1 /* Analyze file differences for GNU DIFF. 2 Copyright (C) 1988, 1989, 1992, 1993, 1997 Free Software Foundation, Inc. 3 4 This file is part of GNU DIFF. 5 6 GNU DIFF is free software; you can redistribute it and/or modify 7 it under the terms of the GNU General Public License as published by 8 the Free Software Foundation; either version 2, or (at your option) 9 any later version. 10 11 GNU DIFF is distributed in the hope that it will be useful, 12 but WITHOUT ANY WARRANTY; without even the implied warranty of 13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 14 GNU General Public License for more details. 15 16 You should have received a copy of the GNU General Public License 17 along with GNU DIFF; see the file COPYING. If not, write to 18 the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. */ 19 20 /* The basic algorithm is described in: 21 "An O(ND) Difference Algorithm and its Variations", Eugene Myers, 22 Algorithmica Vol. 1 No. 2, 1986, pp. 251-266; 23 see especially section 4.2, which describes the variation used below. 24 Unless the --minimal option is specified, this code uses the TOO_EXPENSIVE 25 heuristic, by Paul Eggert, to limit the cost to O(N**1.5 log N) 26 at the price of producing suboptimal output for large inputs with 27 many differences. 28 29 The basic algorithm was independently discovered as described in: 30 "Algorithms for Approximate String Matching", E. Ukkonen, 31 Information and Control Vol. 64, 1985, pp. 100-118. */ 32 33 #include "diff.h" 34 #include "cmpbuf.h" 35 36 extern int no_discards; 37 38 static int *xvec, *yvec; /* Vectors being compared. */ 39 static int *fdiag; /* Vector, indexed by diagonal, containing 40 1 + the X coordinate of the point furthest 41 along the given diagonal in the forward 42 search of the edit matrix. */ 43 static int *bdiag; /* Vector, indexed by diagonal, containing 44 the X coordinate of the point furthest 45 along the given diagonal in the backward 46 search of the edit matrix. */ 47 static int too_expensive; /* Edit scripts longer than this are too 48 expensive to compute. */ 49 50 #define SNAKE_LIMIT 20 /* Snakes bigger than this are considered `big'. */ 51 52 struct partition 53 { 54 int xmid, ymid; /* Midpoints of this partition. */ 55 int lo_minimal; /* Nonzero if low half will be analyzed minimally. */ 56 int hi_minimal; /* Likewise for high half. */ 57 }; 58 59 static int diag PARAMS((int, int, int, int, int, struct partition *)); 60 static struct change *add_change PARAMS((int, int, int, int, struct change *)); 61 static struct change *build_reverse_script PARAMS((struct file_data const[])); 62 static struct change *build_script PARAMS((struct file_data const[])); 63 static void briefly_report PARAMS((int, struct file_data const[])); 64 static void compareseq PARAMS((int, int, int, int, int)); 65 static void discard_confusing_lines PARAMS((struct file_data[])); 66 static void shift_boundaries PARAMS((struct file_data[])); 67 68 /* Find the midpoint of the shortest edit script for a specified 69 portion of the two files. 70 71 Scan from the beginnings of the files, and simultaneously from the ends, 72 doing a breadth-first search through the space of edit-sequence. 73 When the two searches meet, we have found the midpoint of the shortest 74 edit sequence. 75 76 If MINIMAL is nonzero, find the minimal edit script regardless 77 of expense. Otherwise, if the search is too expensive, use 78 heuristics to stop the search and report a suboptimal answer. 79 80 Set PART->(XMID,YMID) to the midpoint (XMID,YMID). The diagonal number 81 XMID - YMID equals the number of inserted lines minus the number 82 of deleted lines (counting only lines before the midpoint). 83 Return the approximate edit cost; this is the total number of 84 lines inserted or deleted (counting only lines before the midpoint), 85 unless a heuristic is used to terminate the search prematurely. 86 87 Set PART->LEFT_MINIMAL to nonzero iff the minimal edit script for the 88 left half of the partition is known; similarly for PART->RIGHT_MINIMAL. 