src/roff/troff/number.cpp

   1 // -*- C++ -*-
   2 /* Copyright (C) 1989, 1990, 1991, 1992, 2001, 2002, 2004
   3    Free Software Foundation, Inc.
   4      Written by James Clark (jjc@jclark.com)
   5
   6 This file is part of groff.
   7
   8 groff is free software; you can redistribute it and/or modify it under
   9 the terms of the GNU General Public License as published by the Free
  10 Software Foundation; either version 2, or (at your option) any later
  11 version.
  12
  13 groff is distributed in the hope that it will be useful, but WITHOUT ANY
  14 WARRANTY; without even the implied warranty of MERCHANTABILITY or
  15 FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
  16 for more details.
  17
  18 You should have received a copy of the GNU General Public License along
  19 with groff; see the file COPYING.  If not, write to the Free Software
  20 Foundation, 51 Franklin St - Fifth Floor, Boston, MA 02110-1301, USA. */
  21
  22
  23 #include "troff.h"
  24 #include "hvunits.h"
  25 #include "stringclass.h"
  26 #include "mtsm.h"
  27 #include "env.h"
  28 #include "token.h"
  29 #include "div.h"
  30
  31 vunits V0;
  32 hunits H0;
  33
  34 int hresolution = 1;
  35 int vresolution = 1;
  36 int units_per_inch;
  37 int sizescale;
  38
  39 static int parse_expr(units *v, int scale_indicator,
  40                       int parenthesised, int rigid = 0);
  41 static int start_number();
  42
  43 int get_vunits(vunits *res, unsigned char si)
  44 {
  45   if (!start_number())
  46     return 0;
  47   units x;
  48   if (parse_expr(&x, si, 0)) {
  49     *res = vunits(x);
  50     return 1;
  51   }
  52   else
  53     return 0;
  54 }
  55
  56 int get_hunits(hunits *res, unsigned char si)
  57 {
  58   if (!start_number())
  59     return 0;
  60   units x;
  61   if (parse_expr(&x, si, 0)) {
  62     *res = hunits(x);
  63     return 1;
  64   }
  65   else
  66     return 0;
  67 }
  68
  69 // for \B
  70
  71 int get_number_rigidly(units *res, unsigned char si)
  72 {
  73   if (!start_number())
  74     return 0;
  75   units x;
  76   if (parse_expr(&x, si, 0, 1)) {
  77     *res = x;
  78     return 1;
  79   }
  80   else
  81     return 0;
  82 }
  83
  84 int get_number(units *res, unsigned char si)
  85 {
  86   if (!start_number())
  87     return 0;
  88   units x;
  89   if (parse_expr(&x, si, 0)) {
  90     *res = x;
  91     return 1;
  92   }
  93   else
  94     return 0;
  95 }
  96
  97 int get_integer(int *res)
  98 {
  99   if (!start_number())
 100     return 0;
 101   units x;
 102   if (parse_expr(&x, 0, 0)) {
 103     *res = x;
 104     return 1;
 105   }
 106   else
 107     return 0;
 108 }
 109
 110 enum incr_number_result { BAD, ABSOLUTE, INCREMENT, DECREMENT };
 111
 112 static incr_number_result get_incr_number(units *res, unsigned char);
 113
 114 int get_vunits(vunits *res, unsigned char si, vunits prev_value)
 115 {
 116   units v;
 117   switch (get_incr_number(&v, si)) {
 118   case BAD:
 119     return 0;
 120   case ABSOLUTE:
 121     *res = v;
 122     break;
 123   case INCREMENT:
 124     *res = prev_value + v;
 125     break;
 126   case DECREMENT:
 127     *res = prev_value - v;
 128     break;
 129   default:
 130     assert(0);
 131   }
 132   return 1;
 133 }
 134
 135 int get_hunits(hunits *res, unsigned char si, hunits prev_value)
 136 {
 137   units v;
 138   switch (get_incr_number(&v, si)) {
 139   case BAD:
 140     return 0;
 141   case ABSOLUTE:
 142     *res = v;
 143     break;
 144   case INCREMENT:
 145     *res = prev_value + v;
 146     break;
 147   case DECREMENT:
 148     *res = prev_value - v;
