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;
9 import java.util.Arrays;
10 import java.util.LinkedHashSet;
11 import java.util.List;
13 import java.util.logging.Logger;
14 import java.util.regex.Matcher;
15 import java.util.regex.Pattern;
17 import com.hughes.android.dictionary.engine.DictionaryBuilder;
18 import com.hughes.android.dictionary.engine.IndexedEntry;
19 import com.hughes.android.dictionary.engine.EntryTypeName;
20 import com.hughes.android.dictionary.engine.IndexBuilder;
21 import com.hughes.android.dictionary.engine.Language;
22 import com.hughes.android.dictionary.engine.PairEntry;
23 import com.hughes.android.dictionary.engine.PairEntry.Pair;
25 public class DictFileParser {
27 static final Logger logger = Logger.getLogger(DictFileParser.class.getName());
30 public static final Pattern TAB = Pattern.compile("\\t");
33 public static final Pattern DOUBLE_COLON = Pattern.compile(" :: ");
34 public static final Pattern PIPE = Pattern.compile("\\|");
36 static final Pattern SPACES = Pattern.compile("\\s+");
37 // static final Pattern DE_NOUN = Pattern.compile("([^ ]+) *\\{(m|f|n|pl)\\}");
38 // static final Pattern EN_VERB = Pattern.compile("^to ([^ ]+)");
40 static final Pattern BRACKETED = Pattern.compile("\\[([^]]+)\\]");
41 static final Pattern PARENTHESIZED = Pattern.compile("\\(([^)]+)\\)");
42 static final Pattern CURLY_BRACED = Pattern.compile("\\{([^}]+)\\}");
44 static final Pattern NON_CHAR_DASH = Pattern.compile("[^-'\\p{L}0-9]+");
45 static final Pattern NON_CHAR = Pattern.compile("[^\\p{L}0-9]+");
47 static final Pattern TRIM_PUNC = Pattern.compile("^[^\\p{L}0-9]+|[^\\p{L}0-9]+$");
49 final Charset charset;
50 final boolean flipCols;
52 final Pattern fieldSplit;
53 final Pattern subfieldSplit;
55 final DictionaryBuilder dictBuilder;
56 final IndexBuilder[] langIndexBuilders;
57 final IndexBuilder bothIndexBuilder;
59 // final Set<String> alreadyDone = new HashSet<String>();
61 public DictFileParser(final Charset charset, boolean flipCols,
62 final Pattern fieldSplit, final Pattern subfieldSplit,
63 final DictionaryBuilder dictBuilder, final IndexBuilder[] langIndexBuilders,
64 final IndexBuilder bothIndexBuilder) {
65 this.charset = charset;
66 this.flipCols = flipCols;
67 this.fieldSplit = fieldSplit;
68 this.subfieldSplit = subfieldSplit;
69 this.dictBuilder = dictBuilder;
70 this.langIndexBuilders = langIndexBuilders;
71 this.bothIndexBuilder = bothIndexBuilder;
74 public void parseFile(final File file) throws IOException {
75 final BufferedReader reader = new BufferedReader(new InputStreamReader(new FileInputStream(file), charset));
78 while ((line = reader.readLine()) != null) {
79 if (count % 10000 == 0) {
80 logger.info("count=" + count + ", line=" + line);
87 private void parseLine(final String line) {
88 if (line.startsWith("#") || line.length() == 0) {
89 logger.info("Skipping comment line: " + line);
92 final String[] fields = fieldSplit.split(line);
93 if (fields.length != 2) {
94 logger.warning("Malformed line: " + line);
98 fields[0] = SPACES.matcher(fields[0]).replaceAll(" ").trim();
99 fields[1] = SPACES.matcher(fields[1]).replaceAll(" ").trim();
101 final String temp = fields[0];
102 fields[0] = fields[1];
106 final String[][] subfields = new String[2][];
107 if (subfieldSplit != null) {
108 subfields[0] = subfieldSplit.split(fields[0]);
109 subfields[1] = subfieldSplit.split(fields[1]);
110 if (subfields[0].length != subfields[1].length) {
111 logger.warning("Number of subfields doesn't match: " + line);
115 subfields[0] = new String[] { fields[0] };
116 subfields[1] = new String[] { fields[1] };
119 final PairEntry pairEntry = new PairEntry();
120 for (int i = 0; i < subfields[0].length; ++i) {
121 subfields[0][i] = subfields[0][i].trim();
122 subfields[1][i] = subfields[1][i].trim();
123 pairEntry.pairs.add(new Pair(subfields[0][i], subfields[1][i]));
125 final IndexedEntry entryData = new IndexedEntry(pairEntry);
127 for (int l = 0; l < 2; ++l) {
128 // alreadyDone.clear();
130 for (int j = 0; j < subfields[l].length; ++j) {
131 String subfield = subfields[l][j];
132 final IndexBuilder indexBuilder = langIndexBuilders[l];
133 if (indexBuilder.index.sortLanguage == Language.de) {
134 subfield = parseField_DE(indexBuilder, subfield, entryData, j);
135 } else if (indexBuilder.index.sortLanguage == Language.en) {
136 subfield = parseField_EN(indexBuilder, subfield, entryData, j);
138 parseFieldGeneric(indexBuilder, subfield, entryData, j, subfields[l].length);
143 private void parseFieldGeneric(final IndexBuilder indexBuilder, String field,
144 final IndexedEntry entryData, final int subfieldIdx, final int numSubFields) {
145 // remove bracketed and parenthesized stuff.
