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.Collections;
25 import java.util.LinkedHashSet;
26 import java.util.List;
28 import java.util.logging.Logger;
29 import java.util.regex.Matcher;
30 import java.util.regex.Pattern;
32 import com.hughes.android.dictionary.engine.DictionaryBuilder;
33 import com.hughes.android.dictionary.engine.EntrySource;
34 import com.hughes.android.dictionary.engine.IndexedEntry;
35 import com.hughes.android.dictionary.engine.EntryTypeName;
36 import com.hughes.android.dictionary.engine.IndexBuilder;
37 import com.hughes.android.dictionary.engine.Language;
38 import com.hughes.android.dictionary.engine.PairEntry;
39 import com.hughes.android.dictionary.engine.PairEntry.Pair;
41 public class DictFileParser {
43 static final Logger logger = Logger.getLogger(DictFileParser.class.getName());
46 public static final Pattern TAB = Pattern.compile("\\t");
49 public static final Pattern DOUBLE_COLON = Pattern.compile(" :: ");
50 public static final Pattern PIPE = Pattern.compile("\\|");
52 static final Pattern SPACES = Pattern.compile("\\s+");
53 // static final Pattern DE_NOUN = Pattern.compile("([^ ]+) *\\{(m|f|n|pl)\\}");
54 // static final Pattern EN_VERB = Pattern.compile("^to ([^ ]+)");
56 static final Pattern BRACKETED = Pattern.compile("\\[([^]]+)\\]");
57 static final Pattern PARENTHESIZED = Pattern.compile("\\(([^)]+)\\)");
58 static final Pattern CURLY_BRACED = Pattern.compile("\\{([^}]+)\\}");
60 static final Pattern NON_CHAR_DASH = Pattern.compile("[^-'\\p{L}0-9]+");
61 public static final Pattern NON_CHAR = Pattern.compile("[^\\p{L}0-9]+");
63 static final Pattern TRIM_PUNC = Pattern.compile("^[^\\p{L}0-9]+|[^\\p{L}0-9]+$");
65 final Charset charset;
66 final boolean flipCols;
68 final Pattern fieldSplit;
69 final Pattern subfieldSplit;
71 final DictionaryBuilder dictBuilder;
72 final IndexBuilder[] langIndexBuilders;
73 final IndexBuilder bothIndexBuilder;
75 EntrySource entrySource;
77 // final Set<String> alreadyDone = new HashSet<String>();
79 public DictFileParser(final Charset charset, boolean flipCols,
80 final Pattern fieldSplit, final Pattern subfieldSplit,
81 final DictionaryBuilder dictBuilder, final IndexBuilder[] langIndexBuilders,
82 final IndexBuilder bothIndexBuilder) {
83 this.charset = charset;
84 this.flipCols = flipCols;
85 this.fieldSplit = fieldSplit;
86 this.subfieldSplit = subfieldSplit;
87 this.dictBuilder = dictBuilder;
88 this.langIndexBuilders = langIndexBuilders;
89 this.bothIndexBuilder = bothIndexBuilder;
92 public void parseFile(final File file, final EntrySource entrySouce) throws IOException {
93 this.entrySource = entrySouce;
94 final BufferedReader reader = new BufferedReader(new InputStreamReader(new FileInputStream(file), charset));
97 while ((line = reader.readLine()) != null) {
98 if (count % 10000 == 0) {
99 logger.info("count=" + count + ", line=" + line);
106 private void parseLine(final String line) {
107 if (line.startsWith("#") || line.length() == 0) {
108 logger.info("Skipping comment line: " + line);
111 final String[] fields = fieldSplit.split(line);
112 if (fields.length != 2) {
113 logger.warning("Malformed line: " + line);
117 fields[0] = SPACES.matcher(fields[0]).replaceAll(" ").trim();
118 fields[1] = SPACES.matcher(fields[1]).replaceAll(" ").trim();
120 final String temp = fields[0];
121 fields[0] = fields[1];
125 final String[][] subfields = new String[2][];
126 if (subfieldSplit != null) {
127 subfields[0] = subfieldSplit.split(fields[0]);
128 subfields[1] = subfieldSplit.split(fields[1]);
129 if (subfields[0].length != subfields[1].length) {
130 logger.warning("Number of subfields doesn't match: " + line);
134 subfields[0] = new String[] { fields[0] };
135 subfields[1] = new String[] { fields[1] };
138 final PairEntry pairEntry = new PairEntry(entrySource);
139 for (int i = 0; i < subfields[0].length; ++i) {
140 subfields[0][i] = subfields[0][i].trim();
141 subfields[1][i] = subfields[1][i].trim();
142 if (subfields[0][i].length() == 0 && subfields[1][i].length() == 0) {
143 logger.warning("Empty pair: " + line);
146 if (subfields[0][i].length() == 0) {
147 subfields[0][i] = "__";
149 if (subfields[1][i].length() == 0) {
150 subfields[1][i] = "__";
152 pairEntry.pairs.add(new Pair(subfields[0][i], subfields[1][i]));
154 final IndexedEntry entryData = new IndexedEntry(pairEntry);
156 for (int l = 0; l < 2; ++l) {
157 // alreadyDone.clear();
159 for (int j = 0; j < subfields[l].length; ++j) {
160 String subfield = subfields[l][j];
161 final IndexBuilder indexBuilder = langIndexBuilders[l];
162 if (indexBuilder.index.sortLanguage == Language.de) {
163 subfield = parseField_DE(indexBuilder, subfield, entryData, j);
164 } else if (indexBuilder.index.sortLanguage == Language.en) {
165 subfield = parseField_EN(indexBuilder, subfield, entryData, j);
167 parseFieldGeneric(indexBuilder, subfield, entryData, j, subfields[l].length);
172 private void parseFieldGeneric(final IndexBuilder indexBuilder, String field,
173 final IndexedEntry entryData, final int subfieldIdx, final int numSubFields) {
174 // remove bracketed and parenthesized stuff.