 149     break;
 150   default:
 151     assert(0);
 152   }
 153   return 1;
 154 }
 155
 156 int get_number(units *res, unsigned char si, units prev_value)
 157 {
 158   units v;
 159   switch (get_incr_number(&v, si)) {
 160   case BAD:
 161     return 0;
 162   case ABSOLUTE:
 163     *res = v;
 164     break;
 165   case INCREMENT:
 166     *res = prev_value + v;
 167     break;
 168   case DECREMENT:
 169     *res = prev_value - v;
 170     break;
 171   default:
 172     assert(0);
 173   }
 174   return 1;
 175 }
 176
 177 int get_integer(int *res, int prev_value)
 178 {
 179   units v;
 180   switch (get_incr_number(&v, 0)) {
 181   case BAD:
 182     return 0;
 183   case ABSOLUTE:
 184     *res = v;
 185     break;
 186   case INCREMENT:
 187     *res = prev_value + int(v);
 188     break;
 189   case DECREMENT:
 190     *res = prev_value - int(v);
 191     break;
 192   default:
 193     assert(0);
 194   }
 195   return 1;
 196 }
 197
 198
 199 static incr_number_result get_incr_number(units *res, unsigned char si)
 200 {
 201   if (!start_number())
 202     return BAD;
 203   incr_number_result result = ABSOLUTE;
 204   if (tok.ch() == '+') {
 205     tok.next();
 206     result = INCREMENT;
 207   }
 208   else if (tok.ch() == '-') {
 209     tok.next();
 210     result = DECREMENT;
 211   }
 212   if (parse_expr(res, si, 0))
 213     return result;
 214   else
 215     return BAD;
 216 }
 217
 218 static int start_number()
 219 {
 220   while (tok.space())
 221     tok.next();
 222   if (tok.newline()) {
 223     warning(WARN_MISSING, "missing number");
 224     return 0;
 225   }
 226   if (tok.tab()) {
 227     warning(WARN_TAB, "tab character where number expected");
 228     return 0;
 229   }
 230   if (tok.right_brace()) {
 231     warning(WARN_RIGHT_BRACE, "`\\}' where number expected");
 232     return 0;
 233   }
 234   return 1;
 235 }
 236
 237 enum { OP_LEQ = 'L', OP_GEQ = 'G', OP_MAX = 'X', OP_MIN = 'N' };
 238
 239 #define SCALE_INDICATOR_CHARS "icfPmnpuvMsz"
 240
 241 static int parse_term(units *v, int scale_indicator,
 242                       int parenthesised, int rigid);
 243
 244 static int parse_expr(units *v, int scale_indicator,
 245                       int parenthesised, int rigid)
 246 {
 247   int result = parse_term(v, scale_indicator, parenthesised, rigid);
 248   while (result) {
 249     if (parenthesised)
 250       tok.skip();
 251     int op = tok.ch();
 252     switch (op) {
 253     case '+':
 254     case '-':
 255     case '/':
 256     case '*':
 257     case '%':
 258     case ':':
 259     case '&':
 260       tok.next();
 261       break;
 262     case '>':
 263       tok.next();
 264       if (tok.ch() == '=') {
 265         tok.next();
 266         op = OP_GEQ;
 267       }
 268       else if (tok.ch() == '?') {
 269         tok.next();
 270         op = OP_MAX;
 271       }
 272       break;
 273     case '<':
 274       tok.next();
 275       if (tok.ch() == '=') {
 276         tok.next();
 277         op = OP_LEQ;
 278       }
 279       else if (tok.ch() == '?') {
 280         tok.next();
 281         op = OP_MIN;
 282       }
 283       break;
 284     case '=':
 285       tok.next();
 286       if (tok.ch() == '=')
 287         tok.next();
 288       break;
 289     default:
 290       return result;
 291     }
 292     units v2;
 293     if (!parse_term(&v2, scale_indicator, parenthesised, rigid))
 294       return 0;
 295     int overflow = 0;
 296     switch (op) {
 297     case '<':
 298       *v = *v < v2;
 299       break;
 300     case '>':
 301       *v = *v > v2;
 302       break;
