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;
9 import java.util.HashSet;
10 import java.util.List;
12 import java.util.logging.Logger;
13 import java.util.regex.Matcher;
14 import java.util.regex.Pattern;
16 import com.hughes.android.dictionary.Language;
17 import com.hughes.android.dictionary.engine.PairEntry.Pair;
19 public class DictFileParser {
21 static final Logger logger = Logger.getLogger(DictFileParser.class.getName());
24 static final Pattern TAB = Pattern.compile("\\t");
27 static final Pattern DOUBLE_COLON = Pattern.compile(" :: ");
28 static final Pattern PIPE = Pattern.compile(" \\| ");
30 static final Pattern SPACES = Pattern.compile("\\s+");
31 static final Pattern DE_NOUN = Pattern.compile("([^ ]+) *\\{(m|f|n|pl)\\}");
32 static final Pattern EN_VERB = Pattern.compile("^to ([^ ]+)");
34 static final Pattern BRACKETED = Pattern.compile("\\[([^]]+)\\]");
35 static final Pattern PARENTHESIZED = Pattern.compile("\\(([^)]+)\\]");
37 static final Pattern NON_CHAR_DASH = Pattern.compile("[^-'\\p{L}0-9]+");
38 static final Pattern NON_CHAR = Pattern.compile("[^\\p{L}0-9]+");
40 static final Pattern TRIM_PUNC = Pattern.compile("^[^\\p{L}0-9]+|[^\\p{L}0-9]+$");
42 final Charset charset;
43 final boolean flipCols;
45 final Pattern fieldSplit;
46 final Pattern subfieldSplit;
48 final DictionaryBuilder dictBuilder;
49 final IndexBuilder[] langIndexBuilders;
50 final IndexBuilder bothIndexBuilder;
52 final Set<String> alreadyDone = new HashSet<String>();
54 public DictFileParser(final Charset charset, boolean flipCols,
55 final Pattern fieldSplit, final Pattern subfieldSplit,
56 final DictionaryBuilder dictBuilder, final IndexBuilder[] langIndexBuilders,
57 final IndexBuilder bothIndexBuilder) {
58 this.charset = charset;
59 this.flipCols = flipCols;
60 this.fieldSplit = fieldSplit;
61 this.subfieldSplit = subfieldSplit;
62 this.dictBuilder = dictBuilder;
63 this.langIndexBuilders = langIndexBuilders;
64 this.bothIndexBuilder = bothIndexBuilder;
67 public void parseFile(final File file) throws IOException {
68 final BufferedReader reader = new BufferedReader(new InputStreamReader(new FileInputStream(file), charset));
70 while ((line = reader.readLine()) != null) {
75 private void parseLine(final String line) {
76 if (line.startsWith("#") || line.length() == 0) {
77 logger.info("Skipping comment line: " + line);
80 final String[] fields = fieldSplit.split(line);
81 if (fields.length != 2) {
82 logger.warning("Malformed line: " + line);
86 fields[0] = SPACES.matcher(fields[0]).replaceAll(" ").trim();
87 fields[1] = SPACES.matcher(fields[1]).replaceAll(" ").trim();
89 final String temp = fields[0];
90 fields[0] = fields[1];
94 final String[][] subfields = new String[2][];
95 if (subfieldSplit != null) {
96 subfields[0] = subfieldSplit.split(fields[0]);
97 subfields[1] = subfieldSplit.split(fields[1]);
98 if (subfields[0].length != subfields[1].length) {
99 logger.warning("Number of subfields doesn't match: " + line);
103 subfields[0] = new String[] { fields[0] };
104 subfields[1] = new String[] { fields[1] };
107 final Pair[] pairs = new Pair[subfields[0].length];
108 for (int i = 0; i < pairs.length; ++i) {
109 pairs[i] = new Pair(subfields[0][i], subfields[1][i]);
111 final PairEntry pairEntry = new PairEntry(pairs);
112 final EntryData entryData = new EntryData(dictBuilder.dictionary.pairEntries.size(), pairEntry);
113 dictBuilder.dictionary.pairEntries.add(pairEntry);
114 dictBuilder.entryDatas.add(entryData); // TODO: delete me.
116 for (int l = 0; l < 2; ++l) {
119 for (int j = 0; j < subfields[l].length; ++j) {
120 String subfield = subfields[l][j];
121 final IndexBuilder indexBuilder = langIndexBuilders[l];
122 if (indexBuilder.index.sortLanguage == Language.de) {
123 subfield = parseField_DE(indexBuilder, subfield, entryData, j);
124 } else if (indexBuilder.index.sortLanguage == Language.en) {
125 subfield = parseField_EN(indexBuilder, subfield, entryData, j);
127 parseFieldGeneric(indexBuilder, subfield, entryData, j, subfields.length);
132 private void parseFieldGeneric(final IndexBuilder indexBuilder, String field,
133 final EntryData entryData, final int subfieldIdx, final int numSubFields) {
134 // remove bracketed and parenthesized stuff.
