chrome/tools/convert_dict/dic_reader.cc

   1 // Copyright (c) 2006-2008 The Chromium Authors. All rights reserved.
   2 // Use of this source code is governed by a BSD-style license that can be
   3 // found in the LICENSE file.
   4
   5 #include "chrome/tools/convert_dict/dic_reader.h"
   6
   7 #include <algorithm>
   8 #include <set>
   9
  10 #include "base/files/file_util.h"
  11 #include "base/strings/string_util.h"
  12 #include "chrome/tools/convert_dict/aff_reader.h"
  13 #include "chrome/tools/convert_dict/hunspell_reader.h"
  14
  15 namespace convert_dict {
  16
  17 namespace {
  18
  19 // Maps each unique word to the unique affix group IDs associated with it.
  20 typedef std::map<std::string, std::set<int> > WordSet;
  21
  22 void SplitDicLine(const std::string& line, std::vector<std::string>* output) {
  23   // We split the line on a slash not preceded by a backslash. A slash at the
  24   // beginning of the line is not a separator either.
  25   size_t slash_index = line.size();
  26   for (size_t i = 0; i < line.size(); i++) {
  27     if (line[i] == '/' && i > 0 && line[i - 1] != '\\') {
  28       slash_index = i;
  29       break;
  30     }
  31   }
  32
  33   output->clear();
  34
  35   // Everything before the slash index is the first term. We also need to
  36   // convert all escaped slashes ("\/" sequences) to regular slashes.
  37   std::string word = line.substr(0, slash_index);
  38   ReplaceSubstringsAfterOffset(&word, 0, "\\/", "/");
  39   output->push_back(word);
  40
  41   // Everything (if anything) after the slash is the second.
  42   if (slash_index < line.size() - 1)
  43     output->push_back(line.substr(slash_index + 1));
  44 }
  45
  46 // This function reads words from a .dic file, or a .dic_delta file. Note that
  47 // we read 'all' the words in the file, irrespective of the word count given
  48 // in the first non empty line of a .dic file. Also note that, for a .dic_delta
  49 // file, the first line actually does _not_ have the number of words. In order
  50 // to control this, we use the |file_has_word_count_in_the_first_line|
  51 // parameter to tell this method whether the first non empty line in the file
  52 // contains the number of words or not. If it does, skip the first line. If it
  53 // does not, then the first line contains a word.
  54 bool PopulateWordSet(WordSet* word_set, FILE* file, AffReader* aff_reader,
  55                      const char* file_type, const char* encoding,
  56                      bool file_has_word_count_in_the_first_line) {
  57   int line_number = 0;
  58   while (!feof(file)) {
  59     std::string line = ReadLine(file);
  60     line_number++;
  61     StripComment(&line);
  62     if (line.empty())
  63       continue;
  64
  65     if (file_has_word_count_in_the_first_line) {
  66       // Skip the first nonempty line, this is the line count. We don't bother
  67       // with it and just read all the lines.
  68       file_has_word_count_in_the_first_line = false;
  69       continue;
  70     }
  71
  72     std::vector<std::string> split;
  73     SplitDicLine(line, &split);
  74     if (split.empty() || split.size() > 2) {
  75       printf("Line %d has extra slashes in the %s file\n", line_number,
  76              file_type);
  77       return false;
  78     }
  79
  80     // The first part is the word, the second (optional) part is the affix. We
  81     // always use UTF-8 as the encoding to simplify life.
  82     std::string utf8word;
  83     std::string encoding_string(encoding);
  84     if (encoding_string == "UTF-8") {
  85       utf8word = split[0];
  86     } else if (!aff_reader->EncodingToUTF8(split[0], &utf8word)) {
  87       printf("Unable to convert line %d from %s to UTF-8 in the %s file\n",
  88              line_number, encoding, file_type);
  89       return false;
  90     }
  91
  92     // We always convert the affix to an index. 0 means no affix.