 303     case OP_LEQ:
 304       *v = *v <= v2;
 305       break;
 306     case OP_GEQ:
 307       *v = *v >= v2;
 308       break;
 309     case OP_MIN:
 310       if (*v > v2)
 311         *v = v2;
 312       break;
 313     case OP_MAX:
 314       if (*v < v2)
 315         *v = v2;
 316       break;
 317     case '=':
 318       *v = *v == v2;
 319       break;
 320     case '&':
 321       *v = *v > 0 && v2 > 0;
 322       break;
 323     case ':':
 324       *v = *v > 0 || v2 > 0;
 325       break;
 326     case '+':
 327       if (v2 < 0) {
 328         if (*v < INT_MIN - v2)
 329           overflow = 1;
 330       }
 331       else if (v2 > 0) {
 332         if (*v > INT_MAX - v2)
 333           overflow = 1;
 334       }
 335       if (overflow) {
 336         error("addition overflow");
 337         return 0;
 338       }
 339       *v += v2;
 340       break;
 341     case '-':
 342       if (v2 < 0) {
 343         if (*v > INT_MAX + v2)
 344           overflow = 1;
 345       }
 346       else if (v2 > 0) {
 347         if (*v < INT_MIN + v2)
 348           overflow = 1;
 349       }
 350       if (overflow) {
 351         error("subtraction overflow");
 352         return 0;
 353       }
 354       *v -= v2;
 355       break;
 356     case '*':
 357       if (v2 < 0) {
 358         if (*v > 0) {
 359           if (*v > -(unsigned)INT_MIN / -(unsigned)v2)
 360             overflow = 1;
 361         }
 362         else if (-(unsigned)*v > INT_MAX / -(unsigned)v2)
 363           overflow = 1;
 364       }
 365       else if (v2 > 0) {
 366         if (*v > 0) {
 367           if (*v > INT_MAX / v2)
 368             overflow = 1;
 369         }
 370         else if (-(unsigned)*v > -(unsigned)INT_MIN / v2)
 371           overflow = 1;
 372       }
 373       if (overflow) {
 374         error("multiplication overflow");
 375         return 0;
 376       }
 377       *v *= v2;
 378       break;
 379     case '/':
 380       if (v2 == 0) {
 381         error("division by zero");
 382         return 0;
 383       }
 384       *v /= v2;
 385       break;
 386     case '%':
 387       if (v2 == 0) {
 388         error("modulus by zero");
 389         return 0;
 390       }
 391       *v %= v2;
 392       break;
 393     default:
 394       assert(0);
 395     }
 396   }
 397   return result;
 398 }
 399
 400 static int parse_term(units *v, int scale_indicator,
 401                       int parenthesised, int rigid)
 402 {
 403   int negative = 0;
 404   for (;;)
 405     if (parenthesised && tok.space())
 406       tok.next();
 407     else if (tok.ch() == '+')
 408       tok.next();
 409     else if (tok.ch() == '-') {
 410       tok.next();
 411       negative = !negative;
 412     }
 413     else
 414       break;
 415   unsigned char c = tok.ch();
 416   switch (c) {
 417   case '|':
 418     // | is not restricted to the outermost level
 419     // tbl uses this
 420     tok.next();
 421     if (!parse_term(v, scale_indicator, parenthesised, rigid))
 422       return 0;
 423     int tem;
 424     tem = (scale_indicator == 'v'
 425            ? curdiv->get_vertical_position().to_units()
 426            : curenv->get_input_line_position().to_units());
 427     if (tem >= 0) {
 428       if (*v < INT_MIN + tem) {
 429         error("numeric overflow");
 430         return 0;
 431       }
 432     }
 433     else {
 434       if (*v > INT_MAX + tem) {
 435         error("numeric overflow");
 436         return 0;
 437       }
 438     }
 439     *v -= tem;
 440     if (negative) {
 441       if (*v == INT_MIN) {
 442         error("numeric overflow");
 443         return 0;
