1 // Copyright 2011 Google Inc. All Rights Reserved.
3 // Licensed under the Apache License, Version 2.0 (the "License");
4 // you may not use this file except in compliance with the License.
5 // You may obtain a copy of the License at
7 // http://www.apache.org/licenses/LICENSE-2.0
9 // Unless required by applicable law or agreed to in writing, software
10 // distributed under the License is distributed on an "AS IS" BASIS,
11 // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 // See the License for the specific language governing permissions and
13 // limitations under the License.
15 package com.hughes.android.dictionary.parser;
17 import java.io.BufferedReader;
19 import java.io.FileInputStream;
20 import java.io.IOException;
21 import java.io.InputStreamReader;
22 import java.nio.charset.Charset;
23 import java.util.Arrays;
24 import java.util.LinkedHashSet;
25 import java.util.List;
27 import java.util.logging.Logger;
28 import java.util.regex.Matcher;
29 import java.util.regex.Pattern;
31 import com.hughes.android.dictionary.engine.DictionaryBuilder;
32 import com.hughes.android.dictionary.engine.IndexedEntry;
33 import com.hughes.android.dictionary.engine.EntryTypeName;
34 import com.hughes.android.dictionary.engine.IndexBuilder;
35 import com.hughes.android.dictionary.engine.Language;
36 import com.hughes.android.dictionary.engine.PairEntry;
37 import com.hughes.android.dictionary.engine.PairEntry.Pair;
39 public class DictFileParser {
41 static final Logger logger = Logger.getLogger(DictFileParser.class.getName());
44 public static final Pattern TAB = Pattern.compile("\\t");
47 public static final Pattern DOUBLE_COLON = Pattern.compile(" :: ");
48 public static final Pattern PIPE = Pattern.compile("\\|");
50 static final Pattern SPACES = Pattern.compile("\\s+");
51 // static final Pattern DE_NOUN = Pattern.compile("([^ ]+) *\\{(m|f|n|pl)\\}");
52 // static final Pattern EN_VERB = Pattern.compile("^to ([^ ]+)");
54 static final Pattern BRACKETED = Pattern.compile("\\[([^]]+)\\]");
55 static final Pattern PARENTHESIZED = Pattern.compile("\\(([^)]+)\\)");
56 static final Pattern CURLY_BRACED = Pattern.compile("\\{([^}]+)\\}");
58 static final Pattern NON_CHAR_DASH = Pattern.compile("[^-'\\p{L}0-9]+");
59 public static final Pattern NON_CHAR = Pattern.compile("[^\\p{L}0-9]+");
61 static final Pattern TRIM_PUNC = Pattern.compile("^[^\\p{L}0-9]+|[^\\p{L}0-9]+$");
63 final Charset charset;
64 final boolean flipCols;
66 final Pattern fieldSplit;
67 final Pattern subfieldSplit;
69 final DictionaryBuilder dictBuilder;
70 final IndexBuilder[] langIndexBuilders;
71 final IndexBuilder bothIndexBuilder;
73 // final Set<String> alreadyDone = new HashSet<String>();
75 public DictFileParser(final Charset charset, boolean flipCols,
76 final Pattern fieldSplit, final Pattern subfieldSplit,
77 final DictionaryBuilder dictBuilder, final IndexBuilder[] langIndexBuilders,
78 final IndexBuilder bothIndexBuilder) {
79 this.charset = charset;
80 this.flipCols = flipCols;
81 this.fieldSplit = fieldSplit;
82 this.subfieldSplit = subfieldSplit;
83 this.dictBuilder = dictBuilder;
84 this.langIndexBuilders = langIndexBuilders;
85 this.bothIndexBuilder = bothIndexBuilder;
88 public void parseFile(final File file) throws IOException {
89 final BufferedReader reader = new BufferedReader(new InputStreamReader(new FileInputStream(file), charset));
92 while ((line = reader.readLine()) != null) {
93 if (count % 10000 == 0) {
94 logger.info("count=" + count + ", line=" + line);
101 private void parseLine(final String line) {
102 if (line.startsWith("#") || line.length() == 0) {
103 logger.info("Skipping comment line: " + line);
106 final String[] fields = fieldSplit.split(line);
107 if (fields.length != 2) {
108 logger.warning("Malformed line: " + line);
112 fields[0] = SPACES.matcher(fields[0]).replaceAll(" ").trim();
113 fields[1] = SPACES.matcher(fields[1]).replaceAll(" ").trim();
115 final String temp = fields[0];
116 fields[0] = fields[1];
120 final String[][] subfields = new String[2][];
121 if (subfieldSplit != null) {
122 subfields[0] = subfieldSplit.split(fields[0]);
123 subfields[1] = subfieldSplit.split(fields[1]);
124 if (subfields[0].length != subfields[1].length) {
125 logger.warning("Number of subfields doesn't match: " + line);
129 subfields[0] = new String[] { fields[0] };
130 subfields[1] = new String[] { fields[1] };
133 final PairEntry pairEntry = new PairEntry();
134 for (int i = 0; i < subfields[0].length; ++i) {
135 subfields[0][i] = subfields[0][i].trim();
136 subfields[1][i] = subfields[1][i].trim();
137 pairEntry.pairs.add(new Pair(subfields[0][i], subfields[1][i]));
139 final IndexedEntry entryData = new IndexedEntry(pairEntry);
141 for (int l = 0; l < 2; ++l) {
142 // alreadyDone.clear();
144 for (int j = 0; j < subfields[l].length; ++j) {
145 String subfield = subfields[l][j];
146 final IndexBuilder indexBuilder = langIndexBuilders[l];
147 if (indexBuilder.index.sortLanguage == Language.de) {
148 subfield = parseField_DE(indexBuilder, subfield, entryData, j);
149 } else if (indexBuilder.index.sortLanguage == Language.en) {
150 subfield = parseField_EN(indexBuilder, subfield, entryData, j);
152 parseFieldGeneric(indexBuilder, subfield, entryData, j, subfields[l].length);
157 private void parseFieldGeneric(final IndexBuilder indexBuilder, String field,
158 final IndexedEntry entryData, final int subfieldIdx, final int numSubFields) {
159 // remove bracketed and parenthesized stuff.
