lcs.c revision 299742
1/* 2 * lcs.c : routines for creating an lcs 3 * 4 * ==================================================================== 5 * Licensed to the Apache Software Foundation (ASF) under one 6 * or more contributor license agreements. See the NOTICE file 7 * distributed with this work for additional information 8 * regarding copyright ownership. The ASF licenses this file 9 * to you under the Apache License, Version 2.0 (the 10 * "License"); you may not use this file except in compliance 11 * with the License. You may obtain a copy of the License at 12 * 13 * http://www.apache.org/licenses/LICENSE-2.0 14 * 15 * Unless required by applicable law or agreed to in writing, 16 * software distributed under the License is distributed on an 17 * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY 18 * KIND, either express or implied. See the License for the 19 * specific language governing permissions and limitations 20 * under the License. 21 * ==================================================================== 22 */ 23 24 25#include <apr.h> 26#include <apr_pools.h> 27#include <apr_general.h> 28 29#include "diff.h" 30 31 32/* 33 * Calculate the Longest Common Subsequence (LCS) between two datasources. 34 * This function is what makes the diff code tick. 35 * 36 * The LCS algorithm implemented here is based on the approach described 37 * by Sun Wu, Udi Manber and Gene Meyers in "An O(NP) Sequence Comparison 38 * Algorithm", but has been modified for better performance. 39 * 40 * Let M and N be the lengths (number of tokens) of the two sources 41 * ('files'). The goal is to reach the end of both sources (files) with the 42 * minimum number of insertions + deletions. Since there is a known length 43 * difference N-M between the files, that is equivalent to just the minimum 44 * number of deletions, or equivalently the minimum number of insertions. 45 * For symmetry, we use the lesser number - deletions if M<N, insertions if 46 * M>N. 47 * 48 * Let 'k' be the difference in remaining length between the files, i.e. 49 * if we're at the beginning of both files, k=N-M, whereas k=0 for the 50 * 'end state', at the end of both files. An insertion will increase k by 51 * one, while a deletion decreases k by one. If k<0, then insertions are 52 * 'free' - we need those to reach the end state k=0 anyway - but deletions 53 * are costly: Adding a deletion means that we will have to add an additional 54 * insertion later to reach the end state, so it doesn't matter if we count 55 * deletions or insertions. Similarly, deletions are free for k>0. 56 * 57 * Let a 'state' be a given position in each file {pos1, pos2}. An array 58 * 'fp' keeps track of the best possible state (largest values of 59 * {pos1, pos2}) that can be achieved for a given cost 'p' (# moves away 60 * from k=0), as well as a linked list of what matches were used to reach 61 * that state. For each new value of p, we find for each value of k the 62 * best achievable state for that k - either by doing a costly operation 63 * (deletion if k<0) from a state achieved at a lower p, or doing a free 64 * operation (insertion if k<0) from a state achieved at the same p - 65 * and in both cases advancing past any matching regions found. This is 66 * handled by running loops over k in order of descending absolute value. 67 * 68 * A recent improvement of the algorithm is to ignore tokens that are unique 69 * to one file or the other, as those are known from the start to be 70 * impossible to match. 