gcc/spellcheck.c

   1 /* Find near-matches for strings.
   2    Copyright (C) 2015-2016 Free Software Foundation, Inc.
   3
   4 This file is part of GCC.
   5
   6 GCC is free software; you can redistribute it and/or modify it under
   7 the terms of the GNU General Public License as published by the Free
   8 Software Foundation; either version 3, or (at your option) any later
   9 version.
  10
  11 GCC is distributed in the hope that it will be useful, but WITHOUT ANY
  12 WARRANTY; without even the implied warranty of MERCHANTABILITY or
  13 FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
  14 for more details.
  15
  16 You should have received a copy of the GNU General Public License
  17 along with GCC; see the file COPYING3.  If not see
  18 <http://www.gnu.org/licenses/>.  */
  19
  20 #include "config.h"
  21 #include "system.h"
  22 #include "coretypes.h"
  23 #include "tm.h"
  24 #include "tree.h"
  25 #include "spellcheck.h"
  26 #include "selftest.h"
  27
  28 /* The Levenshtein distance is an "edit-distance": the minimal
  29    number of one-character insertions, removals or substitutions
  30    that are needed to change one string into another.
  31
  32    This implementation uses the Wagner-Fischer algorithm.  */
  33
  34 edit_distance_t
  35 levenshtein_distance (const char *s, int len_s,
  36                       const char *t, int len_t)
  37 {
  38   const bool debug = false;
  39
  40   if (debug)
  41     {
  42       printf ("s: \"%s\" (len_s=%i)\n", s, len_s);
  43       printf ("t: \"%s\" (len_t=%i)\n", t, len_t);
  44     }
  45
  46   if (len_s == 0)
  47     return len_t;
  48   if (len_t == 0)
  49     return len_s;
  50
  51   /* We effectively build a matrix where each (i, j) contains the
  52      Levenshtein distance between the prefix strings s[0:j]
  53      and t[0:i].
  54      Rather than actually build an (len_t + 1) * (len_s + 1) matrix,
  55      we simply keep track of the last row, v0 and a new row, v1,
  56      which avoids an (len_t + 1) * (len_s + 1) allocation and memory accesses
  57      in favor of two (len_s + 1) allocations.  These could potentially be
  58      statically-allocated if we impose a maximum length on the
  59      strings of interest.  */
  60   edit_distance_t *v0 = new edit_distance_t[len_s + 1];
  61   edit_distance_t *v1 = new edit_distance_t[len_s + 1];
  62
  63   /* The first row is for the case of an empty target string, which
  64      we can reach by deleting every character in the source string.  */
  65   for (int i = 0; i < len_s + 1; i++)
  66     v0[i] = i;
  67
  68   /* Build successive rows.  */
  69   for (int i = 0; i < len_t; i++)
  70     {
  71       if (debug)
  72         {
  73           printf ("i:%i v0 = ", i);
  74           for (int j = 0; j < len_s + 1; j++)
  75             printf ("%i ", v0[j]);
  76           printf ("\n");
  77         }
  78
  79       /* The initial column is for the case of an empty source string; we
  80          can reach prefixes of the target string of length i
  81          by inserting i characters.  */
  82       v1[0] = i + 1;
  83
  84       /* Build the rest of the row by considering neighbors to
  85          the north, west and northwest.  */
  86       for (int j = 0; j < len_s; j++)
  87         {
  88           edit_distance_t cost = (s[j] == t[i] ? 0 : 1);
  89           edit_distance_t deletion     = v1[j] + 1;
  90           edit_distance_t insertion    = v0[j + 1] + 1;
  91           edit_distance_t substitution = v0[j] + cost;
  92           edit_distance_t cheapest = MIN (deletion, insertion);
  93           cheapest = MIN (cheapest, substitution);
  94           v1[j + 1] = cheapest;
  95         }
  96
  97       /* Prepare to move on to next row.  */
  98       for (int j = 0; j < len_s + 1; j++)
  99         v0[j] = v1[j];
 100     }
 101
 102   if (debug)
 103     {
 104       printf ("final v1 = ");
 105       for (int j = 0; j < len_s + 1; j++)
 106         printf ("%i ", v1[j]);
 107       printf ("\n");
 108     }
 109
 110   edit_distance_t result = v1[len_s];
 111   delete[] v0;
 112   delete[] v1;
 113   return result;
 114 }
 115
 116 /* Calculate Levenshtein distance between two nil-terminated strings.  */
 117
 118 edit_distance_t
 119 levenshtein_distance (const char *s, const char *t)
 120 {
 121   return levenshtein_distance (s, strlen (s), t, strlen (t));
 122 }
 123
 124 /* Given TARGET, a non-NULL string, and CANDIDATES, a non-NULL ptr to
 125    an autovec of non-NULL strings, determine which element within
 126    CANDIDATES has the lowest edit distance to TARGET.  If there are
 127    multiple elements with the same minimal distance, the first in the
 128    vector wins.
