1 /* $OpenBSD: out.c,v 1.54 2021/09/07 17:05:30 schwarze Exp $ */ 2 /* 3 * Copyright (c) 2009, 2010, 2011 Kristaps Dzonsons <kristaps@bsd.lv> 4 * Copyright (c) 2011, 2014, 2015, 2017, 2018, 2019, 2021 5 * Ingo Schwarze <schwarze@openbsd.org> 6 * 7 * Permission to use, copy, modify, and distribute this software for any 8 * purpose with or without fee is hereby granted, provided that the above 9 * copyright notice and this permission notice appear in all copies. 10 * 11 * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES 12 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF 13 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR 14 * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES 15 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN 16 * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF 17 * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. 18 */ 19 #include <sys/types.h> 20 21 #include <assert.h> 22 #include <ctype.h> 23 #include <stdint.h> 24 #include <stdio.h> 25 #include <stdlib.h> 26 #include <string.h> 27 #include <time.h> 28 29 #include "mandoc_aux.h" 30 #include "mandoc.h" 31 #include "tbl.h" 32 #include "out.h" 33 34 struct tbl_colgroup { 35 struct tbl_colgroup *next; 36 size_t wanted; 37 int startcol; 38 int endcol; 39 }; 40 41 static size_t tblcalc_data(struct rofftbl *, struct roffcol *, 42 const struct tbl_opts *, const struct tbl_dat *, 43 size_t); 44 static size_t tblcalc_literal(struct rofftbl *, struct roffcol *, 45 const struct tbl_dat *, size_t); 46 static size_t tblcalc_number(struct rofftbl *, struct roffcol *, 47 const struct tbl_opts *, const struct tbl_dat *); 48 49 50 /* 51 * Parse the *src string and store a scaling unit into *dst. 52 * If the string doesn't specify the unit, use the default. 53 * If no default is specified, fail. 54 * Return a pointer to the byte after the last byte used, 55 * or NULL on total failure. 56 */ 57 const char * 58 a2roffsu(const char *src, struct roffsu *dst, enum roffscale def) 59 { 60 char *endptr; 61 62 dst->unit = def == SCALE_MAX ? SCALE_BU : def; 63 dst->scale = strtod(src, &endptr); 64 if (endptr == src) 65 return NULL; 66 67 switch (*endptr++) { 68 case 'c': 69 dst->unit = SCALE_CM; 70 break; 71 case 'i': 72 dst->unit = SCALE_IN; 73 break; 74 case 'f': 75 dst->unit = SCALE_FS; 76 break; 77 case 'M': 78 dst->unit = SCALE_MM; 79 break; 80 case 'm': 81 dst->unit = SCALE_EM; 82 break; 83 case 'n': 84 dst->unit = SCALE_EN; 85 break; 86 case 'P': 87 dst->unit = SCALE_PC; 88 break; 89 case 'p': 90 dst->unit = SCALE_PT; 91 break; 92 case 'u': 93 dst->unit = SCALE_BU; 94 break; 95 case 'v': 96 dst->unit = SCALE_VS; 97 break; 98 default: 99 endptr--; 100 if (SCALE_MAX == def) 101 return NULL; 102 dst->unit = def; 103 break; 104 } 105 return endptr; 106 } 107 108 /* 109 * Calculate the abstract widths and decimal positions of columns in a 110 * table. This routine allocates the columns structures then runs over 111 * all rows and cells in the table. The function pointers in "tbl" are 112 * used for the actual width calculations. 113 */ 114 void 115 tblcalc(struct rofftbl *tbl, const struct tbl_span *sp_first, 116 size_t offset, size_t rmargin) 117 { 118 struct roffsu su; 119 const struct tbl_opts *opts; 120 const struct tbl_span *sp; 121 const struct tbl_dat *dp; 122 struct roffcol *col; 123 struct tbl_colgroup *first_group, **gp, *g; 124 size_t ewidth, min1, min2, wanted, width, xwidth; 125 int done, icol, maxcol, necol, nxcol, quirkcol; 126 127 /* 128 * Allocate the master column specifiers. These will hold the 129 * widths and decimal positions for all cells in the column. It 130 * must be freed and nullified by the caller. 