89 90 This function assumes that the first lines of the specified portions 91 of the two files do not match, and likewise that the last lines do not 92 match. The caller must trim matching lines from the beginning and end 93 of the portions it is going to specify. 94 95 If we return the "wrong" partitions, 96 the worst this can do is cause suboptimal diff output. 97 It cannot cause incorrect diff output. */ 98 99 static int 100 diag (xoff, xlim, yoff, ylim, minimal, part) 101 int xoff, xlim, yoff, ylim, minimal; 102 struct partition *part; 103 { 104 int *const fd = fdiag; /* Give the compiler a chance. */ 105 int *const bd = bdiag; /* Additional help for the compiler. */ 106 int const *const xv = xvec; /* Still more help for the compiler. */ 107 int const *const yv = yvec; /* And more and more . . . */ 108 int const dmin = xoff - ylim; /* Minimum valid diagonal. */ 109 int const dmax = xlim - yoff; /* Maximum valid diagonal. */ 110 int const fmid = xoff - yoff; /* Center diagonal of top-down search. */ 111 int const bmid = xlim - ylim; /* Center diagonal of bottom-up search. */ 112 int fmin = fmid, fmax = fmid; /* Limits of top-down search. */ 113 int bmin = bmid, bmax = bmid; /* Limits of bottom-up search. */ 114 int c; /* Cost. */ 115 int odd = (fmid - bmid) & 1; /* True if southeast corner is on an odd 116 diagonal with respect to the northwest. */ 117 118 fd[fmid] = xoff; 119 bd[bmid] = xlim; 120 121 for (c = 1;; ++c) 122 { 123 int d; /* Active diagonal. */ 124 int big_snake = 0; 125 126 /* Extend the top-down search by an edit step in each diagonal. */ 127 fmin > dmin ? fd[--fmin - 1] = -1 : ++fmin; 128 fmax < dmax ? fd[++fmax + 1] = -1 : --fmax; 129 for (d = fmax; d >= fmin; d -= 2) 130 { 131 int x, y, oldx, tlo = fd[d - 1], thi = fd[d + 1]; 132 133 if (tlo >= thi) 134 x = tlo + 1; 135 else 136 x = thi; 137 oldx = x; 138 y = x - d; 139 while (x < xlim && y < ylim && xv[x] == yv[y]) 140 ++x, ++y; 141 if (x - oldx > SNAKE_LIMIT) 142 big_snake = 1; 143 fd[d] = x; 144 if (odd && bmin <= d && d <= bmax && bd[d] <= x) 145 { 146 part->xmid = x; 147 part->ymid = y; 148 part->lo_minimal = part->hi_minimal = 1; 149 return 2 * c - 1; 150 } 151 } 152 153 /* Similarly extend the bottom-up search. */ 154 bmin > dmin ? bd[--bmin - 1] = INT_MAX : ++bmin; 155 bmax < dmax ? bd[++bmax + 1] = INT_MAX : --bmax; 156 for (d = bmax; d >= bmin; d -= 2) 157 { 158 int x, y, oldx, tlo = bd[d - 1], thi = bd[d + 1]; 159 160 if (tlo < thi) 161 x = tlo; 162 else 163 x = thi - 1; 164 oldx = x; 165 y = x - d; 166 while (x > xoff && y > yoff && xv[x - 1] == yv[y - 1]) 167 --x, --y; 168 if (oldx - x > SNAKE_LIMIT) 169 big_snake = 1; 170 bd[d] = x; 171 if (!odd && fmin <= d && d <= fmax && x <= fd[d]) 172 { 173 part->xmid = x; 174 part->ymid = y; 175 part->lo_minimal = part->hi_minimal = 1; 176 return 2 * c; 177 } 178 } 179 180 if (minimal) 181 continue; 182 183 /* Heuristic: check occasionally for a diagonal that has made 184 lots of progress compared with the edit distance. 185 If we have any such, find the one that has made the most 186 progress and return it as if it had succeeded. 187 188 With this heuristic, for files with a constant small density 189 of changes, the algorithm is linear in the file size. */ 190 191 if (c > 200 && big_snake && heuristic) 192 { 193 int best; 194 195 best = 0; 196 for (d = fmax; d >= fmin; d -= 2) 197 { 198 int dd = d - fmid; 199 int x = fd[d]; 200 int y = x - d; 201 int v = (x - xoff) * 2 - dd; 202 if (v > 12 * (c + (dd < 0 ? -dd : dd))) 203 { 204 if (v > best 205 && xoff + SNAKE_LIMIT <= x && x < xlim 206 && yoff + SNAKE_LIMIT <= y && y < ylim) 207 { 208 /* We have a good enough best diagonal; 209 now insist that it end with a significant snake. */ 210 int k; 211 212 for (k = 1; xv[x - k] == yv[y - k]; k++) 213 if (k == SNAKE_LIMIT) 214 { 215 best = v; 216 part->xmid = x; 217 part->ymid = y; 218 break; 219 } 220 } 221 } 222 } 223 if (best > 0) 224 { 225 