71 */ 72 73typedef struct svn_diff__snake_t svn_diff__snake_t; 74 75struct svn_diff__snake_t 76{ 77 apr_off_t y; 78 svn_diff__lcs_t *lcs; 79 svn_diff__position_t *position[2]; 80}; 81 82static APR_INLINE void 83svn_diff__snake(svn_diff__snake_t *fp_k, 84 svn_diff__token_index_t *token_counts[2], 85 svn_diff__lcs_t **freelist, 86 apr_pool_t *pool) 87{ 88 svn_diff__position_t *start_position[2]; 89 svn_diff__position_t *position[2]; 90 svn_diff__lcs_t *lcs; 91 svn_diff__lcs_t *previous_lcs; 92 93 /* The previous entry at fp[k] is going to be replaced. See if we 94 * can mark that lcs node for reuse, because the sequence up to this 95 * point was a dead end. 96 */ 97 lcs = fp_k[0].lcs; 98 while (lcs) 99 { 100 lcs->refcount--; 101 if (lcs->refcount) 102 break; 103 104 previous_lcs = lcs->next; 105 lcs->next = *freelist; 106 *freelist = lcs; 107 lcs = previous_lcs; 108 } 109 110 if (fp_k[-1].y >= fp_k[1].y) 111 { 112 start_position[0] = fp_k[-1].position[0]; 113 start_position[1] = fp_k[-1].position[1]->next; 114 115 previous_lcs = fp_k[-1].lcs; 116 } 117 else 118 { 119 start_position[0] = fp_k[1].position[0]->next; 120 start_position[1] = fp_k[1].position[1]; 121 122 previous_lcs = fp_k[1].lcs; 123 } 124 125 126 if (previous_lcs) 127 { 128 previous_lcs->refcount++; 129 } 130 131 /* ### Optimization, skip all positions that don't have matchpoints 132 * ### anyway. Beware of the sentinel, don't skip it! 133 */ 134 135 position[0] = start_position[0]; 136 position[1] = start_position[1]; 137 138 while (1) 139 { 140 while (position[0]->token_index == position[1]->token_index) 141 { 142 position[0] = position[0]->next; 143 position[1] = position[1]->next; 144 } 145 146 if (position[1] != start_position[1]) 147 { 148 lcs = *freelist; 149 if (lcs) 150 { 151 *freelist = lcs->next; 152 } 153 else 154 { 155 lcs = apr_palloc(pool, sizeof(*lcs)); 156 } 157 158 lcs->position[0] = start_position[0]; 159 lcs->position[1] = start_position[1]; 160 lcs->length = position[1]->offset - start_position[1]->offset; 161 lcs->next = previous_lcs; 162 lcs->refcount = 1; 163 previous_lcs = lcs; 164 start_position[0] = position[0]; 165 start_position[1] = position[1]; 166 } 167 168 /* Skip any and all tokens that only occur in one of the files */ 169 if (position[0]->token_index >= 0 170 && token_counts[1][position[0]->token_index] == 0) 171 start_position[0] = position[0] = position[0]->next; 172 else if (position[1]->token_index >= 0 173 && token_counts[0][position[1]->token_index] == 0) 174 start_position[1] = position[1] = position[1]->next; 175 else 176 break; 177 } 178 179 fp_k[0].lcs = previous_lcs; 180 fp_k[0].position[0] = position[0]; 181 fp_k[0].position[1] = position[1]; 182 183 fp_k[0].y = position[1]->offset; 184} 185 186 187static svn_diff__lcs_t * 188svn_diff__lcs_reverse(svn_diff__lcs_t *lcs) 189{ 190 svn_diff__lcs_t *next; 191 svn_diff__lcs_t *prev; 192 193 next = NULL; 194 while (lcs != NULL) 195 { 196 prev = lcs->next; 197 lcs->next = next; 198 next = lcs; 199 lcs = prev; 200 } 201 202 return next; 203} 204 205 206/* Prepends a new lcs chunk for the amount of LINES at the given positions 207 * POS0_OFFSET and POS1_OFFSET to the given LCS chain, and returns it. 208 * This function assumes LINES > 0. */ 209static svn_diff__lcs_t * 210prepend_lcs(svn_diff__lcs_t *lcs, apr_off_t lines, 211 apr_off_t pos0_offset, apr_off_t pos1_offset, 212 apr_pool_t *pool) 213{ 214 svn_diff__lcs_t *new_lcs; 215 216 SVN_ERR_ASSERT_NO_RETURN(lines > 0); 217 218 new_lcs = apr_palloc(pool, sizeof(*new_lcs)); 219 new_lcs->position[0] = apr_pcalloc(pool, sizeof(*new_lcs->position[0])); 220 new_lcs->position[0]->offset = pos0_offset; 221 new_lcs->position[1] = apr_pcalloc(pool, sizeof(*new_lcs->position[1])); 222 new_lcs->position[1]->offset = pos1_offset; 223 new_lcs->length = lines; 224 new_lcs->refcount = 1; 225 new_lcs->next = lcs; 226 227 return new_lcs; 228} 229 230 231svn_diff__lcs_t * 232svn_diff__lcs(svn_diff__position_t *position_list1, /* pointer to tail (ring) */ 233 svn_diff__position_t *position_list2, /* pointer to tail (ring) */ 234 svn_diff__token_index_t *token_counts_list1, /* array of counts */ 235 svn_diff__token_index_t *token_counts_list2, /* array of counts */ 236 svn_diff__token_index_t num_tokens, 237 apr_off_t prefix_lines, 238 apr_off_t suffix_lines, 239 apr_pool_t *pool) 240{ 241 apr_off_t length[2]; 242 svn_diff__token_index_t *token_counts[2]; 243 svn_diff__token_index_t unique_count[2]; 244 svn_diff__token_index_t token_index; 245 svn_diff__snake_t *fp; 246 apr_off_t d; 247 apr_off_t k; 248 apr_off_t p = 0; 249 svn_diff__lcs_t *lcs, *lcs_freelist = NULL; 250 251 svn_diff__position_t