| 1181 | } |
| 1182 | } |
| 1183 | } |
| 1184 | sfree(sitem); |
| 1185 | break; |
| 1186 | case tok_cmd: |
| 1187 | switch (t.cmd) { |
| 1188 | case c__comment: |
| 1189 | /* |
| 1190 | * In-paragraph comment: \#{ balanced braces } |
| 1191 | * |
| 1192 | * Anything goes here; even tok_eop. We should |
| 1193 | * eat whitespace after the close brace _if_ |
| 1194 | * there was whitespace before the \#. |
| 1195 | */ |
| 1196 | dtor(t), t = get_token(in); |
| 1197 | if (t.type != tok_lbrace) { |
| 1198 | error(err_explbr, &t.pos); |
| 1199 | } else { |
| 1200 | int braces = 1; |
| 1201 | while (braces > 0) { |
| 1202 | dtor(t), t = get_token(in); |
| 1203 | if (t.type == tok_lbrace) |
| 1204 | braces++; |
| 1205 | else if (t.type == tok_rbrace) |
| 1206 | braces--; |
| 1207 | else if (t.type == tok_eof) { |
| 1208 | error(err_commenteof, &t.pos); |
| 1209 | break; |
| 1210 | } |
| 1211 | } |
| 1212 | } |
| 1213 | if (seenwhite) { |
| 1214 | already = TRUE; |
| 1215 | dtor(t), t = get_token(in); |
| 1216 | if (t.type == tok_white) { |
| 1217 | iswhite = TRUE; |
| 1218 | already = FALSE; |
| 1219 | } |
| 1220 | } |
| 1221 | break; |
| 1222 | case c_q: |
| 1223 | case c_cq: |
| 1224 | type = t.cmd; |
| 1225 | dtor(t), t = get_token(in); |
| 1226 | if (t.type != tok_lbrace) { |
| 1227 | error(err_explbr, &t.pos); |
| 1228 | } else { |
| 1229 | /* |
| 1230 | * Enforce that \q may not be used anywhere |
| 1231 | * within \c. (It shouldn't be necessary |
| 1232 | * since the whole point of \c should be |
| 1233 | * that the user wants to exercise exact |
| 1234 | * control over the glyphs used, and |
| 1235 | * forbidding it has the useful effect of |
| 1236 | * relieving some backends of having to |
| 1237 | * make difficult decisions.) |
| 1238 | */ |
| 1239 | int stype; |
| 1240 | |
| 1241 | if (style != word_Code && style != word_WeakCode) { |
| 1242 | wd.text = NULL; |
| 1243 | wd.type = toquotestyle(style); |
| 1244 | wd.alt = NULL; |
| 1245 | wd.aux = quote_Open; |
| 1246 | wd.fpos = t.pos; |
| 1247 | wd.breaks = FALSE; |
| 1248 | if (!indexing || index_visible) |
| 1249 | addword(wd, &whptr); |
| 1250 | if (indexing) { |
| 1251 | rdadd(&indexstr, L'"'); |
| 1252 | addword(wd, &idximplicit); |
| 1253 | } |
| 1254 | stype = stack_quote; |
| 1255 | } else { |
| 1256 | error(err_codequote, &t.pos); |
| 1257 | stype = stack_nop; |
| 1258 | } |
| 1259 | sitem = snew(struct stack_item); |
| 1260 | sitem->fpos = t.pos; |
| 1261 | sitem->type = stype; |
| 1262 | if (type == c_cq) { |
| 1263 | if (style != word_Normal) { |
| 1264 | error(err_nestedstyles, &t.pos); |
| 1265 | } else { |
| 1266 | style = word_WeakCode; |
| 1267 | spcstyle = tospacestyle(style); |
| 1268 | sitem->type |= stack_style; |
| 1269 | } |
| 1270 | } |
| 1271 | stk_push(parsestk, sitem); |
| 1272 | } |
| 1273 | break; |
| 1274 | case c_K: |