part->lo_minimal = 1; 226 part->hi_minimal = 0; 227 return 2 * c - 1; 228 } 229 230 best = 0; 231 for (d = bmax; d >= bmin; d -= 2) 232 { 233 int dd = d - bmid; 234 int x = bd[d]; 235 int y = x - d; 236 int v = (xlim - x) * 2 + dd; 237 if (v > 12 * (c + (dd < 0 ? -dd : dd))) 238 { 239 if (v > best 240 && xoff < x && x <= xlim - SNAKE_LIMIT 241 && yoff < y && y <= ylim - SNAKE_LIMIT) 242 { 243 /* We have a good enough best diagonal; 244 now insist that it end with a significant snake. */ 245 int k; 246 247 for (k = 0; xv[x + k] == yv[y + k]; k++) 248 if (k == SNAKE_LIMIT - 1) 249 { 250 best = v; 251 part->xmid = x; 252 part->ymid = y; 253 break; 254 } 255 } 256 } 257 } 258 if (best > 0) 259 { 260 part->lo_minimal = 0; 261 part->hi_minimal = 1; 262 return 2 * c - 1; 263 } 264 } 265 266 /* Heuristic: if we've gone well beyond the call of duty, 267 give up and report halfway between our best results so far. */ 268 if (c >= too_expensive) 269 { 270 int fxybest, fxbest; 271 int bxybest, bxbest; 272 273 fxbest = bxbest = 0; /* Pacify `gcc -Wall'. */ 274 275 /* Find forward diagonal that maximizes X + Y. */ 276 fxybest = -1; 277 for (d = fmax; d >= fmin; d -= 2) 278 { 279 int x = min (fd[d], xlim); 280 int y = x - d; 281 if (ylim < y) 282 x = ylim + d, y = ylim; 283 if (fxybest < x + y) 284 { 285 fxybest = x + y; 286 fxbest = x; 287 } 288 } 289 290 /* Find backward diagonal that minimizes X + Y. */ 291 bxybest = INT_MAX; 292 for (d = bmax; d >= bmin; d -= 2) 293 { 294 int x = max (xoff, bd[d]); 295 int y = x - d; 296 if (y < yoff) 297 x = yoff + d, y = yoff; 298 if (x + y < bxybest) 299 { 300 bxybest = x + y; 301 bxbest = x; 302 } 303 } 304 305 /* Use the better of the two diagonals. */ 306 if ((xlim + ylim) - bxybest < fxybest - (xoff + yoff)) 307 { 308 part->xmid = fxbest; 309 part->ymid = fxybest - fxbest; 310 part->lo_minimal = 1; 311 part->hi_minimal = 0; 312 } 313 else 314 { 315 part->xmid = bxbest; 316 part->ymid = bxybest - bxbest; 317 part->lo_minimal = 0; 318 part->hi_minimal = 1; 319 } 320 return 2 * c - 1; 321 } 322 } 323 } 324 325 /* Compare in detail contiguous subsequences of the two files 326 which are known, as a whole, to match each other. 327 328 The results are recorded in the vectors files[N].changed_flag, by 329 storing a 1 in the element for each line that is an insertion or deletion. 330 331 The subsequence of file 0 is [XOFF, XLIM) and likewise for file 1. 332 333 Note that XLIM, YLIM are exclusive bounds. 334 All line numbers are origin-0 and discarded lines are not counted. 335 336 If MINIMAL is nonzero, find a minimal difference no matter how 337 expensive it is. */ 338 339 static void 340 compareseq (xoff, xlim, yoff, ylim, minimal) 341 int xoff, xlim, yoff, ylim, minimal; 342 { 343 int * const xv = xvec; /* Help the compiler. */ 344 int * const yv = yvec; 345 346 /* Slide down the bottom initial diagonal. */ 347 while (xoff < xlim && yoff < ylim && xv[xoff] == yv[yoff]) 348 ++xoff, ++yoff; 349 /* Slide up the top initial diagonal. */ 350 while (xlim > xoff && ylim > yoff && xv[xlim - 1] == yv[ylim - 1]) 351 --xlim, --ylim; 352 353 /* Handle simple cases. */ 354 if (xoff == xlim) 355 while (yoff < ylim) 356 files[1].changed_flag[files[1].realindexes[yoff++]] = 1; 357 else if (yoff == ylim) 358 while (xoff < xlim) 359 files[0].changed_flag[files[0].realindexes[xoff++]] = 1; 360 else 361 { 362 int c; 363 struct partition part; 364 365 /* Find a point of correspondence in the middle of the files. */ 366 367 c = diag (xoff, xlim, yoff, ylim, minimal, &part); 368 369 if (c == 1) 370 { 371 /* This should be impossible, because it implies that 372 one of the two subsequences is empty, 373 and that case was handled above without calling `diag'. 