1 package com.hughes.android.dictionary.engine;
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.PairEntry.Pair;
16 public class DictFileParser {
18 static final Logger logger = Logger.getLogger(DictFileParser.class.getName());
21 static final Pattern TAB = Pattern.compile("\\t");
24 static final Pattern DOUBLE_COLON = Pattern.compile(" :: ");
25 static final Pattern PIPE = Pattern.compile("\\|");
27 static final Pattern SPACES = Pattern.compile("\\s+");
28 static final Pattern DE_NOUN = Pattern.compile("([^ ]+) *\\{(m|f|n|pl)\\}");
29 static final Pattern EN_VERB = Pattern.compile("^to ([^ ]+)");
31 static final Pattern BRACKETED = Pattern.compile("\\[([^]]+)\\]");
32 static final Pattern PARENTHESIZED = Pattern.compile("\\(([^)]+)\\)");
33 static final Pattern CURLY_BRACED = Pattern.compile("\\{([^}]+)\\}");
35 static final Pattern NON_CHAR_DASH = Pattern.compile("[^-'\\p{L}0-9]+");
36 static final Pattern NON_CHAR = Pattern.compile("[^\\p{L}0-9]+");
38 static final Pattern TRIM_PUNC = Pattern.compile("^[^\\p{L}0-9]+|[^\\p{L}0-9]+$");
40 final Charset charset;
41 final boolean flipCols;
43 final Pattern fieldSplit;
44 final Pattern subfieldSplit;
46 final DictionaryBuilder dictBuilder;
47 final IndexBuilder[] langIndexBuilders;
48 final IndexBuilder bothIndexBuilder;
50 // final Set<String> alreadyDone = new HashSet<String>();
52 public DictFileParser(final Charset charset, boolean flipCols,
53 final Pattern fieldSplit, final Pattern subfieldSplit,
54 final DictionaryBuilder dictBuilder, final IndexBuilder[] langIndexBuilders,
55 final IndexBuilder bothIndexBuilder) {
56 this.charset = charset;
57 this.flipCols = flipCols;
58 this.fieldSplit = fieldSplit;
59 this.subfieldSplit = subfieldSplit;
60 this.dictBuilder = dictBuilder;
61 this.langIndexBuilders = langIndexBuilders;
62 this.bothIndexBuilder = bothIndexBuilder;
65 public void parseFile(final File file) throws IOException {
66 final BufferedReader reader = new BufferedReader(new InputStreamReader(new FileInputStream(file), charset));
69 while ((line = reader.readLine()) != null) {
70 if (count % 10000 == 0) {
71 logger.info("count=" + count + ", line=" + line);
78 private void parseLine(final String line) {
79 if (line.startsWith("#") || line.length() == 0) {
80 logger.info("Skipping comment line: " + line);
83 final String[] fields = fieldSplit.split(line);
84 if (fields.length != 2) {
85 logger.warning("Malformed line: " + line);
89 fields[0] = SPACES.matcher(fields[0]).replaceAll(" ").trim();
90 fields[1] = SPACES.matcher(fields[1]).replaceAll(" ").trim();
92 final String temp = fields[0];
93 fields[0] = fields[1];
97 final String[][] subfields = new String[2][];
98 if (subfieldSplit != null) {
99 subfields[0] = subfieldSplit.split(fields[0]);
100 subfields[1] = subfieldSplit.split(fields[1]);
101 if (subfields[0].length != subfields[1].length) {
102 logger.warning("Number of subfields doesn't match: " + line);
106 subfields[0] = new String[] { fields[0] };
107 subfields[1] = new String[] { fields[1] };
110 final Pair[] pairs = new Pair[subfields[0].length];
111 for (int i = 0; i < pairs.length; ++i) {
112 subfields[0][i] = subfields[0][i].trim();
113 subfields[1][i] = subfields[1][i].trim();
114 pairs[i] = new Pair(subfields[0][i], subfields[1][i]);
116 final PairEntry pairEntry = new PairEntry(pairs);
117 final EntryData entryData = new EntryData(dictBuilder.dictionary.pairEntries.size(), pairEntry);
118 dictBuilder.dictionary.pairEntries.add(pairEntry);
120 for (int l = 0; l < 2; ++l) {
121 // alreadyDone.clear();
123 for (int j = 0; j < subfields[l].length; ++j) {
124 String subfield = subfields[l][j];
125 final IndexBuilder indexBuilder = langIndexBuilders[l];
126 if (indexBuilder.index.sortLanguage == Language.de) {
127 subfield = parseField_DE(indexBuilder, subfield, entryData, j);
128 } else if (indexBuilder.index.sortLanguage == Language.en) {
129 subfield = parseField_EN(indexBuilder, subfield, entryData, j);
131 parseFieldGeneric(indexBuilder, subfield, entryData, j, subfields[l].length);
136 private void parseFieldGeneric(final IndexBuilder indexBuilder, String field,
137 final EntryData entryData, final int subfieldIdx, final int numSubFields) {
138 // remove bracketed and parenthesized stuff.