sentinel_position[2]; 252 253 /* Since EOF is always a sync point we tack on an EOF link 254 * with sentinel positions 255 */ 256 lcs = apr_palloc(pool, sizeof(*lcs)); 257 lcs->position[0] = apr_pcalloc(pool, sizeof(*lcs->position[0])); 258 lcs->position[0]->offset = position_list1 259 ? position_list1->offset + suffix_lines + 1 260 : prefix_lines + suffix_lines + 1; 261 lcs->position[1] = apr_pcalloc(pool, sizeof(*lcs->position[1])); 262 lcs->position[1]->offset = position_list2 263 ? position_list2->offset + suffix_lines + 1 264 : prefix_lines + suffix_lines + 1; 265 lcs->length = 0; 266 lcs->refcount = 1; 267 lcs->next = NULL; 268 269 if (position_list1 == NULL || position_list2 == NULL) 270 { 271 if (suffix_lines) 272 lcs = prepend_lcs(lcs, suffix_lines, 273 lcs->position[0]->offset - suffix_lines, 274 lcs->position[1]->offset - suffix_lines, 275 pool); 276 if (prefix_lines) 277 lcs = prepend_lcs(lcs, prefix_lines, 1, 1, pool); 278 279 return lcs; 280 } 281 282 unique_count[1] = unique_count[0] = 0; 283 for (token_index = 0; token_index < num_tokens; token_index++) 284 { 285 if (token_counts_list1[token_index] == 0) 286 unique_count[1] += token_counts_list2[token_index]; 287 if (token_counts_list2[token_index] == 0) 288 unique_count[0] += token_counts_list1[token_index]; 289 } 290 291 /* Calculate lengths M and N of the sequences to be compared. Do not 292 * count tokens unique to one file, as those are ignored in __snake. 293 */ 294 length[0] = position_list1->offset - position_list1->next->offset + 1 295 - unique_count[0]; 296 length[1] = position_list2->offset - position_list2->next->offset + 1 297 - unique_count[1]; 298 299 /* strikerXXX: here we allocate the furthest point array, which is 300 * strikerXXX: sized M + N + 3 (!) 301 */ 302 fp = apr_pcalloc(pool, 303 sizeof(*fp) * (apr_size_t)(length[0] + length[1] + 3)); 304 305 /* The origo of fp corresponds to the end state, where we are 306 * at the end of both files. The valid states thus span from 307 * -N (at end of first file and at the beginning of the second 308 * file) to +M (the opposite :). Finally, svn_diff__snake needs 309 * 1 extra slot on each side to work. 310 */ 311 fp += length[1] + 1; 312 313 sentinel_position[0].next = position_list1->next; 314 position_list1->next = &sentinel_position[0]; 315 sentinel_position[0].offset = position_list1->offset + 1; 316 token_counts[0] = token_counts_list1; 317 318 sentinel_position[1].next = position_list2->next; 319 position_list2->next = &sentinel_position[1]; 320 sentinel_position[1].offset = position_list2->offset + 1; 321 token_counts[1] = token_counts_list2; 322 323 /* Negative indices will not be used elsewhere 324 */ 325 sentinel_position[0].token_index = -1; 326 sentinel_position[1].token_index = -2; 327 328 /* position d = M - N corresponds to the initial state, where 329 * we are at the beginning of both files. 330 */ 331 d = length[0] - length[1]; 332 333 /* k = d - 1 will be the first to be used to get previous 334 * position information from, make sure it holds sane 335 * data 336 */ 337 fp[d - 1].position[0] = sentinel_position[0].next; 338 fp[d - 1].position[1] = &sentinel_position[1]; 339 340 p = 0; 341 do 342 { 343 /* For k < 0, insertions are free */ 344 for (k = (d < 0 ? d : 0) - p; k < 0; k++) 345 { 346 svn_diff__snake(fp + k, token_counts, &lcs_freelist, pool); 347 } 348 /* for k > 0, deletions are free */ 349 for (k = (d > 0 ? d : 0) + p; k >= 0; k--) 350 { 351 svn_diff__snake(fp + k, token_counts, &lcs_freelist, pool); 352 } 353 354 p++; 355 } 356 while (fp[0].position[1] != &sentinel_position[1]); 357 358 if (suffix_lines) 359 lcs->next = prepend_lcs(fp[0].lcs, suffix_lines, 360 lcs->position[0]->offset - suffix_lines, 361 lcs->position[1]->offset - suffix_lines, 362 pool); 363 else 364 lcs->next = fp[0].lcs; 365 366 lcs = svn_diff__lcs_reverse(lcs); 367 368 position_list1->next = sentinel_position[0].next; 369 position_list2->next = sentinel_position[1].next; 370 371 if (prefix_lines) 372 return prepend_lcs(lcs, prefix_lines, 1, 1, pool); 373 else 374 return lcs; 375} 376