175 final StringBuilder bracketed = new StringBuilder();
176 final StringBuilder parenthesized = new StringBuilder();
179 while ((matcher = BRACKETED.matcher(field)).find()) {
180 bracketed.append(matcher.group(1)).append(" ");
181 field = matcher.replaceFirst(" ");
184 while ((matcher = PARENTHESIZED.matcher(field)).find()) {
185 parenthesized.append(matcher.group(1)).append(" ");
186 field = matcher.replaceFirst(" ");
189 field = SPACES.matcher(field).replaceAll(" ").trim();
191 // split words on non -A-z0-9, do them.
192 final String[] tokens = NON_CHAR_DASH.split(field);
194 final EntryTypeName entryTypeName;
195 if (numSubFields == 1) {
196 assert subfieldIdx == 0;
197 if (tokens.length == 1) {
198 entryTypeName = EntryTypeName.ONE_WORD;
199 } else if (tokens.length == 2) {
200 entryTypeName = EntryTypeName.TWO_WORDS;
201 } else if (tokens.length == 3) {
202 entryTypeName = EntryTypeName.THREE_WORDS;
203 } else if (tokens.length == 4) {
204 entryTypeName = EntryTypeName.FOUR_WORDS;
206 entryTypeName = EntryTypeName.FIVE_OR_MORE_WORDS;
209 assert numSubFields > 1;
210 if (subfieldIdx == 0) {
211 if (tokens.length == 1) {
212 entryTypeName = EntryTypeName.MULTIROW_HEAD_ONE_WORD;
214 entryTypeName = EntryTypeName.MULTIROW_HEAD_MANY_WORDS;
217 assert subfieldIdx > 0;
218 if (tokens.length == 1) {
219 entryTypeName = EntryTypeName.MULTIROW_TAIL_ONE_WORD;
221 entryTypeName = EntryTypeName.MULTIROW_TAIL_MANY_WORDS;
226 for (String token : tokens) {
227 token = TRIM_PUNC.matcher(token).replaceAll("");
228 if (/*!alreadyDone.contains(token) && */token.length() > 0) {
229 indexBuilder.addEntryWithTokens(entryData, Collections.singleton(token), entryTypeName);
230 // alreadyDone.add(token);
232 // also split words on dashes, do them, too.
233 if (token.contains("-")) {
234 final String[] dashed = token.split("-");
235 for (final String dashedToken : dashed) {
236 if (/*!alreadyDone.contains(dashedToken) && */dashedToken.length() > 0) {
237 indexBuilder.addEntryWithTokens(entryData, Collections.singleton(dashedToken), EntryTypeName.PART_OF_HYPHENATED);
242 } // if (!alreadyDone.contains(token)) {
243 } // for (final String token : tokens) {
245 // process bracketed stuff (split on spaces and dashes always)
246 final String[] bracketedTokens = NON_CHAR.split(bracketed.toString());
247 for (final String token : bracketedTokens) {
248 assert !token.contains("-");
249 if (/*!alreadyDone.contains(token) && */token.length() > 0) {
250 indexBuilder.addEntryWithTokens(entryData, Collections.singleton(token), EntryTypeName.BRACKETED);
254 // process paren stuff
255 final String[] parenTokens = NON_CHAR.split(parenthesized.toString());
256 for (final String token : parenTokens) {
257 assert !token.contains("-");
258 if (/*!alreadyDone.contains(token) && */token.length() > 0) {
259 indexBuilder.addEntryWithTokens(entryData, Collections.singleton(token), EntryTypeName.PARENTHESIZED);
265 private String parseField_DE(final IndexBuilder indexBuilder, String field,
266 final IndexedEntry entryData, final int subfieldIdx) {
268 // final Matcher matcher = DE_NOUN.matcher(field);
269 // while (matcher.find()) {
270 // final String noun = matcher.group(1);
271 //final String gender = matcher.group(2);
272 // if (alreadyDone.add(noun)) {
273 // System.out.println("Found DE noun " + noun + ", " + gender);
274 // final List<EntryData> entries = indexBuilder.getOrCreateEntries(noun, EntryTypeName.NOUN);
275 // entries.add(entryData);
279 // In English, curly braces are used for different tenses.
280 field = CURLY_BRACED.matcher(field).replaceAll(" ");
285 private String parseField_EN(final IndexBuilder indexBuilder, String field,
286 final IndexedEntry entryData, final int subfieldIdx) {
287 if (field.startsWith("to ")) {
288 field = field.substring(3);
293 public static final Set<String> tokenize(final String text, final Pattern pattern) {
294 final String[] split = pattern.split(text);
295 final Set<String> result = new LinkedHashSet<String>(Arrays.asList(split));