374 Let's verify that this is true. */ 375 abort (); 376 #if 0 377 /* The two subsequences differ by a single insert or delete; 378 record it and we are done. */ 379 if (part.xmid - part.ymid < xoff - yoff) 380 files[1].changed_flag[files[1].realindexes[part.ymid - 1]] = 1; 381 else 382 files[0].changed_flag[files[0].realindexes[part.xmid]] = 1; 383 #endif 384 } 385 else 386 { 387 /* Use the partitions to split this problem into subproblems. */ 388 compareseq (xoff, part.xmid, yoff, part.ymid, part.lo_minimal); 389 compareseq (part.xmid, xlim, part.ymid, ylim, part.hi_minimal); 390 } 391 } 392 } 393 394 /* Discard lines from one file that have no matches in the other file. 395 396 A line which is discarded will not be considered by the actual 397 comparison algorithm; it will be as if that line were not in the file. 398 The file's `realindexes' table maps virtual line numbers 399 (which don't count the discarded lines) into real line numbers; 400 this is how the actual comparison algorithm produces results 401 that are comprehensible when the discarded lines are counted. 402 403 When we discard a line, we also mark it as a deletion or insertion 404 so that it will be printed in the output. */ 405 406 static void 407 discard_confusing_lines (filevec) 408 struct file_data filevec[]; 409 { 410 unsigned int f, i; 411 char *discarded[2]; 412 int *equiv_count[2]; 413 int *p; 414 415 /* Allocate our results. */ 416 p = (int *) xmalloc ((filevec[0].buffered_lines + filevec[1].buffered_lines) 417 * (2 * sizeof (int))); 418 for (f = 0; f < 2; f++) 419 { 420 filevec[f].undiscarded = p; p += filevec[f].buffered_lines; 421 filevec[f].realindexes = p; p += filevec[f].buffered_lines; 422 } 423 424 /* Set up equiv_count[F][I] as the number of lines in file F 425 that fall in equivalence class I. */ 426 427 p = (int *) xmalloc (filevec[0].equiv_max * (2 * sizeof (int))); 428 equiv_count[0] = p; 429 equiv_count[1] = p + filevec[0].equiv_max; 430 bzero (p, filevec[0].equiv_max * (2 * sizeof (int))); 431 432 for (i = 0; i < filevec[0].buffered_lines; ++i) 433 ++equiv_count[0][filevec[0].equivs[i]]; 434 for (i = 0; i < filevec[1].buffered_lines; ++i) 435 ++equiv_count[1][filevec[1].equivs[i]]; 436 437 /* Set up tables of which lines are going to be discarded. */ 438 439 discarded[0] = xmalloc (sizeof (char) 440 * (filevec[0].buffered_lines 441 + filevec[1].buffered_lines)); 442 discarded[1] = discarded[0] + filevec[0].buffered_lines; 443 bzero (discarded[0], sizeof (char) * (filevec[0].buffered_lines 444 + filevec[1].buffered_lines)); 445 446 /* Mark to be discarded each line that matches no line of the other file. 447 If a line matches many lines, mark it as provisionally discardable. */ 448 449 for (f = 0; f < 2; f++) 450 { 451 unsigned int end = filevec[f].buffered_lines; 452 char *discards = discarded[f]; 453 int *counts = equiv_count[1 - f]; 454 int *equivs = filevec[f].equivs; 455 unsigned int many = 5; 456 unsigned int tem = end / 64; 457 458 /* Multiply MANY by approximate square root of number of lines. 459 That is the threshold for provisionally discardable lines. */ 460 while ((tem = tem >> 2) > 0) 461 many *= 2; 462 463 for (i = 0; i < end; i++) 464 { 465 int nmatch; 466 if (equivs[i] == 0) 467 continue; 468 nmatch = counts[equivs[i]]; 469 if (nmatch == 0) 470 discards[i] = 1; 471 else if (nmatch > many) 472 discards[i] = 2; 473 } 474 } 475 476 /* Don't really discard the provisional lines except when they occur 477 in a run of discardables, with nonprovisionals at the beginning 478 and end. */ 479 480 for (f = 0; f < 2; f++) 481 { 482 unsigned int end = filevec[f].buffered_lines; 483 register char *discards = discarded[f]; 484 485 for (i = 0; i < end; i++) 486 { 487 /* Cancel provisional discards not in middle of run of discards. */ 488 if (discards[i] == 2) 489 discards[i] = 0; 490 else if (discards[i] != 0) 491 { 492 /* We have found a nonprovisional discard. */ 493 register int j; 494 unsigned int length; 495 unsigned int provisional = 0; 496 497 /* Find end of this run of discardable lines. 