131 */ 132 133 assert(tbl->cols == NULL); 134 tbl->cols = mandoc_calloc((size_t)sp_first->opts->cols, 135 sizeof(struct roffcol)); 136 opts = sp_first->opts; 137 138 maxcol = -1; 139 first_group = NULL; 140 for (sp = sp_first; sp != NULL; sp = sp->next) { 141 if (sp->pos != TBL_SPAN_DATA) 142 continue; 143 144 /* 145 * Account for the data cells in the layout, matching it 146 * to data cells in the data section. 147 */ 148 149 gp = &first_group; 150 for (dp = sp->first; dp != NULL; dp = dp->next) { 151 icol = dp->layout->col; 152 while (maxcol < icol + dp->hspans) 153 tbl->cols[++maxcol].spacing = SIZE_MAX; 154 col = tbl->cols + icol; 155 col->flags |= dp->layout->flags; 156 if (dp->layout->flags & TBL_CELL_WIGN) 157 continue; 158 159 /* Handle explicit width specifications. */ 160 161 if (dp->layout->wstr != NULL && 162 dp->layout->width == 0 && 163 a2roffsu(dp->layout->wstr, &su, SCALE_EN) 164 != NULL) 165 dp->layout->width = 166 (*tbl->sulen)(&su, tbl->arg); 167 if (col->width < dp->layout->width) 168 col->width = dp->layout->width; 169 if (dp->layout->spacing != SIZE_MAX && 170 (col->spacing == SIZE_MAX || 171 col->spacing < dp->layout->spacing)) 172 col->spacing = dp->layout->spacing; 173 174 /* 175 * Calculate an automatic width. 176 * Except for spanning cells, apply it. 177 */ 178 179 width = tblcalc_data(tbl, 180 dp->hspans == 0 ? col : NULL, 181 opts, dp, 182 dp->block == 0 ? 0 : 183 dp->layout->width ? dp->layout->width : 184 rmargin ? (rmargin + sp->opts->cols / 2) 185 / (sp->opts->cols + 1) : 0); 186 if (dp->hspans == 0) 187 continue; 188 189 /* 190 * Build an ordered, singly linked list 191 * of all groups of columns joined by spans, 192 * recording the minimum width for each group. 193 */ 194 195 while (*gp != NULL && ((*gp)->startcol < icol || 196 (*gp)->endcol < icol + dp->hspans)) 197 gp = &(*gp)->next; 198 if (*gp == NULL || (*gp)->startcol > icol || 199 (*gp)->endcol > icol + dp->hspans) { 200 g = mandoc_malloc(sizeof(*g)); 201 g->next = *gp; 202 g->wanted = width; 203 g->startcol = icol; 204 g->endcol = icol + dp->hspans; 205 *gp = g; 206 } else if ((*gp)->wanted < width) 207 (*gp)->wanted = width; 208 } 209 } 210 211 /* 212 * The minimum width of columns explicitly specified 213 * in the layout is 1n. 214 */ 215 216 if (maxcol < sp_first->opts->cols - 1) 217 maxcol = sp_first->opts->cols - 1; 218 for (icol = 0; icol <= maxcol; icol++) { 219 col = tbl->cols + icol; 220 if (col->width < 1) 221 col->width = 1; 222 223 /* 224 * Column spacings are needed for span width 225 * calculations, so set the default values now. 226 */ 227 228 if (col->spacing == SIZE_MAX || icol == maxcol) 229 col->spacing = 3; 230 } 231 232 /* 233 * Replace the minimum widths with the missing widths, 234 * and dismiss groups that are already wide enough. 235 */ 236 237 gp = &first_group; 238 while ((g = *gp) != NULL) { 239 done = 0; 240 for (icol = g->startcol; icol <= g->endcol; icol++) { 241 width = tbl->cols[icol].width; 242 if (icol < g->endcol) 243 width += tbl->cols[icol].spacing; 244 if (g->wanted <= width) { 245 done = 1; 246 break; 247 } else 248 (*gp)->wanted -= width; 249 } 250 if (done) { 251 *gp = g->next; 252 free(g); 253 } else 254 gp = &(*gp)->next; 255 } 256 257 while (first_group != NULL) { 258 259 /* 260 * Find the smallest and second smallest column width 261 * among the columns which may need expamsion. 