160 final StringBuilder bracketed = new StringBuilder();
161 final StringBuilder parenthesized = new StringBuilder();
164 while ((matcher = BRACKETED.matcher(field)).find()) {
165 bracketed.append(matcher.group(1)).append(" ");
166 field = matcher.replaceFirst(" ");
169 while ((matcher = PARENTHESIZED.matcher(field)).find()) {
170 parenthesized.append(matcher.group(1)).append(" ");
171 field = matcher.replaceFirst(" ");
174 field = SPACES.matcher(field).replaceAll(" ").trim();
176 // split words on non -A-z0-9, do them.
177 final String[] tokens = NON_CHAR_DASH.split(field);
179 final EntryTypeName entryTypeName;
180 if (numSubFields == 1) {
181 assert subfieldIdx == 0;
182 if (tokens.length == 1) {
183 entryTypeName = EntryTypeName.ONE_WORD;
184 } else if (tokens.length == 2) {
185 entryTypeName = EntryTypeName.TWO_WORDS;
186 } else if (tokens.length == 3) {
187 entryTypeName = EntryTypeName.THREE_WORDS;
188 } else if (tokens.length == 4) {
189 entryTypeName = EntryTypeName.FOUR_WORDS;
191 entryTypeName = EntryTypeName.FIVE_OR_MORE_WORDS;
194 assert numSubFields > 1;
195 if (subfieldIdx == 0) {
196 if (tokens.length == 1) {
197 entryTypeName = EntryTypeName.MULTIROW_HEAD_ONE_WORD;
199 entryTypeName = EntryTypeName.MULTIROW_HEAD_MANY_WORDS;
202 assert subfieldIdx > 0;
203 if (tokens.length == 1) {
204 entryTypeName = EntryTypeName.MULTIROW_TAIL_ONE_WORD;
206 entryTypeName = EntryTypeName.MULTIROW_TAIL_MANY_WORDS;
211 for (String token : tokens) {
212 token = TRIM_PUNC.matcher(token).replaceAll("");
213 if (/*!alreadyDone.contains(token) && */token.length() > 0) {
214 final List<IndexedEntry> entries = indexBuilder.getOrCreateEntries(token, entryTypeName);
215 entries.add(entryData);
216 // alreadyDone.add(token);
218 // also split words on dashes, do them, too.
219 if (token.contains("-")) {
220 final String[] dashed = token.split("-");
221 for (final String dashedToken : dashed) {
222 if (/*!alreadyDone.contains(dashedToken) && */dashedToken.length() > 0) {
223 final List<IndexedEntry> dashEntries = indexBuilder.getOrCreateEntries(dashedToken, EntryTypeName.PART_OF_HYPHENATED);
224 dashEntries.add(entryData);
229 } // if (!alreadyDone.contains(token)) {
230 } // for (final String token : tokens) {
232 // process bracketed stuff (split on spaces and dashes always)
233 final String[] bracketedTokens = NON_CHAR.split(bracketed.toString());
234 for (final String token : bracketedTokens) {
235 assert !token.contains("-");
236 if (/*!alreadyDone.contains(token) && */token.length() > 0) {
237 final List<IndexedEntry> entries = indexBuilder.getOrCreateEntries(token, EntryTypeName.BRACKETED);
238 entries.add(entryData);
242 // process paren stuff
243 final String[] parenTokens = NON_CHAR.split(parenthesized.toString());
244 for (final String token : parenTokens) {
245 assert !token.contains("-");
246 if (/*!alreadyDone.contains(token) && */token.length() > 0) {
247 final List<IndexedEntry> entries = indexBuilder.getOrCreateEntries(token, EntryTypeName.PARENTHESIZED);
248 entries.add(entryData);
254 private String parseField_DE(final IndexBuilder indexBuilder, String field,
255 final IndexedEntry entryData, final int subfieldIdx) {
257 // final Matcher matcher = DE_NOUN.matcher(field);
258 // while (matcher.find()) {
259 // final String noun = matcher.group(1);
260 //final String gender = matcher.group(2);
261 // if (alreadyDone.add(noun)) {
262 // System.out.println("Found DE noun " + noun + ", " + gender);
263 // final List<EntryData> entries = indexBuilder.getOrCreateEntries(noun, EntryTypeName.NOUN);
264 // entries.add(entryData);
268 // In English, curly braces are used for different tenses.
269 field = CURLY_BRACED.matcher(field).replaceAll(" ");
274 private String parseField_EN(final IndexBuilder indexBuilder, String field,
275 final IndexedEntry entryData, final int subfieldIdx) {
276 if (field.startsWith("to ")) {
277 field = field.substring(3);
282 public static final Set<String> tokenize(final String text, final Pattern pattern) {
283 final String[] split = pattern.split(text);
284 final Set<String> result = new LinkedHashSet<String>(Arrays.asList(split));