498 Count how many are provisionally discardable. */ 499 for (j = i; j < end; j++) 500 { 501 if (discards[j] == 0) 502 break; 503 if (discards[j] == 2) 504 ++provisional; 505 } 506 507 /* Cancel provisional discards at end, and shrink the run. */ 508 while (j > i && discards[j - 1] == 2) 509 discards[--j] = 0, --provisional; 510 511 /* Now we have the length of a run of discardable lines 512 whose first and last are not provisional. */ 513 length = j - i; 514 515 /* If 1/4 of the lines in the run are provisional, 516 cancel discarding of all provisional lines in the run. */ 517 if (provisional * 4 > length) 518 { 519 while (j > i) 520 if (discards[--j] == 2) 521 discards[j] = 0; 522 } 523 else 524 { 525 register unsigned int consec; 526 unsigned int minimum = 1; 527 unsigned int tem = length / 4; 528 529 /* MINIMUM is approximate square root of LENGTH/4. 530 A subrun of two or more provisionals can stand 531 when LENGTH is at least 16. 532 A subrun of 4 or more can stand when LENGTH >= 64. */ 533 while ((tem = tem >> 2) > 0) 534 minimum *= 2; 535 minimum++; 536 537 /* Cancel any subrun of MINIMUM or more provisionals 538 within the larger run. */ 539 for (j = 0, consec = 0; j < length; j++) 540 if (discards[i + j] != 2) 541 consec = 0; 542 else if (minimum == ++consec) 543 /* Back up to start of subrun, to cancel it all. */ 544 j -= consec; 545 else if (minimum < consec) 546 discards[i + j] = 0; 547 548 /* Scan from beginning of run 549 until we find 3 or more nonprovisionals in a row 550 or until the first nonprovisional at least 8 lines in. 551 Until that point, cancel any provisionals. */ 552 for (j = 0, consec = 0; j < length; j++) 553 { 554 if (j >= 8 && discards[i + j] == 1) 555 break; 556 if (discards[i + j] == 2) 557 consec = 0, discards[i + j] = 0; 558 else if (discards[i + j] == 0) 559 consec = 0; 560 else 561 consec++; 562 if (consec == 3) 563 break; 564 } 565 566 /* I advances to the last line of the run. */ 567 i += length - 1; 568 569 /* Same thing, from end. */ 570 for (j = 0, consec = 0; j < length; j++) 571 { 572 if (j >= 8 && discards[i - j] == 1) 573 break; 574 if (discards[i - j] == 2) 575 consec = 0, discards[i - j] = 0; 576 else if (discards[i - j] == 0) 577 consec = 0; 578 else 579 consec++; 580 if (consec == 3) 581 break; 582 } 583 } 584 } 585 } 586 } 587 588 /* Actually discard the lines. */ 589 for (f = 0; f < 2; f++) 590 { 591 char *discards = discarded[f]; 592 unsigned int end = filevec[f].buffered_lines; 593 unsigned int j = 0; 594 for (i = 0; i < end; ++i) 595 if (no_discards || discards[i] == 0) 596 { 597 filevec[f].undiscarded[j] = filevec[f].equivs[i]; 598 filevec[f].realindexes[j++] = i; 599 } 600 else 601 filevec[f].changed_flag[i] = 1; 602 filevec[f].nondiscarded_lines = j; 603 } 604 605 free (discarded[0]); 606 free (equiv_count[0]); 607 } 608 609 /* Adjust inserts/deletes of identical lines to join changes 610 as much as possible. 611 612 We do something when a run of changed lines include a 613 line at one end and have an excluded, identical line at the other. 614 We are free to choose which identical line is included. 615 `compareseq' usually chooses the one at the beginning, 616 but usually it is cleaner to consider the following identical line 617 to be the "change". */ 618 619 int inhibit; 620 621 static void 622 shift_boundaries (filevec) 623 struct file_data filevec[]; 624 { 625 int f; 626 int inhibit_hunk_merge = horizon_lines != context; 627 628 for (f = 0; f < 2; f++) 629 { 630 char *changed = filevec[f].changed_flag; 631 char const *other_changed = filevec[1-f].changed_flag; 632 int const *equivs = filevec[f].equivs; 633 int i = 0; 634 int j = 0; 635 int i_end = filevec[f].buffered_lines; 636 637 while (1) 638 { 639 int runlength, start, corresponding; 640 641 /* Scan forwards to find beginning of another run of changes. 