 129
 130    If more than half of the letters were misspelled, the suggestion is
 131    likely to be meaningless, so return NULL for this case.  */
 132
 133 const char *
 134 find_closest_string (const char *target,
 135                      const auto_vec<const char *> *candidates)
 136 {
 137   gcc_assert (target);
 138   gcc_assert (candidates);
 139
 140   int i;
 141   const char *candidate;
 142   const char *best_candidate = NULL;
 143   edit_distance_t best_distance = MAX_EDIT_DISTANCE;
 144   size_t len_target = strlen (target);
 145   FOR_EACH_VEC_ELT (*candidates, i, candidate)
 146     {
 147       gcc_assert (candidate);
 148       edit_distance_t dist
 149         = levenshtein_distance (target, len_target,
 150                                 candidate, strlen (candidate));
 151       if (dist < best_distance)
 152         {
 153           best_distance = dist;
 154           best_candidate = candidate;
 155         }
 156     }
 157
 158   /* If more than half of the letters were misspelled, the suggestion is
 159      likely to be meaningless.  */
 160   if (best_candidate)
 161     {
 162       unsigned int cutoff = MAX (len_target, strlen (best_candidate)) / 2;
 163       if (best_distance > cutoff)
 164         return NULL;
 165     }
 166
 167   return best_candidate;
 168 }
 169
 170 #if CHECKING_P
 171
 172 namespace selftest {
 173
 174 /* Selftests.  */
 175
 176 /* Verify that the levenshtein_distance (A, B) equals the expected
 177    value.  */
 178
 179 static void
 180 levenshtein_distance_unit_test_oneway (const char *a, const char *b,
 181                                        edit_distance_t expected)
 182 {
 183   edit_distance_t actual = levenshtein_distance (a, b);
 184   ASSERT_EQ (actual, expected);
 185 }
 186
 187 /* Verify that both
 188      levenshtein_distance (A, B)
 189    and
 190      levenshtein_distance (B, A)
 191    equal the expected value, to ensure that the function is symmetric.  */
 192
 193 static void
 194 levenshtein_distance_unit_test (const char *a, const char *b,
 195                                 edit_distance_t expected)
 196 {
 197   levenshtein_distance_unit_test_oneway (a, b, expected);
 198   levenshtein_distance_unit_test_oneway (b, a, expected);
 199 }
 200
 201 /* Verify levenshtein_distance for a variety of pairs of pre-canned
 202    inputs, comparing against known-good values.  */
 203
 204 void
 205 spellcheck_c_tests ()
 206 {
 207   levenshtein_distance_unit_test ("", "nonempty", strlen ("nonempty"));
 208   levenshtein_distance_unit_test ("saturday", "sunday", 3);
 209   levenshtein_distance_unit_test ("foo", "m_foo", 2);
 210   levenshtein_distance_unit_test ("hello_world", "HelloWorld", 3);
 211   levenshtein_distance_unit_test
 212     ("the quick brown fox jumps over the lazy dog", "dog", 40);
 213   levenshtein_distance_unit_test
 214     ("the quick brown fox jumps over the lazy dog",
 215      "the quick brown dog jumps over the lazy fox",
 216      4);
 217   levenshtein_distance_unit_test
 218     ("Lorem ipsum dolor sit amet, consectetur adipiscing elit,",
 219      "All your base are belong to us",
 220      44);
 221 }
 222
 223 } // namespace selftest
 224
 225 #endif /* #if CHECKING_P */