diff options
author | Nico Weber <nicolasweber@gmx.de> | 2015-09-07 21:26:19 (GMT) |
---|---|---|
committer | Nico Weber <nicolasweber@gmx.de> | 2015-09-07 21:26:19 (GMT) |
commit | fdb885dec781424d1317671db3fac74e6daf5347 (patch) | |
tree | 9e4b175d0b472fe24337cb07da3f6c43c50f5ffd /src/edit_distance.cc | |
parent | 2810ffb2b5ff92b38396cc711775801abb3ce634 (diff) | |
download | Ninja-fdb885dec781424d1317671db3fac74e6daf5347.zip Ninja-fdb885dec781424d1317671db3fac74e6daf5347.tar.gz Ninja-fdb885dec781424d1317671db3fac74e6daf5347.tar.bz2 |
Merge LLVM r242069, make spell checking allocate less memory.
Diffstat (limited to 'src/edit_distance.cc')
-rw-r--r-- | src/edit_distance.cc | 38 |
1 files changed, 20 insertions, 18 deletions
diff --git a/src/edit_distance.cc b/src/edit_distance.cc index a6719d3..3bb62b8 100644 --- a/src/edit_distance.cc +++ b/src/edit_distance.cc @@ -28,40 +28,42 @@ int EditDistance(const StringPiece& s1, // http://en.wikipedia.org/wiki/Levenshtein_distance // // Although the algorithm is typically described using an m x n - // array, only two rows are used at a time, so this implementation - // just keeps two separate vectors for those two rows. + // array, only one row plus one element are used at a time, so this + // implementation just keeps one vector for the row. To update one entry, + // only the entries to the left, top, and top-left are needed. The left + // entry is in row[x-1], the top entry is what's in row[x] from the last + // iteration, and the top-left entry is stored in previous. int m = s1.len_; int n = s2.len_; - vector<int> previous(n + 1); - vector<int> current(n + 1); - - for (int i = 0; i <= n; ++i) - previous[i] = i; + vector<int> row(n + 1); + for (int i = 1; i <= n; ++i) + row[i] = i; for (int y = 1; y <= m; ++y) { - current[0] = y; - int best_this_row = current[0]; + row[0] = y; + int best_this_row = row[0]; + int previous = y - 1; for (int x = 1; x <= n; ++x) { + int old_row = row[x]; if (allow_replacements) { - current[x] = min(previous[x-1] + (s1.str_[y-1] == s2.str_[x-1] ? 0 : 1), - min(current[x-1], previous[x])+1); + row[x] = min(previous + (s1.str_[y - 1] == s2.str_[x - 1] ? 0 : 1), + min(row[x - 1], row[x]) + 1); } else { - if (s1.str_[y-1] == s2.str_[x-1]) - current[x] = previous[x-1]; + if (s1.str_[y - 1] == s2.str_[x - 1]) + row[x] = previous; else - current[x] = min(current[x-1], previous[x]) + 1; + row[x] = min(row[x - 1], row[x]) + 1; } - best_this_row = min(best_this_row, current[x]); + previous = old_row; + best_this_row = min(best_this_row, row[x]); } if (max_edit_distance && best_this_row > max_edit_distance) return max_edit_distance + 1; - - current.swap(previous); } - return previous[n]; + return row[n]; } |