642 Also keep track of the corresponding point in the other file. */ 643 644 while (i < i_end && changed[i] == 0) 645 { 646 while (other_changed[j++]) 647 continue; 648 i++; 649 } 650 651 if (i == i_end) 652 break; 653 654 start = i; 655 656 /* Find the end of this run of changes. */ 657 658 while (changed[++i]) 659 continue; 660 while (other_changed[j]) 661 j++; 662 663 do 664 { 665 /* Record the length of this run of changes, so that 666 we can later determine whether the run has grown. */ 667 runlength = i - start; 668 669 if (! inhibit_hunk_merge) 670 { 671 /* Move the changed region back, so long as the 672 previous unchanged line matches the last changed one. 673 This merges with previous changed regions. */ 674 675 while (start && equivs[start - 1] == equivs[i - 1]) 676 { 677 changed[--start] = 1; 678 changed[--i] = 0; 679 while (changed[start - 1]) 680 start--; 681 while (other_changed[--j]) 682 continue; 683 } 684 } 685 686 /* Set CORRESPONDING to the end of the changed run, at the last 687 point where it corresponds to a changed run in the other file. 688 CORRESPONDING == I_END means no such point has been found. */ 689 corresponding = other_changed[j - 1] ? i : i_end; 690 691 /* Shift the changed region forward, so long as the 692 first changed line matches the following unchanged one, 693 but if INHIBIT_HUNK_MERGE is 1 do not shift if 694 this would merge with another changed region. 695 Do this second, so that if there are no merges, 696 the changed region is moved forward as far as possible. */ 697 698 while (i != i_end && equivs[start] == equivs[i] 699 && ! (inhibit_hunk_merge & other_changed[j + 1])) 700 { 701 changed[start++] = 0; 702 changed[i++] = 1; 703 while (changed[i]) 704 i++; 705 while (other_changed[++j]) 706 corresponding = i; 707 } 708 } 709 while (runlength != i - start); 710 711 /* If possible, move the fully-merged run of changes 712 back to a corresponding run in the other file. */ 713 714 while (corresponding < i) 715 { 716 changed[--start] = 1; 717 changed[--i] = 0; 718 while (other_changed[--j]) 719 continue; 720 } 721 } 722 } 723 } 724 725 /* Cons an additional entry onto the front of an edit script OLD. 726 LINE0 and LINE1 are the first affected lines in the two files (origin 0). 727 DELETED is the number of lines deleted here from file 0. 728 INSERTED is the number of lines inserted here in file 1. 729 730 If DELETED is 0 then LINE0 is the number of the line before 731 which the insertion was done; vice versa for INSERTED and LINE1. */ 732 733 static struct change * 734 add_change (line0, line1, deleted, inserted, old) 735 int line0, line1, deleted, inserted; 736 struct change *old; 737 { 738 struct change *new = (struct change *) xmalloc (sizeof (struct change)); 739 740 new->line0 = line0; 741 new->line1 = line1; 742 new->inserted = inserted; 743 new->deleted = deleted; 744 new->link = old; 745 return new; 746 } 747 748 /* Scan the tables of which lines are inserted and deleted, 749 producing an edit script in reverse order. */ 750 751 static struct change * 752 build_reverse_script (filevec) 753 struct file_data const filevec[]; 754 { 755 struct change *script = 0; 756 char *changed0 = filevec[0].changed_flag; 757 char *changed1 = filevec[1].changed_flag; 758 int len0 = filevec[0].buffered_lines; 759 int len1 = filevec[1].buffered_lines; 760 761 /* Note that changedN[len0] does exist, and contains 0. */ 762 763 int i0 = 0, i1 = 0; 764 765 while (i0 < len0 || i1 < len1) 766 { 767 if (changed0[i0] || changed1[i1]) 768 { 769 int line0 = i0, line1 = i1; 770 771 /* Find # lines changed here in each file. */ 772 while (changed0[i0]) ++i0; 773 while (changed1[i1]) ++i1; 774 775 /* Record this change. */ 776 script = add_change (line0, line1, i0 - line0, i1 - line1, script); 777 } 778 779 /* We have reached lines in the two files that match each other. */ 780 i0++, i1++; 781 } 782 783 return script; 784 } 785 786 /* Scan the tables of which lines are inserted and deleted, 787 producing an edit script in forward order. */ 788 789 static struct change * 790 build_script (filevec) 791 struct file_data const filevec[]; 792 { 793 struct change *script = 0; 794 char *changed0 = filevec[0].changed_flag; 795 char *changed1 = filevec[1].changed_flag; 796 int i0 = filevec[0].buffered_lines, i1 = filevec[1].buffered_lines; 797 798 /* Note that changedN[-1] does