| 1275 | case c_k: |
| 1276 | case c_W: |
| 1277 | case c_date: |
| 1278 | /* |
| 1279 | * Keyword, hyperlink, or \date. We expect a |
| 1280 | * left brace, some text, and then a right |
| 1281 | * brace. No nesting; no arguments. |
| 1282 | */ |
| 1283 | wd.fpos = t.pos; |
| 1284 | wd.breaks = FALSE; |
| 1285 | if (t.cmd == c_K) |
| 1286 | wd.type = word_UpperXref; |
| 1287 | else if (t.cmd == c_k) |
| 1288 | wd.type = word_LowerXref; |
| 1289 | else if (t.cmd == c_W) |
| 1290 | wd.type = word_HyperLink; |
| 1291 | else |
| 1292 | wd.type = word_Normal; |
| 1293 | dtor(t), t = get_token(in); |
| 1294 | if (t.type != tok_lbrace) { |
| 1295 | if (wd.type == word_Normal) { |
| 1296 | time_t thetime = time(NULL); |
| 1297 | struct tm *broken = localtime(&thetime); |
| 1298 | already = TRUE; |
| 1299 | wdtext = ustrftime(NULL, broken); |
| 1300 | wd.type = style; |
| 1301 | } else { |
| 1302 | error(err_explbr, &t.pos); |
| 1303 | wdtext = NULL; |
| 1304 | } |
| 1305 | } else { |
| 1306 | rdstring rs = { 0, 0, NULL }; |
| 1307 | while (dtor(t), t = get_token(in), |
| 1308 | t.type == tok_word || t.type == tok_white) { |
| 1309 | if (t.type == tok_white) |
| 1310 | rdadd(&rs, ' '); |
| 1311 | else |
| 1312 | rdadds(&rs, t.text); |
| 1313 | } |
| 1314 | if (wd.type == word_Normal) { |
| 1315 | time_t thetime = time(NULL); |
| 1316 | struct tm *broken = localtime(&thetime); |
| 1317 | wdtext = ustrftime(rs.text, broken); |
| 1318 | wd.type = style; |
| 1319 | } else { |
| 1320 | wdtext = ustrdup(rs.text); |
| 1321 | } |
| 1322 | sfree(rs.text); |
| 1323 | if (t.type != tok_rbrace) { |
| 1324 | error(err_kwexprbr, &t.pos); |
| 1325 | } |
| 1326 | } |
| 1327 | wd.alt = NULL; |
| 1328 | wd.aux = 0; |
| 1329 | if (!indexing || index_visible) { |
| 1330 | wd.text = ustrdup(wdtext); |
| 1331 | addword(wd, &whptr); |
| 1332 | } |
| 1333 | if (indexing) { |
| 1334 | wd.text = ustrdup(wdtext); |
| 1335 | addword(wd, &idximplicit); |
| 1336 | } |
| 1337 | sfree(wdtext); |
| 1338 | if (wd.type == word_HyperLink) { |
| 1339 | /* |
| 1340 | * Hyperlinks are different: they then |
| 1341 | * expect another left brace, to begin |
| 1342 | * delimiting the text marked by the link. |
| 1343 | */ |
| 1344 | dtor(t), t = get_token(in); |
| 1345 | sitem = snew(struct stack_item); |
| 1346 | sitem->fpos = wd.fpos; |
| 1347 | sitem->type = stack_hyper; |
| 1348 | /* |
| 1349 | * Special cases: \W{}\i, \W{}\ii |
| 1350 | */ |
| 1351 | if (t.type == tok_cmd && |
| 1352 | (t.cmd == c_i || t.cmd == c_ii)) { |
| 1353 | if (indexing) { |
| 1354 | error(err_nestedindex, &t.pos); |
| 1355 | } else { |
| 1356 | /* Add an index-reference word with no |
| 1357 | * text as yet */ |
| 1358 | wd.type = word_IndexRef; |
| 1359 | wd.text = NULL; |
| 1360 | wd.alt = NULL; |
| 1361 | wd.aux = 0; |
| 1362 | wd.breaks = FALSE; |
| 1363 | indexword = addword(wd, &whptr); |
