1 /* Functions to make fuzzy comparisons between strings
2 Copyright (C) 1988-1989, 1992-1993, 1995, 2001-2003, 2006 Free Software Foundation, Inc.
4 This program is free software: you can redistribute it and/or modify
5 it under the terms of the GNU General Public License as published by
6 the Free Software Foundation; either version 3 of the License, or
7 (at your option) any later version.
9 This program is distributed in the hope that it will be useful,
10 but WITHOUT ANY WARRANTY; without even the implied warranty of
11 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 GNU General Public License for more details.
14 You should have received a copy of the GNU General Public License
15 along with this program. If not, see <http://www.gnu.org/licenses/>.
18 Derived from GNU diff 2.7, analyze.c et al.
20 The basic idea is to consider two vectors as similar if, when
21 transforming the first vector into the second vector through a
22 sequence of edits (inserts and deletes of one element each),
23 this sequence is short - or equivalently, if the ordered list
24 of elements that are untouched by these edits is long. For a
25 good introduction to the subject, read about the "Levenshtein
26 distance" in Wikipedia.
28 The basic algorithm is described in:
29 "An O(ND) Difference Algorithm and its Variations", Eugene Myers,
30 Algorithmica Vol. 1 No. 2, 1986, pp. 251-266;
31 see especially section 4.2, which describes the variation used below.
33 The basic algorithm was independently discovered as described in:
34 "Algorithms for Approximate String Matching", E. Ukkonen,
35 Information and Control Vol. 64, 1985, pp. 100-118.
37 Unless the 'find_minimal' flag is set, this code uses the TOO_EXPENSIVE
38 heuristic, by Paul Eggert, to limit the cost to O(N**1.5 log N)
39 at the price of producing suboptimal output for large inputs with
59 # define uintptr_t unsigned long
64 #define EQUAL(x,y) ((x) == (y))
66 #define EXTRA_CONTEXT_FIELDS \
67 /* The number of elements inserted or deleted. */ \
68 int xvec_edit_count; \
70 #define NOTE_DELETE(ctxt, xoff) ctxt->xvec_edit_count++
71 #define NOTE_INSERT(ctxt, yoff) ctxt->yvec_edit_count++
72 /* We don't need USE_HEURISTIC, since it is unlikely in typical uses of
77 /* Because fstrcmp is typically called multiple times, attempt to minimize
78 the number of memory allocations performed. Thus, let a call reuse the
79 memory already allocated by the previous call, if it is sufficient.
80 To make it multithread-safe, without need for a lock that protects the
81 already allocated memory, store the allocated memory per thread. Free
82 it only when the thread exits. */
84 static gl_tls_key_t buffer_key; /* TLS key for a 'int *' */
85 static gl_tls_key_t bufmax_key; /* TLS key for a 'size_t' */
90 gl_tls_key_init (buffer_key, free);
91 gl_tls_key_init (bufmax_key, NULL);
92 /* The per-thread initial values are NULL and 0, respectively. */
95 /* Ensure that keys_init is called once only. */
96 gl_once_define(static, keys_init_once)
100 fstrcmp - fuzzy string compare
103 double fstrcmp(const char *, const char *);
106 The fstrcmp function may be used to compare two string for
107 similarity. It is very useful in reducing "cascade" or
108 "secondary" errors in compilers or other situations where
112 double; 0 if the strings are entirly dissimilar, 1 if the
113 strings are identical, and a number in between if they are
117 fstrcmp (const char *string1, const char *string2)
128 /* set the info for each string. */
130 xvec_length = strlen (string1);
132 yvec_length = strlen (string2);
134 /* short-circuit obvious comparisons */
135 if (xvec_length == 0 && yvec_length == 0)
137 if (xvec_length == 0 || yvec_length == 0)
140 /* Set TOO_EXPENSIVE to be approximate square root of input size,
141 bounded below by 256. */
142 ctxt.too_expensive = 1;
143 for (i = xvec_length + yvec_length;
146 ctxt.too_expensive <<= 1;
147 if (ctxt.too_expensive < 256)
148 ctxt.too_expensive = 256;
150 /* Allocate memory for fdiag and bdiag from a thread-local pool. */
151 fdiag_len = xvec_length + yvec_length + 3;
152 gl_once (keys_init_once, keys_init);
153 buffer = (int *) gl_tls_get (buffer_key);
154 bufmax = (size_t) (uintptr_t) gl_tls_get (bufmax_key);
155 if (fdiag_len > bufmax)
157 /* Need more memory. */
159 if (fdiag_len > bufmax)
161 /* Calling xrealloc would be a waste: buffer's contents does not need
165 buffer = (int *) xnmalloc (bufmax, 2 * sizeof (int));
166 gl_tls_set (buffer_key, buffer);
167 gl_tls_set (bufmax_key, (void *) (uintptr_t) bufmax);
169 ctxt.fdiag = buffer + yvec_length + 1;
170 ctxt.bdiag = ctxt.fdiag + fdiag_len;
172 /* Now do the main comparison algorithm */
173 ctxt.xvec_edit_count = 0;
174 ctxt.yvec_edit_count = 0;
175 compareseq (0, xvec_length, 0, yvec_length, 0,
179 ((number of chars in common) / (average length of the strings)).
180 This is admittedly biased towards finding that the strings are
181 similar, however it does produce meaningful results. */
182 return ((double) (xvec_length + yvec_length
183 - ctxt.yvec_edit_count - ctxt.xvec_edit_count)
184 / (xvec_length + yvec_length));