135 final StringBuilder bracketed = new StringBuilder();
136 final StringBuilder parenthesized = new StringBuilder();
139 while ((matcher = BRACKETED.matcher(field)).matches()) {
140 bracketed.append(matcher.group(1)).append(" ");
141 field = matcher.replaceFirst(" ");
144 while ((matcher = PARENTHESIZED.matcher(field)).matches()) {
145 parenthesized.append(matcher.group(1)).append(" ");
146 field = matcher.replaceFirst(" ");
149 field = SPACES.matcher(field).replaceAll(" ").trim();
151 // split words on non -A-z0-9, do them.
152 final String[] tokens = NON_CHAR_DASH.split(field);
154 final EntryTypeName entryTypeName;
155 if (numSubFields == 1) {
156 assert subfieldIdx == 0;
157 if (tokens.length == 1) {
158 entryTypeName = EntryTypeName.ONE_WORD;
159 } else if (tokens.length == 2) {
160 entryTypeName = EntryTypeName.TWO_WORDS;
161 } else if (tokens.length == 3) {
162 entryTypeName = EntryTypeName.THREE_WORDS;
163 } else if (tokens.length == 4) {
164 entryTypeName = EntryTypeName.FOUR_WORDS;
166 entryTypeName = EntryTypeName.FIVE_OR_MORE_WORDS;
169 assert numSubFields > 1;
170 if (subfieldIdx == 0) {
171 if (tokens.length == 1) {
172 entryTypeName = EntryTypeName.MULTIROW_HEAD_ONE_WORD;
174 entryTypeName = EntryTypeName.MULTIROW_HEAD_MANY_WORDS;
177 assert subfieldIdx > 0;
178 if (tokens.length == 1) {
179 entryTypeName = EntryTypeName.MULTIROW_TAIL_ONE_WORD;
181 entryTypeName = EntryTypeName.MULTIROW_TAIL_MANY_WORDS;
186 for (String token : tokens) {
187 token = TRIM_PUNC.matcher(token).replaceAll("");
188 if (!alreadyDone.contains(token) && token.length() > 0) {
189 final List<EntryData> entries = indexBuilder.getOrCreateEntries(token, entryTypeName);
190 entries.add(entryData);
191 alreadyDone.add(token);
193 // also split words on dashes, do them, too.
194 if (token.contains("-")) {
195 final String[] dashed = token.split("-");
196 for (final String dashedToken : dashed) {
197 if (!alreadyDone.contains(dashedToken) && dashedToken.length() > 0) {
198 final List<EntryData> dashEntries = indexBuilder.getOrCreateEntries(dashedToken, EntryTypeName.PART_OF_HYPHENATED);
199 dashEntries.add(entryData);
204 } // if (!alreadyDone.contains(token)) {
205 } // for (final String token : tokens) {
207 // process bracketed stuff (split on spaces and dashes always)
208 final String[] bracketedTokens = NON_CHAR.split(bracketed.toString());
209 for (final String token : bracketedTokens) {
210 assert !token.contains("-");
211 if (!alreadyDone.contains(token) && token.length() > 0) {
212 final List<EntryData> entries = indexBuilder.getOrCreateEntries(token, EntryTypeName.BRACKETED);
213 entries.add(entryData);
217 // process paren stuff
218 final String[] parenTokens = NON_CHAR.split(bracketed.toString());
219 for (final String token : parenTokens) {
220 assert !token.contains("-");
221 if (!alreadyDone.contains(token) && token.length() > 0) {
222 final List<EntryData> entries = indexBuilder.getOrCreateEntries(token, EntryTypeName.PARENTHESIZED);
223 entries.add(entryData);
229 private String parseField_DE(final IndexBuilder indexBuilder, String field,
230 final EntryData entryData, final int subfieldIdx) {
231 final Matcher matcher = DE_NOUN.matcher(field);
232 while (matcher.find()) {
233 final String noun = matcher.group(1);
234 //final String gender = matcher.group(2);
235 if (alreadyDone.add(noun)) {
236 // System.out.println("Found DE noun " + noun + ", " + gender);
237 final List<EntryData> entries = indexBuilder.getOrCreateEntries(noun, EntryTypeName.NOUN);
238 entries.add(entryData);
244 private String parseField_EN(final IndexBuilder indexBuilder, String field,
245 final EntryData entryData, final int subfieldIdx) {
246 if (field.startsWith("to ")) {
247 field = field.substring(3);