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;
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.EntrySource;
33 import com.hughes.android.dictionary.engine.EntryTypeName;
34 import com.hughes.android.dictionary.engine.IndexBuilder;
35 import com.hughes.android.dictionary.engine.IndexedEntry;
36 import com.hughes.android.dictionary.engine.Language;
37 import com.hughes.android.dictionary.engine.PairEntry;
38 import com.hughes.util.StringUtil;
40 public class DictFileParser implements Parser {
42 static final Logger logger = Logger.getLogger(DictFileParser.class.getName());
45 public static final String TAB = "\t";
48 public static final String DOUBLE_COLON = " :: ";
49 public static final String PIPE = "|";
51 static final Pattern SPACES = Pattern.compile("\\s+");
53 static final Pattern CURLY_BRACED = Pattern.compile("\\{([^}]+)\\}");
55 // http://www.regular-expressions.info/unicode.html
56 static final Pattern NON_CHAR_DASH = Pattern.compile("[^-'\\p{L}\\p{M}\\p{N}]+");
57 public static final Pattern NON_CHAR = Pattern.compile("[^\\p{L}\\p{M}\\p{N}]+");
59 static final Pattern TRIM_PUNC = Pattern.compile("^[^\\p{L}\\p{M}\\p{N}]+|[^\\p{L}\\p{M}\\p{N}]+$");
61 final Charset charset;
62 final boolean flipCols;
64 final String fieldSplit;
65 final String subfieldSplit;
67 final DictionaryBuilder dictBuilder;
69 EntrySource entrySource;
71 // final Set<String> alreadyDone = new HashSet<String>();
73 public DictFileParser(final Charset charset, boolean flipCols,
74 final String fieldSplit, final String subfieldSplit,
75 final DictionaryBuilder dictBuilder) {
76 this.charset = charset;
77 this.flipCols = flipCols;
78 this.fieldSplit = fieldSplit;
79 this.subfieldSplit = subfieldSplit;
80 this.dictBuilder = dictBuilder;
84 public void parse(final File file, final EntrySource entrySouce, final int pageLimit) throws IOException {
85 this.entrySource = entrySouce;
86 final BufferedReader reader = new BufferedReader(new InputStreamReader(new FileInputStream(file), charset));
89 while ((line = reader.readLine()) != null) {
90 if (pageLimit >= 0 && count >= pageLimit) {
93 if (count % 10000 == 0) {
94 logger.info("count=" + count + ", line=" + line);
101 private void parseLine(final String line) {
102 if (line.startsWith("#") || line.isEmpty()) {
103 logger.info("Skipping comment line: " + line);
106 final String[] fields = StringUtil.split(line, fieldSplit);
107 if (fields.length < 2 || fields.length > 4) {
108 logger.warning("Malformed line, expected 3 or 4 fields, got " + fields.length + ": " + 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] = StringUtil.split(fields[0], subfieldSplit);
123 subfields[1] = StringUtil.split(fields[1], subfieldSplit);
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(entrySource);
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 if (subfields[0][i].isEmpty() && subfields[1][i].isEmpty()) {
138 logger.warning("Empty pair: " + line);
141 if (subfields[0][i].isEmpty()) {
142 subfields[0][i] = "__";
144 if (subfields[1][i].isEmpty()) {
145 subfields[1][i] = "__";
147 pairEntry.pairs.add(new PairEntry.Pair(subfields[0][i], subfields[1][i]));
149 final IndexedEntry entryData = new IndexedEntry(pairEntry);
150 entryData.isValid = true;
152 for (int l = 0; l < 2; ++l) {
153 // alreadyDone.clear();
155 final IndexBuilder indexBuilder = dictBuilder.indexBuilders.get(l);
156 for (int j = 0; j < subfields[l].length; ++j) {
157 String subfield = subfields[l][j];
158 if (indexBuilder.index.sortLanguage == Language.de) {
159 subfield = parseField_DE(indexBuilder, subfield, entryData, j);
160 } else if (indexBuilder.index.sortLanguage == Language.en) {
161 subfield = parseField_EN(indexBuilder, subfield, entryData, j);
163 parseFieldGeneric(indexBuilder, subfield, entryData, j, subfields[l].length);
168 private StringBuilder extractParenthesized(StringBuilder in, String startChar, String endChar) {
169 StringBuilder res = new StringBuilder();
171 while ((pos = in.indexOf(startChar, pos)) != -1) {
172 int end = in.indexOf(endChar, pos + 1);
173 if (end == -1) break;
174 res.append(in, pos + 1, end).append(" ");
175 in.replace(pos, end + 1, " ");
176 pos++; // skip the just appended space
181 private void parseFieldGeneric(final IndexBuilder indexBuilder, String field,
182 final IndexedEntry entryData, final int subfieldIdx, final int numSubFields) {
183 final StringBuilder fieldsb = new StringBuilder(field);
184 // remove bracketed and parenthesized stuff.