 444       }
 445       *v = -*v;
 446     }
 447     return 1;
 448   case '(':
 449     tok.next();
 450     c = tok.ch();
 451     if (c == ')') {
 452       if (rigid)
 453         return 0;
 454       warning(WARN_SYNTAX, "empty parentheses");
 455       tok.next();
 456       *v = 0;
 457       return 1;
 458     }
 459     else if (c != 0 && strchr(SCALE_INDICATOR_CHARS, c) != 0) {
 460       tok.next();
 461       if (tok.ch() == ';') {
 462         tok.next();
 463         scale_indicator = c;
 464       }
 465       else {
 466         error("expected `;' after scale-indicator (got %1)",
 467               tok.description());
 468         return 0;
 469       }
 470     }
 471     else if (c == ';') {
 472       scale_indicator = 0;
 473       tok.next();
 474     }
 475     if (!parse_expr(v, scale_indicator, 1, rigid))
 476       return 0;
 477     tok.skip();
 478     if (tok.ch() != ')') {
 479       if (rigid)
 480         return 0;
 481       warning(WARN_SYNTAX, "missing `)' (got %1)", tok.description());
 482     }
 483     else
 484       tok.next();
 485     if (negative) {
 486       if (*v == INT_MIN) {
 487         error("numeric overflow");
 488         return 0;
 489       }
 490       *v = -*v;
 491     }
 492     return 1;
 493   case '.':
 494     *v = 0;
 495     break;
 496   case '0':
 497   case '1':
 498   case '2':
 499   case '3':
 500   case '4':
 501   case '5':
 502   case '6':
 503   case '7':
 504   case '8':
 505   case '9':
 506     *v = 0;
 507     do {
 508       if (*v > INT_MAX/10) {
 509         error("numeric overflow");
 510         return 0;
 511       }
 512       *v *= 10;
 513       if (*v > INT_MAX - (int(c) - '0')) {
 514         error("numeric overflow");
 515         return 0;
 516       }
 517       *v += c - '0';
 518       tok.next();
 519       c = tok.ch();
 520     } while (csdigit(c));
 521     break;
 522   case '/':
 523   case '*':
 524   case '%':
 525   case ':':
 526   case '&':
 527   case '>':
 528   case '<':
 529   case '=':
 530     warning(WARN_SYNTAX, "empty left operand");
 531     *v = 0;
 532     return rigid ? 0 : 1;
 533   default:
 534     warning(WARN_NUMBER, "numeric expression expected (got %1)",
 535             tok.description());
 536     return 0;
 537   }
 538   int divisor = 1;
 539   if (tok.ch() == '.') {
 540     tok.next();
 541     for (;;) {
 542       c = tok.ch();
 543       if (!csdigit(c))
 544         break;
 545       // we may multiply the divisor by 254 later on
 546       if (divisor <= INT_MAX/2540 && *v <= (INT_MAX - 9)/10) {
 547         *v *= 10;
 548         *v += c - '0';
 549         divisor *= 10;
 550       }
 551       tok.next();
 552     }
 553   }
 554   int si = scale_indicator;
 555   int do_next = 0;
 556   if ((c = tok.ch()) != 0 && strchr(SCALE_INDICATOR_CHARS, c) != 0) {
 557     switch (scale_indicator) {
 558     case 'z':
 559       if (c != 'u' && c != 'z') {
 560         warning(WARN_SCALE,
 561                 "only `z' and `u' scale indicators valid in this context");
 562         break;
 563       }
 564       si = c;
 565       break;
 566     case 0:
 567       warning(WARN_SCALE, "scale indicator invalid in this context");
 568       break;
 569     case 'u':
 570       si = c;
 571       break;
 572     default:
 573       if (c == 'z') {
 574         warning(WARN_SCALE, "`z' scale indicator invalid in this context");
 575         break;
 576       }
 577       si = c;
 578       break;
 579     }
 580     // Don't do tok.next() here because the next token might be \s, which
 581     // would affect the interpretation of m.