262 */ 263 264 min1 = min2 = SIZE_MAX; 265 for (icol = 0; icol <= maxcol; icol++) { 266 width = tbl->cols[icol].width; 267 if (min1 > width) { 268 min2 = min1; 269 min1 = width; 270 } else if (min1 < width && min2 > width) 271 min2 = width; 272 } 273 274 /* 275 * Find the minimum wanted width 276 * for any one of the narrowest columns, 277 * and mark the columns wanting that width. 278 */ 279 280 wanted = min2; 281 for (g = first_group; g != NULL; g = g->next) { 282 necol = 0; 283 for (icol = g->startcol; icol <= g->endcol; icol++) 284 if (tbl->cols[icol].width == min1) 285 necol++; 286 if (necol == 0) 287 continue; 288 width = min1 + (g->wanted - 1) / necol + 1; 289 if (width > min2) 290 width = min2; 291 if (wanted > width) 292 wanted = width; 293 } 294 295 /* Record the effect of the widening. */ 296 297 gp = &first_group; 298 while ((g = *gp) != NULL) { 299 done = 0; 300 for (icol = g->startcol; icol <= g->endcol; icol++) { 301 if (tbl->cols[icol].width != min1) 302 continue; 303 if (g->wanted <= wanted - min1) { 304 tbl->cols[icol].width += g->wanted; 305 done = 1; 306 break; 307 } 308 tbl->cols[icol].width = wanted; 309 g->wanted -= wanted - min1; 310 } 311 if (done) { 312 *gp = g->next; 313 free(g); 314 } else 315 gp = &(*gp)->next; 316 } 317 } 318 319 /* 320 * Align numbers with text. 321 * Count columns to equalize and columns to maximize. 322 * Find maximum width of the columns to equalize. 323 * Find total width of the columns *not* to maximize. 324 */ 325 326 necol = nxcol = 0; 327 ewidth = xwidth = 0; 328 for (icol = 0; icol <= maxcol; icol++) { 329 col = tbl->cols + icol; 330 if (col->width > col->nwidth) 331 col->decimal += (col->width - col->nwidth) / 2; 332 if (col->flags & TBL_CELL_EQUAL) { 333 necol++; 334 if (ewidth < col->width) 335 ewidth = col->width; 336 } 337 if (col->flags & TBL_CELL_WMAX) 338 nxcol++; 339 else 340 xwidth += col->width; 341 } 342 343 /* 344 * Equalize columns, if requested for any of them. 345 * Update total width of the columns not to maximize. 346 */ 347 348 if (necol) { 349 for (icol = 0; icol <= maxcol; icol++) { 350 col = tbl->cols + icol; 351 if ( ! (col->flags & TBL_CELL_EQUAL)) 352 continue; 353 if (col->width == ewidth) 354 continue; 355 if (nxcol && rmargin) 356 xwidth += ewidth - col->width; 357 col->width = ewidth; 358 } 359 } 360 361 /* 362 * If there are any columns to maximize, find the total 363 * available width, deducting 3n margins between columns. 364 * Distribute the available width evenly. 365 */ 366 367 if (nxcol && rmargin) { 368 xwidth += 3*maxcol + 369 (opts->opts & (TBL_OPT_BOX | TBL_OPT_DBOX) ? 370 2 : !!opts->lvert + !!opts->rvert); 371 if (rmargin <= offset + xwidth) 372 return; 373 xwidth = rmargin - offset - xwidth; 374 375 /* 376 * Emulate a bug in GNU tbl width calculation that 377 * manifests itself for large numbers of x-columns. 378 * Emulating it for 5 x-columns gives identical 379 * behaviour for up to 6 x-columns. 380 */ 381 382 if (nxcol == 5) { 383 quirkcol = xwidth % nxcol + 2; 384 if (quirkcol != 3 && quirkcol != 4) 385 quirkcol = -1; 386 } else 387 quirkcol = -1; 388 389 necol = 0; 390 ewidth = 0; 391 for (icol = 0; icol <= maxcol; icol++) { 392 col = tbl->cols + icol; 393 if ( ! (col->flags & TBL_CELL_WMAX)) 394 continue; 395 col->width = (double)xwidth * ++necol / nxcol 396 - ewidth + 0.4995; 397 if (necol == quirkcol) 398 col->width--; 399 ewidth += col->width; 400 } 401 } 402 } 403 404 static size_t 405 tblcalc_data(struct rofftbl *tbl, struct roffcol *col, 406 const struct tbl_opts *opts, const struct tbl_dat *dp, size_t mw) 407 { 408 