  93     int affix_index = 0;
  94     if (split.size() == 2) {
  95       // Got a rule, which is the stuff after the slash. The line may also have
  96       // an optional term separated by a tab. This is the morphological
  97       // description. We don't care about this (it is used in the tests to
  98       // generate a nice dump), so we remove it.
  99       size_t split1_tab_offset = split[1].find('\t');
 100       if (split1_tab_offset != std::string::npos)
 101         split[1] = split[1].substr(0, split1_tab_offset);
 102
 103       if (aff_reader->has_indexed_affixes())
 104         affix_index = atoi(split[1].c_str());
 105       else
 106         affix_index = aff_reader->GetAFIndexForAFString(split[1]);
 107     }
 108
 109     // Discard the morphological description if it is attached to the first
 110     // token. (It is attached to the first token if a word doesn't have affix
 111     // rules.)
 112     size_t word_tab_offset = utf8word.find('\t');
 113     if (word_tab_offset != std::string::npos)
 114       utf8word = utf8word.substr(0, word_tab_offset);
 115
 116     WordSet::iterator found = word_set->find(utf8word);
 117     std::set<int> affix_vector;
 118     affix_vector.insert(affix_index);
 119
 120     if (found == word_set->end())
 121       word_set->insert(std::make_pair(utf8word, affix_vector));
 122     else
 123       found->second.insert(affix_index);
 124   }
 125
 126   return true;
 127 }
 128
 129 }  // namespace
 130
 131 DicReader::DicReader(const base::FilePath& path) {
 132   file_ = base::OpenFile(path, "r");
 133
 134   base::FilePath additional_path =
 135       path.ReplaceExtension(FILE_PATH_LITERAL("dic_delta"));
 136   additional_words_file_ = base::OpenFile(additional_path, "r");
 137
 138   if (additional_words_file_)
 139     printf("Reading %" PRFilePath " ...\n", additional_path.value().c_str());
 140   else
 141     printf("%" PRFilePath " not found.\n", additional_path.value().c_str());
 142 }
 143
 144 DicReader::~DicReader() {
 145   if (file_)
 146     base::CloseFile(file_);
 147   if (additional_words_file_)
 148     base::CloseFile(additional_words_file_);
 149 }
 150
 151 bool DicReader::Read(AffReader* aff_reader) {
 152   if (!file_)
 153     return false;
 154
 155   WordSet word_set;
 156
 157   // Add words from the dic file to the word set.
 158   // Note that the first line is the word count in the file.
 159   if (!PopulateWordSet(&word_set, file_, aff_reader, "dic",
 160                        aff_reader->encoding(), true))
 161     return false;
 162
 163   // Add words from the .dic_delta file to the word set, if it exists.
 164   // The first line is the first word to add. Word count line is not present.
 165   // NOTE: These additional words should be encoded as UTF-8.
 166   if (additional_words_file_ != NULL) {
 167     PopulateWordSet(&word_set, additional_words_file_, aff_reader, "dic delta",
 168                     "UTF-8", false);
 169   }
 170   // Make sure the words are sorted, they may be unsorted in the input.
 171   for (WordSet::iterator word = word_set.begin(); word != word_set.end();
 172        ++word) {
 173     std::vector<int> affixes;
 174     for (std::set<int>::iterator aff = word->second.begin();
 175          aff != word->second.end(); ++aff)
 176       affixes.push_back(*aff);
 177
 178     // Double check that the affixes are sorted. This isn't strictly necessary
 179     // but it's nice for the file to have a fixed layout.
 180     std::sort(affixes.begin(), affixes.end());
 181     std::reverse(affixes.begin(), affixes.end());
 182     words_.push_back(std::make_pair(word->first, affixes));
 183   }
 184
 185   // Double-check that the words are sorted.
 186   std::sort(words_.begin(), words_.end());
 187   return true;
 188 }
 189
 190 }  // namespace convert_dict