 582     do_next = 1;
 583   }
 584   switch (si) {
 585   case 'i':
 586     *v = scale(*v, units_per_inch, divisor);
 587     break;
 588   case 'c':
 589     *v = scale(*v, units_per_inch*100, divisor*254);
 590     break;
 591   case 0:
 592   case 'u':
 593     if (divisor != 1)
 594       *v /= divisor;
 595     break;
 596   case 'f':
 597     *v = scale(*v, 65536, divisor);
 598     break;
 599   case 'p':
 600     *v = scale(*v, units_per_inch, divisor*72);
 601     break;
 602   case 'P':
 603     *v = scale(*v, units_per_inch, divisor*6);
 604     break;
 605   case 'm':
 606     {
 607       // Convert to hunits so that with -Tascii `m' behaves as in nroff.
 608       hunits em = curenv->get_size();
 609       *v = scale(*v, em.is_zero() ? hresolution : em.to_units(), divisor);
 610     }
 611     break;
 612   case 'M':
 613     {
 614       hunits em = curenv->get_size();
 615       *v = scale(*v, em.is_zero() ? hresolution : em.to_units(), divisor*100);
 616     }
 617     break;
 618   case 'n':
 619     {
 620       // Convert to hunits so that with -Tascii `n' behaves as in nroff.
 621       hunits en = curenv->get_size()/2;
 622       *v = scale(*v, en.is_zero() ? hresolution : en.to_units(), divisor);
 623     }
 624     break;
 625   case 'v':
 626     *v = scale(*v, curenv->get_vertical_spacing().to_units(), divisor);
 627     break;
 628   case 's':
 629     while (divisor > INT_MAX/(sizescale*72)) {
 630       divisor /= 10;
 631       *v /= 10;
 632     }
 633     *v = scale(*v, units_per_inch, divisor*sizescale*72);
 634     break;
 635   case 'z':
 636     *v = scale(*v, sizescale, divisor);
 637     break;
 638   default:
 639     assert(0);
 640   }
 641   if (do_next)
 642     tok.next();
 643   if (negative) {
 644     if (*v == INT_MIN) {
 645       error("numeric overflow");
 646       return 0;
 647     }
 648     *v = -*v;
 649   }
 650   return 1;
 651 }
 652
 653 units scale(units n, units x, units y)
 654 {
 655   assert(x >= 0 && y > 0);
 656   if (x == 0)
 657     return 0;
 658   if (n >= 0) {
 659     if (n <= INT_MAX/x)
 660       return (n*x)/y;
 661   }
 662   else {
 663     if (-(unsigned)n <= -(unsigned)INT_MIN/x)
 664       return (n*x)/y;
 665   }
 666   double res = n*double(x)/double(y);
 667   if (res > INT_MAX) {
 668     error("numeric overflow");
 669     return INT_MAX;
 670   }
 671   else if (res < INT_MIN) {
 672     error("numeric overflow");
 673     return INT_MIN;
 674   }
 675   return int(res);
 676 }
 677
 678 vunits::vunits(units x)
 679 {
 680   // don't depend on the rounding direction for division of negative integers
 681   if (vresolution == 1)
 682     n = x;
 683   else
 684     n = (x < 0
 685          ? -((-x + vresolution/2 - 1)/vresolution)
 686          : (x + vresolution/2 - 1)/vresolution);
 687 }
 688
 689 hunits::hunits(units x)
 690 {
 691   // don't depend on the rounding direction for division of negative integers
 692   if (hresolution == 1)
 693     n = x;
 694   else
 695     n = (x < 0
 696          ? -((-x + hresolution/2 - 1)/hresolution)
 697          : (x + hresolution/2 - 1)/hresolution);
 698 }