size_t sz; 409 410 /* Branch down into data sub-types. */ 411 412 switch (dp->layout->pos) { 413 case TBL_CELL_HORIZ: 414 case TBL_CELL_DHORIZ: 415 sz = (*tbl->len)(1, tbl->arg); 416 if (col != NULL && col->width < sz) 417 col->width = sz; 418 return sz; 419 case TBL_CELL_LONG: 420 case TBL_CELL_CENTRE: 421 case TBL_CELL_LEFT: 422 case TBL_CELL_RIGHT: 423 return tblcalc_literal(tbl, col, dp, mw); 424 case TBL_CELL_NUMBER: 425 return tblcalc_number(tbl, col, opts, dp); 426 case TBL_CELL_DOWN: 427 return 0; 428 default: 429 abort(); 430 } 431 } 432 433 static size_t 434 tblcalc_literal(struct rofftbl *tbl, struct roffcol *col, 435 const struct tbl_dat *dp, size_t mw) 436 { 437 const char *str; /* Beginning of the first line. */ 438 const char *beg; /* Beginning of the current line. */ 439 char *end; /* End of the current line. */ 440 size_t lsz; /* Length of the current line. */ 441 size_t wsz; /* Length of the current word. */ 442 size_t msz; /* Length of the longest line. */ 443 444 if (dp->string == NULL || *dp->string == '\0') 445 return 0; 446 str = mw ? mandoc_strdup(dp->string) : dp->string; 447 msz = lsz = 0; 448 for (beg = str; beg != NULL && *beg != '\0'; beg = end) { 449 end = mw ? strchr(beg, ' ') : NULL; 450 if (end != NULL) { 451 *end++ = '\0'; 452 while (*end == ' ') 453 end++; 454 } 455 wsz = (*tbl->slen)(beg, tbl->arg); 456 if (mw && lsz && lsz + 1 + wsz <= mw) 457 lsz += 1 + wsz; 458 else 459 lsz = wsz; 460 if (msz < lsz) 461 msz = lsz; 462 } 463 if (mw) 464 free((void *)str); 465 if (col != NULL && col->width < msz) 466 col->width = msz; 467 return msz; 468 } 469 470 static size_t 471 tblcalc_number(struct rofftbl *tbl, struct roffcol *col, 472 const struct tbl_opts *opts, const struct tbl_dat *dp) 473 { 474 const char *cp, *lastdigit, *lastpoint; 475 size_t intsz, totsz; 476 char buf[2]; 477 478 if (dp->string == NULL || *dp->string == '\0') 479 return 0; 480 481 totsz = (*tbl->slen)(dp->string, tbl->arg); 482 if (col == NULL) 483 return totsz; 484 485 /* 486 * Find the last digit and 487 * the last decimal point that is adjacent to a digit. 488 * The alignment indicator "\&" overrides everything. 489 */ 490 491 lastdigit = lastpoint = NULL; 492 for (cp = dp->string; cp[0] != '\0'; cp++) { 493 if (cp[0] == '\\' && cp[1] == '&') { 494 lastdigit = lastpoint = cp; 495 break; 496 } else if (cp[0] == opts->decimal && 497 (isdigit((unsigned char)cp[1]) || 498 (cp > dp->string && isdigit((unsigned char)cp[-1])))) 499 lastpoint = cp; 500 else if (isdigit((unsigned char)cp[0])) 501 lastdigit = cp; 502 } 503 504 /* Not a number, treat as a literal string. */ 505 506 if (lastdigit == NULL) { 507 if (col != NULL && col->width < totsz) 508 col->width = totsz; 509 return totsz; 510 } 511 512 /* Measure the width of the integer part. */ 513 514 if (lastpoint == NULL) 515 lastpoint = lastdigit + 1; 516 intsz = 0; 517 buf[1] = '\0'; 518 for (cp = dp->string; cp < lastpoint; cp++) { 519 buf[0] = cp[0]; 520 intsz += (*tbl->slen)(buf, tbl->arg); 521 } 522 523 /* 524 * If this number has more integer digits than all numbers 525 * seen on earlier lines, shift them all to the right. 526 * If it has fewer, shift this number to the right. 527 */ 528 529 if (intsz > col->decimal) { 530 col->nwidth += intsz - col->decimal; 531 col->decimal = intsz; 532 } else 533 totsz += col->decimal - intsz; 534 535 /* Update the maximum total width seen so far. */ 536 537 if (totsz > col->nwidth) 538 col->nwidth = totsz; 539 if (col->nwidth > col->width) 540 col->width = col->nwidth; 541 return totsz; 542 } 543