185 final StringBuilder bracketed = extractParenthesized(fieldsb, "[", "]");
186 final StringBuilder parenthesized = extractParenthesized(fieldsb, "(", ")");
188 field = fieldsb.toString().trim();
190 // split words on non -A-z0-9, do them.
191 final String[] tokens = NON_CHAR_DASH.split(field);
193 final EntryTypeName entryTypeName;
194 if (numSubFields == 1) {
195 assert subfieldIdx == 0;
196 if (tokens.length == 1) {
197 entryTypeName = EntryTypeName.ONE_WORD;
198 } else if (tokens.length == 2) {
199 entryTypeName = EntryTypeName.TWO_WORDS;
200 } else if (tokens.length == 3) {
201 entryTypeName = EntryTypeName.THREE_WORDS;
202 } else if (tokens.length == 4) {
203 entryTypeName = EntryTypeName.FOUR_WORDS;
205 entryTypeName = EntryTypeName.FIVE_OR_MORE_WORDS;
208 assert numSubFields > 1;
209 if (subfieldIdx == 0) {
210 if (tokens.length == 1) {
211 entryTypeName = EntryTypeName.MULTIROW_HEAD_ONE_WORD;
213 entryTypeName = EntryTypeName.MULTIROW_HEAD_MANY_WORDS;
216 assert subfieldIdx > 0;
217 if (tokens.length == 1) {
218 entryTypeName = EntryTypeName.MULTIROW_TAIL_ONE_WORD;
220 entryTypeName = EntryTypeName.MULTIROW_TAIL_MANY_WORDS;
225 for (String token : tokens) {
226 token = TRIM_PUNC.matcher(token).replaceAll("");
227 if (/*!alreadyDone.contains(token) && */!token.isEmpty()) {
228 indexBuilder.addEntryWithTokens(entryData, Collections.singleton(token), entryTypeName);
229 // alreadyDone.add(token);
231 // also split words on dashes, do them, too.
232 if (token.indexOf('-') != -1) {
233 final String[] dashed = StringUtil.split(token, "-");
234 for (final String dashedToken : dashed) {
235 if (/*!alreadyDone.contains(dashedToken) && */!dashedToken.isEmpty()) {
236 indexBuilder.addEntryWithTokens(entryData, Collections.singleton(dashedToken), EntryTypeName.PART_OF_HYPHENATED);
241 } // if (!alreadyDone.contains(token)) {
242 } // for (final String token : tokens) {
244 // process bracketed stuff (split on spaces and dashes always)
245 if (bracketed.length() > 0) {
246 final String[] bracketedTokens = NON_CHAR.split(bracketed.toString());
247 for (final String token : bracketedTokens) {
248 assert token.indexOf("-") == -1;
249 if (/*!alreadyDone.contains(token) && */!token.isEmpty()) {
250 indexBuilder.addEntryWithTokens(entryData, Collections.singleton(token), EntryTypeName.BRACKETED);
255 // process paren stuff
256 if (parenthesized.length() > 0) {
257 final String[] parenTokens = NON_CHAR.split(parenthesized.toString());
258 for (final String token : parenTokens) {
259 assert token.indexOf("-") == -1;
260 if (/*!alreadyDone.contains(token) && */!token.isEmpty()) {
261 indexBuilder.addEntryWithTokens(entryData, Collections.singleton(token), EntryTypeName.PARENTHESIZED);
268 private String parseField_DE(final IndexBuilder indexBuilder, String field,
269 final IndexedEntry entryData, final int subfieldIdx) {
271 // final Matcher matcher = DE_NOUN.matcher(field);
272 // while (matcher.find()) {
273 // final String noun = matcher.group(1);
274 //final String gender = matcher.group(2);
275 // if (alreadyDone.add(noun)) {
276 // System.out.println("Found DE noun " + noun + ", " + gender);
277 // final List<EntryData> entries = indexBuilder.getOrCreateEntries(noun, EntryTypeName.NOUN);
278 // entries.add(entryData);
282 if (field.indexOf('{') == -1) return field;
284 // In English, curly braces are used for different tenses.
285 field = CURLY_BRACED.matcher(field).replaceAll(" ");
290 private String parseField_EN(final IndexBuilder indexBuilder, String field,
291 final IndexedEntry entryData, final int subfieldIdx) {
292 if (field.startsWith("to ")) {
293 field = field.substring(3);
298 public static Set<String> tokenize(final String text, final Pattern pattern) {
299 final String[] split = pattern.split(text);
300 final Set<String> result = new LinkedHashSet<>(Arrays.asList(split));