1b2441318SGreg Kroah-Hartman // SPDX-License-Identifier: GPL-2.0 286470930SIngo Molnar #include "levenshtein.h" 3175729fcSArnaldo Carvalho de Melo #include <errno.h> 4175729fcSArnaldo Carvalho de Melo #include <stdlib.h> 5175729fcSArnaldo Carvalho de Melo #include <string.h> 686470930SIngo Molnar 786470930SIngo Molnar /* 886470930SIngo Molnar * This function implements the Damerau-Levenshtein algorithm to 986470930SIngo Molnar * calculate a distance between strings. 1086470930SIngo Molnar * 1186470930SIngo Molnar * Basically, it says how many letters need to be swapped, substituted, 1286470930SIngo Molnar * deleted from, or added to string1, at least, to get string2. 1386470930SIngo Molnar * 1486470930SIngo Molnar * The idea is to build a distance matrix for the substrings of both 1586470930SIngo Molnar * strings. To avoid a large space complexity, only the last three rows 1686470930SIngo Molnar * are kept in memory (if swaps had the same or higher cost as one deletion 1786470930SIngo Molnar * plus one insertion, only two rows would be needed). 1886470930SIngo Molnar * 1986470930SIngo Molnar * At any stage, "i + 1" denotes the length of the current substring of 2086470930SIngo Molnar * string1 that the distance is calculated for. 2186470930SIngo Molnar * 2286470930SIngo Molnar * row2 holds the current row, row1 the previous row (i.e. for the substring 2386470930SIngo Molnar * of string1 of length "i"), and row0 the row before that. 2486470930SIngo Molnar * 2586470930SIngo Molnar * In other words, at the start of the big loop, row2[j + 1] contains the 2686470930SIngo Molnar * Damerau-Levenshtein distance between the substring of string1 of length 2786470930SIngo Molnar * "i" and the substring of string2 of length "j + 1". 2886470930SIngo Molnar * 2986470930SIngo Molnar * All the big loop does is determine the partial minimum-cost paths. 3086470930SIngo Molnar * 3186470930SIngo Molnar * It does so by calculating the costs of the path ending in characters 3286470930SIngo Molnar * i (in string1) and j (in string2), respectively, given that the last 3386470930SIngo Molnar * operation is a substition, a swap, a deletion, or an insertion. 3486470930SIngo Molnar * 3586470930SIngo Molnar * This implementation allows the costs to be weighted: 3686470930SIngo Molnar * 3786470930SIngo Molnar * - w (as in "sWap") 3886470930SIngo Molnar * - s (as in "Substitution") 3986470930SIngo Molnar * - a (for insertion, AKA "Add") 4086470930SIngo Molnar * - d (as in "Deletion") 4186470930SIngo Molnar * 4286470930SIngo Molnar * Note that this algorithm calculates a distance _iff_ d == a. 4386470930SIngo Molnar */ 4486470930SIngo Molnar int levenshtein(const char *string1, const char *string2, 4586470930SIngo Molnar int w, int s, int a, int d) 4686470930SIngo Molnar { 4786470930SIngo Molnar int len1 = strlen(string1), len2 = strlen(string2); 4886470930SIngo Molnar int *row0 = malloc(sizeof(int) * (len2 + 1)); 4986470930SIngo Molnar int *row1 = malloc(sizeof(int) * (len2 + 1)); 5086470930SIngo Molnar int *row2 = malloc(sizeof(int) * (len2 + 1)); 5186470930SIngo Molnar int i, j; 5286470930SIngo Molnar 5386470930SIngo Molnar for (j = 0; j <= len2; j++) 5486470930SIngo Molnar row1[j] = j * a; 5586470930SIngo Molnar for (i = 0; i < len1; i++) { 5686470930SIngo Molnar int *dummy; 5786470930SIngo Molnar 5886470930SIngo Molnar row2[0] = (i + 1) * d; 5986470930SIngo Molnar for (j = 0; j < len2; j++) { 6086470930SIngo Molnar /* substitution */ 6186470930SIngo Molnar row2[j + 1] = row1[j] + s * (string1[i] != string2[j]); 6286470930SIngo Molnar /* swap */ 6386470930SIngo Molnar if (i > 0 && j > 0 && string1[i - 1] == string2[j] && 6486470930SIngo Molnar string1[i] == string2[j - 1] && 6586470930SIngo Molnar row2[j + 1] > row0[j - 1] + w) 6686470930SIngo Molnar row2[j + 1] = row0[j - 1] + w; 6786470930SIngo Molnar /* deletion */ 6886470930SIngo Molnar if (row2[j + 1] > row1[j + 1] + d) 6986470930SIngo Molnar row2[j + 1] = row1[j + 1] + d; 7086470930SIngo Molnar /* insertion */ 7186470930SIngo Molnar if (row2[j + 1] > row2[j] + a) 7286470930SIngo Molnar row2[j + 1] = row2[j] + a; 7386470930SIngo Molnar } 7486470930SIngo Molnar 7586470930SIngo Molnar dummy = row0; 7686470930SIngo Molnar row0 = row1; 7786470930SIngo Molnar row1 = row2; 7886470930SIngo Molnar row2 = dummy; 7986470930SIngo Molnar } 8086470930SIngo Molnar 8186470930SIngo Molnar i = row1[len2]; 8286470930SIngo Molnar free(row0); 8386470930SIngo Molnar free(row1); 8486470930SIngo Molnar free(row2); 8586470930SIngo Molnar 8686470930SIngo Molnar return i; 8786470930SIngo Molnar } 88