| 1364 | /* Set up a rdstring to read the |
| 1365 | * index text */ |
| 1366 | indexstr = nullrs; |
| 1367 | /* Flags so that we do the Right |
| 1368 | * Things with text */ |
| 1369 | index_visible = (type != c_I); |
| 1370 | index_downcase = (type == c_ii); |
| 1371 | indexing = TRUE; |
| 1372 | idxwordlist = NULL; |
| 1373 | idximplicit = &idxwordlist; |
| 1374 | |
| 1375 | sitem->type |= stack_idx; |
| 1376 | } |
| 1377 | dtor(t), t = get_token(in); |
| 1378 | } |
| 1379 | /* |
| 1380 | * Special cases: \W{}\c, \W{}\e, \W{}\cw |
| 1381 | */ |
| 1382 | if (t.type == tok_cmd && |
| 1383 | (t.cmd == c_e || t.cmd == c_c || t.cmd == c_cw)) { |
| 1384 | if (style != word_Normal) |
| 1385 | error(err_nestedstyles, &t.pos); |
| 1386 | else { |
| 1387 | style = (t.cmd == c_c ? word_Code : |
| 1388 | t.cmd == c_cw ? word_WeakCode : |
| 1389 | word_Emph); |
| 1390 | spcstyle = tospacestyle(style); |
| 1391 | sitem->type |= stack_style; |
| 1392 | } |
| 1393 | dtor(t), t = get_token(in); |
| 1394 | } |
| 1395 | if (t.type != tok_lbrace) { |
| 1396 | error(err_explbr, &t.pos); |
| 1397 | sfree(sitem); |
| 1398 | } else { |
| 1399 | stk_push(parsestk, sitem); |
| 1400 | } |
| 1401 | } |
| 1402 | break; |
| 1403 | case c_c: |
| 1404 | case c_cw: |
| 1405 | case c_e: |
| 1406 | type = t.cmd; |
| 1407 | if (style != word_Normal) { |
| 1408 | error(err_nestedstyles, &t.pos); |
| 1409 | /* Error recovery: eat lbrace, push nop. */ |
| 1410 | dtor(t), t = get_token(in); |
| 1411 | sitem = snew(struct stack_item); |
| 1412 | sitem->fpos = t.pos; |
| 1413 | sitem->type = stack_nop; |
| 1414 | stk_push(parsestk, sitem); |
| 1415 | } |
| 1416 | dtor(t), t = get_token(in); |
| 1417 | if (t.type != tok_lbrace) { |
| 1418 | error(err_explbr, &t.pos); |
| 1419 | } else { |
| 1420 | style = (type == c_c ? word_Code : |
| 1421 | type == c_cw ? word_WeakCode : |
| 1422 | word_Emph); |
| 1423 | spcstyle = tospacestyle(style); |
| 1424 | sitem = snew(struct stack_item); |
| 1425 | sitem->fpos = t.pos; |
| 1426 | sitem->type = stack_style; |
| 1427 | stk_push(parsestk, sitem); |
| 1428 | } |
| 1429 | break; |
| 1430 | case c_i: |
| 1431 | case c_ii: |
| 1432 | case c_I: |
| 1433 | type = t.cmd; |
| 1434 | if (indexing) { |
| 1435 | error(err_nestedindex, &t.pos); |
| 1436 | /* Error recovery: eat lbrace, push nop. */ |
| 1437 | dtor(t), t = get_token(in); |
| 1438 | sitem = snew(struct stack_item); |
| 1439 | sitem->fpos = t.pos; |
| 1440 | sitem->type = stack_nop; |
| 1441 | stk_push(parsestk, sitem); |
| 1442 | } |
| 1443 | sitem = snew(struct stack_item); |
| 1444 | sitem->fpos = t.pos; |
| 1445 | sitem->type = stack_idx; |
| 1446 | dtor(t), t = get_token(in); |
| 1447 | /* |
| 1448 | * Special cases: \i\c, \i\e, \i\cw |
| 1449 | */ |
| 1450 | wd.fpos = t.pos; |
| 1451 | if (t.type == tok_cmd && |