exist, and contains 0. */ 799 800 while (i0 >= 0 || i1 >= 0) 801 { 802 if (changed0[i0 - 1] || changed1[i1 - 1]) 803 { 804 int line0 = i0, line1 = i1; 805 806 /* Find # lines changed here in each file. */ 807 while (changed0[i0 - 1]) --i0; 808 while (changed1[i1 - 1]) --i1; 809 810 /* Record this change. */ 811 script = add_change (i0, i1, line0 - i0, line1 - i1, script); 812 } 813 814 /* We have reached lines in the two files that match each other. */ 815 i0--, i1--; 816 } 817 818 return script; 819 } 820 821 /* If CHANGES, briefly report that two files differed. */ 822 static void 823 briefly_report (changes, filevec) 824 int changes; 825 struct file_data const filevec[]; 826 { 827 if (changes) 828 message (no_details_flag ? "Files %s and %s differ\n" 829 : "Binary files %s and %s differ\n", 830 filevec[0].name, filevec[1].name); 831 } 832 833 /* Report the differences of two files. DEPTH is the current directory 834 depth. */ 835 int 836 diff_2_files (filevec, depth) 837 struct file_data filevec[]; 838 int depth; 839 { 840 int diags; 841 int i; 842 struct change *e, *p; 843 struct change *script; 844 int changes; 845 846 847 /* If we have detected that either file is binary, 848 compare the two files as binary. This can happen 849 only when the first chunk is read. 850 Also, --brief without any --ignore-* options means 851 we can speed things up by treating the files as binary. */ 852 853 if (read_files (filevec, no_details_flag & ~ignore_some_changes)) 854 { 855 /* Files with different lengths must be different. */ 856 if (filevec[0].stat.st_size != filevec[1].stat.st_size 857 && (filevec[0].desc < 0 || S_ISREG (filevec[0].stat.st_mode)) 858 && (filevec[1].desc < 0 || S_ISREG (filevec[1].stat.st_mode))) 859 changes = 1; 860 861 /* Standard input equals itself. */ 862 else if (filevec[0].desc == filevec[1].desc) 863 changes = 0; 864 865 else 866 /* Scan both files, a buffer at a time, looking for a difference. */ 867 { 868 /* Allocate same-sized buffers for both files. */ 869 size_t buffer_size = buffer_lcm (STAT_BLOCKSIZE (filevec[0].stat), 870 STAT_BLOCKSIZE (filevec[1].stat)); 871 for (i = 0; i < 2; i++) 872 filevec[i].buffer = xrealloc (filevec[i].buffer, buffer_size); 873 874 for (;; filevec[0].buffered_chars = filevec[1].buffered_chars = 0) 875 { 876 /* Read a buffer's worth from both files. */ 877 for (i = 0; i < 2; i++) 878 if (0 <= filevec[i].desc) 879 while (filevec[i].buffered_chars != buffer_size) 880 { 881 int r = read (filevec[i].desc, 882 filevec[i].buffer 883 + filevec[i].buffered_chars, 884 buffer_size - filevec[i].buffered_chars); 885 if (r == 0) 886 break; 887 if (r < 0) 888 pfatal_with_name (filevec[i].name); 889 filevec[i].buffered_chars += r; 890 } 891 892 /* If the buffers differ, the files differ. */ 893 if (filevec[0].buffered_chars != filevec[1].buffered_chars 894 || (filevec[0].buffered_chars != 0 895 && memcmp (filevec[0].buffer, 896 filevec[1].buffer, 897 filevec[0].buffered_chars) != 0)) 898 { 899 changes = 1; 900 break; 901 } 902 903 /* If we reach end of file, the files are the same. */ 904 if (filevec[0].buffered_chars != buffer_size) 905 { 906 changes = 0; 907 break; 908 } 909 } 910 } 911 912 briefly_report (changes, filevec); 913 } 914 else 915 { 916 /* Allocate vectors for the results of comparison: 917 a flag for each line of each file, saying whether that line 918 is an insertion or deletion. 919 Allocate an extra element, always zero, at each end of each vector. */ 920 921 size_t s = filevec[0].buffered_lines + filevec[1].buffered_lines + 4; 922 filevec[0].changed_flag = xmalloc (s); 923 bzero (filevec[0].changed_flag, s); 924 filevec[0].changed_flag++; 925 filevec[1].changed_flag = filevec[0].changed_flag 926 + filevec[0].buffered_lines + 2; 927 928 /* Some lines are obviously insertions or deletions 929 because they don't match anything. Detect them now, and 930 avoid even thinking about them in the main comparison algorithm. */ 931 932 discard_confusing_lines (filevec); 933 934 /* Now do the main comparison algorithm, considering just the 935 undiscarded lines. */ 936 937 xvec = filevec[0].undiscarded; 938 yvec = filevec[1].undiscarded; 939 diags = filevec[0].nondiscarded_lines + filevec[1].nondiscarded_lines + 3; 940 fdiag = (int *) xmalloc (diags * (2 * sizeof (int))); 941 bdiag = fdiag + diags; 942 fdiag += filevec[1].nondiscarded_lines + 1; 943 bdiag += filevec[1].nondiscarded_lines + 1; 944 945 /* Set TOO_EXPENSIVE to be approximate square root of input size, 946 bounded below by 256. */ 947 too_expensive = 1; 948 for (i = filevec[0].nondiscarded_lines + filevec[1].nondiscarded_lines; 949 i != 0; i >>= 2) 950 too_expensive <<= 1; 951 too_expensive = max (256, too_expensive); 952 953 files[0] = filevec[0]; 954 files[1] = filevec[1]; 955 956 compareseq (0, filevec[0].nondiscarded_lines, 957 0, filevec[1].nondiscarded_lines, no_discards); 958 959 free (fdiag - (filevec[1].nondiscarded_lines + 1)); 960 961 /* Modify the results slightly to make them prettier 962 in cases where that can validly be done. */ 963 964 shift_boundaries (filevec); 965 966 /* Get the results of comparison in the form of a chain 967 of `struct change's -- an edit script. */ 968 969 if (output_style == OUTPUT_ED) 970 script = build_reverse_script (filevec); 971 else 972 script = build_script (filevec); 973 974 /* Set CHANGES if we had any diffs. 975 If some changes are ignored, we must scan the script to decide. */ 976 if (ignore_blank_lines_flag || ignore_regexp_list) 977 { 978 struct change *next = script; 979 changes = 0; 980 981 while (next && changes == 0) 982 { 983 struct change *this, *end; 984 int first0, last0, first1, last1, deletes, inserts; 985 986 /* Find a set of changes that belong together. */ 987 this = next; 988 end = find_change (next); 989 990 /* Disconnect them from the rest of the changes, making them 991 a hunk, and remember the rest for next iteration. */ 992 next = end->link; 993 end->link = 0; 994 995 /* Determine whether this hunk is really a difference. */ 996 analyze_hunk (this, &first0, &last0, &first1, &last1, 997 &deletes, &inserts); 998 999 /* Reconnect the script so it will all be freed properly. */ 1000 end->link = next; 1001 1002 if (deletes || inserts) 1003 changes = 1; 1004 } 1005 } 1006 else 1007 changes = (script != 0); 1008 1009 if (no_details_flag) 1010 briefly_report (changes, filevec); 1011 else 1012 { 1013 if (changes || ! no_diff_means_no_output) 1014 { 1015 /* Record info for starting up output, 1016 to be used if and when we have some output to print. */ 1017 setup_output (files[0].name, files[1].name, depth); 1018 1019 switch (output_style) 1020 { 1021 case OUTPUT_CONTEXT: 1022 print_context_script (script, 0); 1023 break; 1024 1025 case OUTPUT_UNIFIED: 1026 print_context_script (script, 1); 1027 break; 1028 1029 case OUTPUT_ED: 1030 print_ed_script (script); 1031 break; 1032 1033 case OUTPUT_FORWARD_ED: 1034 pr_forward_ed_script (script); 1035 break; 1036 1037 case OUTPUT_RCS: 1038 print_rcs_script (script); 1039 break; 1040 1041 case OUTPUT_NORMAL: 1042 print_normal_script (script); 1043 break; 1044 1045 case OUTPUT_IFDEF: 1046 print_ifdef_script (script); 1047 break; 1048 1049 case OUTPUT_SDIFF: 1050 print_sdiff_script (script); 1051 } 1052 1053 finish_output (); 1054 } 1055 } 1056 1057 free (filevec[0].undiscarded); 1058 1059 free (filevec[0].changed_flag - 1); 1060 1061 for (i = 1; i >= 0; --i) 1062 free (filevec[i].equivs); 1063 1064 for (i = 0; i < 2; ++i) 1065 free (filevec[i].linbuf + filevec[i].linbuf_base); 1066 1067 for (e = script; e; e = p) 1068 { 1069 p = e->link; 1070 free (e); 1071 } 1072 1073 if (! ROBUST_OUTPUT_STYLE (output_style)) 1074 for (i = 0; i < 2; ++i) 1075 if (filevec[i].missing_newline) 1076 { 1077 diff_error ("No newline at end of file %s", filevec[i].name, ""); 1078 changes = 2; 1079 } 1080 } 1081 1082 if (filevec[0].buffer != filevec[1].buffer) 1083 free (filevec[0].buffer); 1084 free (filevec[1].buffer); 1085 1086 return changes; 1087 } 1088