1 package com.hughes.android.dictionary.parser;
3 import java.io.BufferedReader;
5 import java.io.FileInputStream;
6 import java.io.IOException;
7 import java.io.InputStreamReader;
8 import java.nio.charset.Charset;
10 import java.util.logging.Logger;
11 import java.util.regex.Matcher;
12 import java.util.regex.Pattern;
14 import com.hughes.android.dictionary.engine.DictionaryBuilder;
15 import com.hughes.android.dictionary.engine.EntryData;
16 import com.hughes.android.dictionary.engine.EntryTypeName;
17 import com.hughes.android.dictionary.engine.IndexBuilder;
18 import com.hughes.android.dictionary.engine.Language;
19 import com.hughes.android.dictionary.engine.PairEntry;
20 import com.hughes.android.dictionary.engine.PairEntry.Pair;
22 public class DictFileParser {
24 static final Logger logger = Logger.getLogger(DictFileParser.class.getName());
27 public static final Pattern TAB = Pattern.compile("\\t");
30 public static final Pattern DOUBLE_COLON = Pattern.compile(" :: ");
31 public static final Pattern PIPE = Pattern.compile("\\|");
33 static final Pattern SPACES = Pattern.compile("\\s+");
34 static final Pattern DE_NOUN = Pattern.compile("([^ ]+) *\\{(m|f|n|pl)\\}");
35 static final Pattern EN_VERB = Pattern.compile("^to ([^ ]+)");
37 static final Pattern BRACKETED = Pattern.compile("\\[([^]]+)\\]");
38 static final Pattern PARENTHESIZED = Pattern.compile("\\(([^)]+)\\)");
39 static final Pattern CURLY_BRACED = Pattern.compile("\\{([^}]+)\\}");
41 static final Pattern NON_CHAR_DASH = Pattern.compile("[^-'\\p{L}0-9]+");
42 static final Pattern NON_CHAR = Pattern.compile("[^\\p{L}0-9]+");
44 static final Pattern TRIM_PUNC = Pattern.compile("^[^\\p{L}0-9]+|[^\\p{L}0-9]+$");
46 final Charset charset;
47 final boolean flipCols;
49 final Pattern fieldSplit;
50 final Pattern subfieldSplit;
52 final DictionaryBuilder dictBuilder;
53 final IndexBuilder[] langIndexBuilders;
54 final IndexBuilder bothIndexBuilder;
56 // final Set<String> alreadyDone = new HashSet<String>();
58 public DictFileParser(final Charset charset, boolean flipCols,
59 final Pattern fieldSplit, final Pattern subfieldSplit,
60 final DictionaryBuilder dictBuilder, final IndexBuilder[] langIndexBuilders,
61 final IndexBuilder bothIndexBuilder) {
62 this.charset = charset;
63 this.flipCols = flipCols;
64 this.fieldSplit = fieldSplit;
65 this.subfieldSplit = subfieldSplit;
66 this.dictBuilder = dictBuilder;
67 this.langIndexBuilders = langIndexBuilders;
68 this.bothIndexBuilder = bothIndexBuilder;
71 public void parseFile(final File file) throws IOException {
72 final BufferedReader reader = new BufferedReader(new InputStreamReader(new FileInputStream(file), charset));
75 while ((line = reader.readLine()) != null) {
76 if (count % 10000 == 0) {
77 logger.info("count=" + count + ", line=" + line);
84 private void parseLine(final String line) {
85 if (line.startsWith("#") || line.length() == 0) {
86 logger.info("Skipping comment line: " + line);
89 final String[] fields = fieldSplit.split(line);
90 if (fields.length != 2) {
91 logger.warning("Malformed line: " + line);
95 fields[0] = SPACES.matcher(fields[0]).replaceAll(" ").trim();
96 fields[1] = SPACES.matcher(fields[1]).replaceAll(" ").trim();
98 final String temp = fields[0];
99 fields[0] = fields[1];
103 final String[][] subfields = new String[2][];
104 if (subfieldSplit != null) {
105 subfields[0] = subfieldSplit.split(fields[0]);
106 subfields[1] = subfieldSplit.split(fields[1]);
107 if (subfields[0].length != subfields[1].length) {
108 logger.warning("Number of subfields doesn't match: " + line);
112 subfields[0] = new String[] { fields[0] };
113 subfields[1] = new String[] { fields[1] };
116 final Pair[] pairs = new Pair[subfields[0].length];
117 for (int i = 0; i < pairs.length; ++i) {
118 subfields[0][i] = subfields[0][i].trim();
119 subfields[1][i] = subfields[1][i].trim();
120 pairs[i] = new Pair(subfields[0][i], subfields[1][i]);
122 final PairEntry pairEntry = new PairEntry(pairs);
123 final EntryData entryData = new EntryData(dictBuilder.dictionary.pairEntries.size(), pairEntry);
124 dictBuilder.dictionary.pairEntries.add(pairEntry);
126 for (int l = 0; l < 2; ++l) {
127 // alreadyDone.clear();
129 for (int j = 0; j < subfields[l].length; ++j) {
130 String subfield = subfields[l][j];
131 final IndexBuilder indexBuilder = langIndexBuilders[l];
132 if (indexBuilder.index.sortLanguage == Language.de) {
133 subfield = parseField_DE(indexBuilder, subfield, entryData, j);
134 } else if (indexBuilder.index.sortLanguage == Language.en) {
135 subfield = parseField_EN(indexBuilder, subfield, entryData, j);
137 parseFieldGeneric(indexBuilder, subfield, entryData, j, subfields[l].length);
142 private void parseFieldGeneric(final IndexBuilder indexBuilder, String field,
143 final EntryData entryData, final int subfieldIdx, final int numSubFields) {
144 // remove bracketed and parenthesized stuff.