| 1452 | (t.cmd == c_e || t.cmd == c_c || t.cmd == c_cw)) { |
| 1453 | if (style != word_Normal) |
| 1454 | error(err_nestedstyles, &t.pos); |
| 1455 | else { |
| 1456 | style = (t.cmd == c_c ? word_Code : |
| 1457 | t.cmd == c_cw ? word_WeakCode : |
| 1458 | word_Emph); |
| 1459 | spcstyle = tospacestyle(style); |
| 1460 | sitem->type |= stack_style; |
| 1461 | } |
| 1462 | dtor(t), t = get_token(in); |
| 1463 | } |
| 1464 | if (t.type != tok_lbrace) { |
| 1465 | sfree(sitem); |
| 1466 | error(err_explbr, &t.pos); |
| 1467 | } else { |
| 1468 | /* Add an index-reference word with no text as yet */ |
| 1469 | wd.type = word_IndexRef; |
| 1470 | wd.text = NULL; |
| 1471 | wd.alt = NULL; |
| 1472 | wd.aux = 0; |
| 1473 | wd.breaks = FALSE; |
| 1474 | indexword = addword(wd, &whptr); |
| 1475 | /* Set up a rdstring to read the index text */ |
| 1476 | indexstr = nullrs; |
| 1477 | /* Flags so that we do the Right Things with text */ |
| 1478 | index_visible = (type != c_I); |
| 1479 | index_downcase = (type == c_ii); |
| 1480 | indexing = TRUE; |
| 1481 | idxwordlist = NULL; |
| 1482 | idximplicit = &idxwordlist; |
| 1483 | /* Stack item to close the indexing on exit */ |
| 1484 | stk_push(parsestk, sitem); |
| 1485 | } |
| 1486 | break; |
| 1487 | case c_u: |
| 1488 | uchr = t.aux; |
| 1489 | utext[0] = uchr; utext[1] = 0; |
| 1490 | wd.type = style; |
| 1491 | wd.breaks = FALSE; |
| 1492 | wd.alt = NULL; |
| 1493 | wd.aux = 0; |
| 1494 | wd.fpos = t.pos; |
| 1495 | if (!indexing || index_visible) { |
| 1496 | wd.text = ustrdup(utext); |
| 1497 | uword = addword(wd, &whptr); |
| 1498 | } else |
| 1499 | uword = NULL; |
| 1500 | if (indexing) { |
| 1501 | wd.text = ustrdup(utext); |
| 1502 | iword = addword(wd, &idximplicit); |
| 1503 | } else |
| 1504 | iword = NULL; |
| 1505 | dtor(t), t = get_token(in); |
| 1506 | if (t.type == tok_lbrace) { |
| 1507 | /* |
| 1508 | * \u with a left brace. Until the brace |
| 1509 | * closes, all further words go on a |
| 1510 | * sidetrack from the main thread of the |
| 1511 | * paragraph. |
| 1512 | */ |
| 1513 | sitem = snew(struct stack_item); |
| 1514 | sitem->fpos = t.pos; |
| 1515 | sitem->type = stack_ualt; |
| 1516 | sitem->whptr = whptr; |
| 1517 | sitem->idximplicit = idximplicit; |
| 1518 | stk_push(parsestk, sitem); |
| 1519 | whptr = uword ? &uword->alt : NULL; |
| 1520 | idximplicit = iword ? &iword->alt : NULL; |
| 1521 | } else { |
| 1522 | if (indexing) |
| 1523 | rdadd(&indexstr, uchr); |
| 1524 | already = TRUE; |
| 1525 | } |
| 1526 | break; |
| 1527 | default: |
| 1528 | if (!macrolookup(macros, in, t.text, &t.pos)) |
| 1529 | error(err_badmidcmd, t.text, &t.pos); |
| 1530 | break; |
| 1531 | } |
| 1532 | } |
| 1533 | if (!already) |
| 1534 | dtor(t), t = get_token(in); |
| 1535 | seenwhite = iswhite; |
| 1536 | } |
| 1537 | finished_para: |
| 1538 | /* Check the stack is empty */ |
