• Home
  • History
  • Annotate
  • Line#
  • Navigate
  • Raw
  • Download
  • only in /netgear-WNDR4500v2-V1.0.0.60_1.0.38/ap/gpl/timemachine/gettext-0.17/gettext-tools/gnulib-lib/
1/* Functions to make fuzzy comparisons between strings
2   Copyright (C) 1988-1989, 1992-1993, 1995, 2001-2003, 2006 Free Software Foundation, Inc.
3
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.
8
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.
13
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/>.
16
17
18   Derived from GNU diff 2.7, analyze.c et al.
19
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.
27
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.
32
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.
36
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
40   many differences.  */
41
42#include <config.h>
43
44/* Specification.  */
45#include "fstrcmp.h"
46
47#include <string.h>
48#include <stdbool.h>
49#include <stdio.h>
50#include <stdlib.h>
51#include <limits.h>
52
53#include "lock.h"
54#include "tls.h"
55#include "minmax.h"
56#include "xalloc.h"
57
58#ifndef uintptr_t
59# define uintptr_t unsigned long
60#endif
61
62
63#define ELEMENT char
64#define EQUAL(x,y) ((x) == (y))
65#define OFFSET int
66#define EXTRA_CONTEXT_FIELDS \
67  /* The number of elements inserted or deleted. */ \
68  int xvec_edit_count; \
69  int yvec_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
73   fstrcmp().  */
74#include "diffseq.h"
75
76
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.  */
83
84static gl_tls_key_t buffer_key;	/* TLS key for a 'int *' */
85static gl_tls_key_t bufmax_key;	/* TLS key for a 'size_t' */
86
87static void
88keys_init (void)
89{
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.  */
93}
94
95/* Ensure that keys_init is called once only.  */
96gl_once_define(static, keys_init_once)
97
98
99/* NAME
100	fstrcmp - fuzzy string compare
101
102   SYNOPSIS
103	double fstrcmp(const char *, const char *);
104
105   DESCRIPTION
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
109	symbol tables occur.
110
111   RETURNS
112	double; 0 if the strings are entirly dissimilar, 1 if the
113	strings are identical, and a number in between if they are
114	similar.  */
115
116double
117fstrcmp (const char *string1, const char *string2)
118{
119  struct context ctxt;
120  int xvec_length;
121  int yvec_length;
122  int i;
123
124  size_t fdiag_len;
125  int *buffer;
126  size_t bufmax;
127
128  /* set the info for each string.  */
129  ctxt.xvec = string1;
130  xvec_length = strlen (string1);
131  ctxt.yvec = string2;
132  yvec_length = strlen (string2);
133
134  /* short-circuit obvious comparisons */
135  if (xvec_length == 0 && yvec_length == 0)
136    return 1.0;
137  if (xvec_length == 0 || yvec_length == 0)
138    return 0.0;
139
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;
144       i != 0;
145       i >>= 2)
146    ctxt.too_expensive <<= 1;
147  if (ctxt.too_expensive < 256)
148    ctxt.too_expensive = 256;
149
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)
156    {
157      /* Need more memory.  */
158      bufmax = 2 * bufmax;
159      if (fdiag_len > bufmax)
160	bufmax = fdiag_len;
161      /* Calling xrealloc would be a waste: buffer's contents does not need
162	 to be preserved.  */
163      if (buffer != NULL)
164	free (buffer);
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);
168    }
169  ctxt.fdiag = buffer + yvec_length + 1;
170  ctxt.bdiag = ctxt.fdiag + fdiag_len;
171
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,
176	      &ctxt);
177
178  /* The result is
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));
185}
186