139 final StringBuilder bracketed = new StringBuilder();
140 final StringBuilder parenthesized = new StringBuilder();
143 while ((matcher = BRACKETED.matcher(field)).find()) {
144 bracketed.append(matcher.group(1)).append(" ");
145 field = matcher.replaceFirst(" ");
148 while ((matcher = PARENTHESIZED.matcher(field)).find()) {
149 parenthesized.append(matcher.group(1)).append(" ");
150 field = matcher.replaceFirst(" ");
153 field = SPACES.matcher(field).replaceAll(" ").trim();
155 // split words on non -A-z0-9, do them.
156 final String[] tokens = NON_CHAR_DASH.split(field);
158 final EntryTypeName entryTypeName;
159 if (numSubFields == 1) {
160 assert subfieldIdx == 0;
161 if (tokens.length == 1) {
162 entryTypeName = EntryTypeName.ONE_WORD;
163 } else if (tokens.length == 2) {
164 entryTypeName = EntryTypeName.TWO_WORDS;
165 } else if (tokens.length == 3) {
166 entryTypeName = EntryTypeName.THREE_WORDS;
167 } else if (tokens.length == 4) {
168 entryTypeName = EntryTypeName.FOUR_WORDS;
170 entryTypeName = EntryTypeName.FIVE_OR_MORE_WORDS;
173 assert numSubFields > 1;
174 if (subfieldIdx == 0) {
175 if (tokens.length == 1) {
176 entryTypeName = EntryTypeName.MULTIROW_HEAD_ONE_WORD;
178 entryTypeName = EntryTypeName.MULTIROW_HEAD_MANY_WORDS;
181 assert subfieldIdx > 0;
182 if (tokens.length == 1) {
183 entryTypeName = EntryTypeName.MULTIROW_TAIL_ONE_WORD;
185 entryTypeName = EntryTypeName.MULTIROW_TAIL_MANY_WORDS;
190 for (String token : tokens) {
191 token = TRIM_PUNC.matcher(token).replaceAll("");
192 if (/*!alreadyDone.contains(token) && */token.length() > 0) {
193 final List<EntryData> entries = indexBuilder.getOrCreateEntries(token, entryTypeName);
194 entries.add(entryData);
195 // alreadyDone.add(token);
197 // also split words on dashes, do them, too.
198 if (token.contains("-")) {
199 final String[] dashed = token.split("-");
200 for (final String dashedToken : dashed) {
201 if (/*!alreadyDone.contains(dashedToken) && */dashedToken.length() > 0) {
202 final List<EntryData> dashEntries = indexBuilder.getOrCreateEntries(dashedToken, EntryTypeName.PART_OF_HYPHENATED);
203 dashEntries.add(entryData);
208 } // if (!alreadyDone.contains(token)) {
209 } // for (final String token : tokens) {
211 // process bracketed stuff (split on spaces and dashes always)
212 final String[] bracketedTokens = NON_CHAR.split(bracketed.toString());
213 for (final String token : bracketedTokens) {
214 assert !token.contains("-");
215 if (/*!alreadyDone.contains(token) && */token.length() > 0) {
216 final List<EntryData> entries = indexBuilder.getOrCreateEntries(token, EntryTypeName.BRACKETED);
217 entries.add(entryData);
221 // process paren stuff
222 final String[] parenTokens = NON_CHAR.split(parenthesized.toString());
223 for (final String token : parenTokens) {
224 assert !token.contains("-");
225 if (/*!alreadyDone.contains(token) && */token.length() > 0) {
226 final List<EntryData> entries = indexBuilder.getOrCreateEntries(token, EntryTypeName.PARENTHESIZED);
227 entries.add(entryData);
233 private String parseField_DE(final IndexBuilder indexBuilder, String field,
234 final EntryData entryData, final int subfieldIdx) {
236 // final Matcher matcher = DE_NOUN.matcher(field);
237 // while (matcher.find()) {
238 // final String noun = matcher.group(1);
239 //final String gender = matcher.group(2);
240 // if (alreadyDone.add(noun)) {
241 // System.out.println("Found DE noun " + noun + ", " + gender);
242 // final List<EntryData> entries = indexBuilder.getOrCreateEntries(noun, EntryTypeName.NOUN);
243 // entries.add(entryData);
247 // In English, curly braces are used for different tenses.
248 field = CURLY_BRACED.matcher(field).replaceAll(" ");
253 private String parseField_EN(final IndexBuilder indexBuilder, String field,
254 final EntryData entryData, final int subfieldIdx) {
255 if (field.startsWith("to ")) {
256 field = field.substring(3);