| 1539 | if (stk_top(parsestk)) { |
| 1540 | while ((sitem = stk_pop(parsestk))) |
| 1541 | sfree(sitem); |
| 1542 | error(err_missingrbrace, &t.pos); |
| 1543 | } |
| 1544 | stk_free(parsestk); |
| 1545 | prev_para_type = par.type; |
| 1546 | /* |
| 1547 | * Before we add the paragraph to the output list, we |
| 1548 | * should check that there was any text in it at all; there |
| 1549 | * might not be if (for example) the paragraph contained |
| 1550 | * nothing but an unrecognised command sequence, and if we |
| 1551 | * put an empty paragraph on the list it may confuse the |
| 1552 | * back ends later on. |
| 1553 | */ |
| 1554 | if (par.words) { |
| 1555 | addpara(par, ret); |
| 1556 | } |
| 1557 | if (t.type == tok_eof) |
| 1558 | already = TRUE; |
| 1559 | } |
| 1560 | |
| 1561 | if (stk_top(crossparastk)) { |
| 1562 | void *p; |
| 1563 | |
| 1564 | error(err_missingrbrace2, &t.pos); |
| 1565 | while ((p = stk_pop(crossparastk))) |
| 1566 | sfree(p); |
| 1567 | } |
| 1568 | |
| 1569 | /* |
| 1570 | * We break to here rather than returning, because otherwise |
| 1571 | * this cleanup doesn't happen. |
| 1572 | */ |
| 1573 | dtor(t); |
| 1574 | |
| 1575 | stk_free(crossparastk); |
| 1576 | } |
| 1577 | |
| 1578 | struct { |
| 1579 | char const *magic; |
| 1580 | size_t nmagic; |
| 1581 | void (*reader)(input *); |
| 1582 | } magics[] = { |
| 1583 | { "%!FontType1-", 12, &read_pfa_file }, |
| 1584 | { "%!PS-AdobeFont-", 15, &read_pfa_file }, |
| 1585 | { "StartFontMetrics", 16, &read_afm_file }, |
| 1586 | }; |
| 1587 | |
| 1588 | paragraph *read_input(input *in, indexdata *idx) { |
| 1589 | paragraph *head = NULL; |
| 1590 | paragraph **hptr = &head; |
| 1591 | tree234 *macros; |
| 1592 | char mag[16]; |
| 1593 | size_t len, i; |
| 1594 | void (*reader)(input *); |
| 1595 | |
| 1596 | macros = newtree234(macrocmp); |
| 1597 | |
| 1598 | while (in->currindex < in->nfiles) { |
| 1599 | in->currfp = fopen(in->filenames[in->currindex], "r"); |
| 1600 | if (in->currfp) { |
| 1601 | setpos(in, in->filenames[in->currindex]); |
| 1602 | in->charset = in->defcharset; |
| 1603 | in->csstate = charset_init_state; |
| 1604 | in->wcpos = in->nwc = 0; |
| 1605 | in->pushback_chars = NULL; |
| 1606 | reader = NULL; |
| 1607 | len = fread(mag, 1, sizeof(mag), in->currfp); |
| 1608 | for (i = 0; i < lenof(magics); i++) { |
| 1609 | if (len >= magics[i].nmagic && |
| 1610 | memcmp(mag, magics[i].magic, magics[i].nmagic) == 0) { |
| 1611 | reader = magics[i].reader; |
| 1612 | break; |
| 1613 | } |
| 1614 | } |
| 1615 | rewind(in->currfp); |
| 1616 | if (reader == NULL) |
| 1617 | read_file(&hptr, in, idx, macros); |
| 1618 | else |
| 1619 | (*reader)(in); |
| 1620 | } |
| 1621 | in->currindex++; |
| 1622 | } |
| 1623 | |
| 1624 | macrocleanup(macros); |
| 1625 | |
| 1626 | return head; |
| 1627 | } |