146 final StringBuilder bracketed = new StringBuilder();
147 final StringBuilder parenthesized = new StringBuilder();
150 while ((matcher = BRACKETED.matcher(field)).find()) {
151 bracketed.append(matcher.group(1)).append(" ");
152 field = matcher.replaceFirst(" ");
155 while ((matcher = PARENTHESIZED.matcher(field)).find()) {
156 parenthesized.append(matcher.group(1)).append(" ");
157 field = matcher.replaceFirst(" ");
160 field = SPACES.matcher(field).replaceAll(" ").trim();
162 // split words on non -A-z0-9, do them.
163 final String[] tokens = NON_CHAR_DASH.split(field);
165 final EntryTypeName entryTypeName;
166 if (numSubFields == 1) {
167 assert subfieldIdx == 0;
168 if (tokens.length == 1) {
169 entryTypeName = EntryTypeName.ONE_WORD;
170 } else if (tokens.length == 2) {
171 entryTypeName = EntryTypeName.TWO_WORDS;
172 } else if (tokens.length == 3) {
173 entryTypeName = EntryTypeName.THREE_WORDS;
174 } else if (tokens.length == 4) {
175 entryTypeName = EntryTypeName.FOUR_WORDS;
177 entryTypeName = EntryTypeName.FIVE_OR_MORE_WORDS;
180 assert numSubFields > 1;
181 if (subfieldIdx == 0) {
182 if (tokens.length == 1) {
183 entryTypeName = EntryTypeName.MULTIROW_HEAD_ONE_WORD;
185 entryTypeName = EntryTypeName.MULTIROW_HEAD_MANY_WORDS;
188 assert subfieldIdx > 0;
189 if (tokens.length == 1) {
190 entryTypeName = EntryTypeName.MULTIROW_TAIL_ONE_WORD;
192 entryTypeName = EntryTypeName.MULTIROW_TAIL_MANY_WORDS;
197 for (String token : tokens) {
198 token = TRIM_PUNC.matcher(token).replaceAll("");
199 if (/*!alreadyDone.contains(token) && */token.length() > 0) {
200 final List<IndexedEntry> entries = indexBuilder.getOrCreateEntries(token, entryTypeName);
201 entries.add(entryData);
202 // alreadyDone.add(token);
204 // also split words on dashes, do them, too.
205 if (token.contains("-")) {
206 final String[] dashed = token.split("-");
207 for (final String dashedToken : dashed) {
208 if (/*!alreadyDone.contains(dashedToken) && */dashedToken.length() > 0) {
209 final List<IndexedEntry> dashEntries = indexBuilder.getOrCreateEntries(dashedToken, EntryTypeName.PART_OF_HYPHENATED);
210 dashEntries.add(entryData);
215 } // if (!alreadyDone.contains(token)) {
216 } // for (final String token : tokens) {
218 // process bracketed stuff (split on spaces and dashes always)
219 final String[] bracketedTokens = NON_CHAR.split(bracketed.toString());
220 for (final String token : bracketedTokens) {
221 assert !token.contains("-");
222 if (/*!alreadyDone.contains(token) && */token.length() > 0) {
223 final List<IndexedEntry> entries = indexBuilder.getOrCreateEntries(token, EntryTypeName.BRACKETED);
224 entries.add(entryData);
228 // process paren stuff
229 final String[] parenTokens = NON_CHAR.split(parenthesized.toString());
230 for (final String token : parenTokens) {
231 assert !token.contains("-");
232 if (/*!alreadyDone.contains(token) && */token.length() > 0) {
233 final List<IndexedEntry> entries = indexBuilder.getOrCreateEntries(token, EntryTypeName.PARENTHESIZED);
234 entries.add(entryData);
240 private String parseField_DE(final IndexBuilder indexBuilder, String field,
241 final IndexedEntry entryData, final int subfieldIdx) {
243 // final Matcher matcher = DE_NOUN.matcher(field);
244 // while (matcher.find()) {
245 // final String noun = matcher.group(1);
246 //final String gender = matcher.group(2);
247 // if (alreadyDone.add(noun)) {
248 // System.out.println("Found DE noun " + noun + ", " + gender);
249 // final List<EntryData> entries = indexBuilder.getOrCreateEntries(noun, EntryTypeName.NOUN);
250 // entries.add(entryData);
254 // In English, curly braces are used for different tenses.
255 field = CURLY_BRACED.matcher(field).replaceAll(" ");
260 private String parseField_EN(final IndexBuilder indexBuilder, String field,
261 final IndexedEntry entryData, final int subfieldIdx) {
262 if (field.startsWith("to ")) {
263 field = field.substring(3);
268 public static final Set<String> tokenize(final String text, final Pattern pattern) {
269 final String[] split = pattern.split(text);
270 final Set<String> result = new LinkedHashSet<String>(Arrays.asList(split));