145 final StringBuilder bracketed = new StringBuilder();
146 final StringBuilder parenthesized = new StringBuilder();
149 while ((matcher = BRACKETED.matcher(field)).find()) {
150 bracketed.append(matcher.group(1)).append(" ");
151 field = matcher.replaceFirst(" ");
154 while ((matcher = PARENTHESIZED.matcher(field)).find()) {
155 parenthesized.append(matcher.group(1)).append(" ");
156 field = matcher.replaceFirst(" ");
159 field = SPACES.matcher(field).replaceAll(" ").trim();
161 // split words on non -A-z0-9, do them.
162 final String[] tokens = NON_CHAR_DASH.split(field);
164 final EntryTypeName entryTypeName;
165 if (numSubFields == 1) {
166 assert subfieldIdx == 0;
167 if (tokens.length == 1) {
168 entryTypeName = EntryTypeName.ONE_WORD;
169 } else if (tokens.length == 2) {
170 entryTypeName = EntryTypeName.TWO_WORDS;
171 } else if (tokens.length == 3) {
172 entryTypeName = EntryTypeName.THREE_WORDS;
173 } else if (tokens.length == 4) {
174 entryTypeName = EntryTypeName.FOUR_WORDS;
176 entryTypeName = EntryTypeName.FIVE_OR_MORE_WORDS;
179 assert numSubFields > 1;
180 if (subfieldIdx == 0) {
181 if (tokens.length == 1) {
182 entryTypeName = EntryTypeName.MULTIROW_HEAD_ONE_WORD;
184 entryTypeName = EntryTypeName.MULTIROW_HEAD_MANY_WORDS;
187 assert subfieldIdx > 0;
188 if (tokens.length == 1) {
189 entryTypeName = EntryTypeName.MULTIROW_TAIL_ONE_WORD;
191 entryTypeName = EntryTypeName.MULTIROW_TAIL_MANY_WORDS;
196 for (String token : tokens) {
197 token = TRIM_PUNC.matcher(token).replaceAll("");
198 if (/*!alreadyDone.contains(token) && */token.length() > 0) {
199 final List<EntryData> entries = indexBuilder.getOrCreateEntries(token, entryTypeName);
200 entries.add(entryData);
201 // alreadyDone.add(token);
203 // also split words on dashes, do them, too.
204 if (token.contains("-")) {
205 final String[] dashed = token.split("-");
206 for (final String dashedToken : dashed) {
207 if (/*!alreadyDone.contains(dashedToken) && */dashedToken.length() > 0) {
208 final List<EntryData> dashEntries = indexBuilder.getOrCreateEntries(dashedToken, EntryTypeName.PART_OF_HYPHENATED);
209 dashEntries.add(entryData);
214 } // if (!alreadyDone.contains(token)) {
215 } // for (final String token : tokens) {
217 // process bracketed stuff (split on spaces and dashes always)
218 final String[] bracketedTokens = NON_CHAR.split(bracketed.toString());
219 for (final String token : bracketedTokens) {
220 assert !token.contains("-");
221 if (/*!alreadyDone.contains(token) && */token.length() > 0) {
222 final List<EntryData> entries = indexBuilder.getOrCreateEntries(token, EntryTypeName.BRACKETED);
223 entries.add(entryData);
227 // process paren stuff
228 final String[] parenTokens = NON_CHAR.split(parenthesized.toString());
229 for (final String token : parenTokens) {
230 assert !token.contains("-");
231 if (/*!alreadyDone.contains(token) && */token.length() > 0) {
232 final List<EntryData> entries = indexBuilder.getOrCreateEntries(token, EntryTypeName.PARENTHESIZED);
233 entries.add(entryData);
239 private String parseField_DE(final IndexBuilder indexBuilder, String field,
240 final EntryData entryData, final int subfieldIdx) {
242 // final Matcher matcher = DE_NOUN.matcher(field);
243 // while (matcher.find()) {
244 // final String noun = matcher.group(1);
245 //final String gender = matcher.group(2);
246 // if (alreadyDone.add(noun)) {
247 // System.out.println("Found DE noun " + noun + ", " + gender);
248 // final List<EntryData> entries = indexBuilder.getOrCreateEntries(noun, EntryTypeName.NOUN);
249 // entries.add(entryData);
253 // In English, curly braces are used for different tenses.
254 field = CURLY_BRACED.matcher(field).replaceAll(" ");
259 private String parseField_EN(final IndexBuilder indexBuilder, String field,
260 final EntryData entryData, final int subfieldIdx) {
261 if (field.startsWith("to ")) {
262 field = field.substring(3);