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.Language;
15 import com.hughes.android.dictionary.engine.PairEntry.Pair;
17 public class DictFileParser {
19 static final Logger logger = Logger.getLogger(DictFileParser.class.getName());
22 static final Pattern TAB = Pattern.compile("\\t");
25 static final Pattern DOUBLE_COLON = Pattern.compile(" :: ");
26 static final Pattern PIPE = Pattern.compile(" \\| ");
28 static final Pattern SPACES = Pattern.compile("\\s+");
29 static final Pattern DE_NOUN = Pattern.compile("([^ ]+) *\\{(m|f|n|pl)\\}");
30 static final Pattern EN_VERB = Pattern.compile("^to ([^ ]+)");
32 static final Pattern BRACKETED = Pattern.compile("\\[([^]]+)\\]");
33 static final Pattern PARENTHESIZED = Pattern.compile("\\(([^)]+)\\)");
34 static final Pattern CURLY_BRACED = Pattern.compile("\\{([^}]+)\\}");
36 static final Pattern NON_CHAR_DASH = Pattern.compile("[^-'\\p{L}0-9]+");
37 static final Pattern NON_CHAR = Pattern.compile("[^\\p{L}0-9]+");
39 static final Pattern TRIM_PUNC = Pattern.compile("^[^\\p{L}0-9]+|[^\\p{L}0-9]+$");
41 final Charset charset;
42 final boolean flipCols;
44 final Pattern fieldSplit;
45 final Pattern subfieldSplit;
47 final DictionaryBuilder dictBuilder;
48 final IndexBuilder[] langIndexBuilders;
49 final IndexBuilder bothIndexBuilder;
51 // final Set<String> alreadyDone = new HashSet<String>();
53 public DictFileParser(final Charset charset, boolean flipCols,
54 final Pattern fieldSplit, final Pattern subfieldSplit,
55 final DictionaryBuilder dictBuilder, final IndexBuilder[] langIndexBuilders,
56 final IndexBuilder bothIndexBuilder) {
57 this.charset = charset;
58 this.flipCols = flipCols;
59 this.fieldSplit = fieldSplit;
60 this.subfieldSplit = subfieldSplit;
61 this.dictBuilder = dictBuilder;
62 this.langIndexBuilders = langIndexBuilders;
63 this.bothIndexBuilder = bothIndexBuilder;
66 public void parseFile(final File file) throws IOException {
67 final BufferedReader reader = new BufferedReader(new InputStreamReader(new FileInputStream(file), charset));
69 while ((line = reader.readLine()) != null) {
74 private void parseLine(final String line) {
75 if (line.startsWith("#") || line.length() == 0) {
76 logger.info("Skipping comment line: " + line);
79 final String[] fields = fieldSplit.split(line);
80 if (fields.length != 2) {
81 logger.warning("Malformed line: " + line);
85 fields[0] = SPACES.matcher(fields[0]).replaceAll(" ").trim();
86 fields[1] = SPACES.matcher(fields[1]).replaceAll(" ").trim();
88 final String temp = fields[0];
89 fields[0] = fields[1];
93 final String[][] subfields = new String[2][];
94 if (subfieldSplit != null) {
95 subfields[0] = subfieldSplit.split(fields[0]);
96 subfields[1] = subfieldSplit.split(fields[1]);
97 if (subfields[0].length != subfields[1].length) {
98 logger.warning("Number of subfields doesn't match: " + line);
102 subfields[0] = new String[] { fields[0] };
103 subfields[1] = new String[] { fields[1] };
106 final Pair[] pairs = new Pair[subfields[0].length];
107 for (int i = 0; i < pairs.length; ++i) {
108 pairs[i] = new Pair(subfields[0][i], subfields[1][i]);
110 final PairEntry pairEntry = new PairEntry(pairs);
111 final EntryData entryData = new EntryData(dictBuilder.dictionary.pairEntries.size(), pairEntry);
112 dictBuilder.dictionary.pairEntries.add(pairEntry);
113 dictBuilder.entryDatas.add(entryData); // TODO: delete me.
115 for (int l = 0; l < 2; ++l) {
116 // alreadyDone.clear();
118 for (int j = 0; j < subfields[l].length; ++j) {
119 String subfield = subfields[l][j];
120 final IndexBuilder indexBuilder = langIndexBuilders[l];
121 if (indexBuilder.index.sortLanguage == Language.de) {
122 subfield = parseField_DE(indexBuilder, subfield, entryData, j);
123 } else if (indexBuilder.index.sortLanguage == Language.en) {
124 subfield = parseField_EN(indexBuilder, subfield, entryData, j);
126 parseFieldGeneric(indexBuilder, subfield, entryData, j, subfields[l].length);
131 private void parseFieldGeneric(final IndexBuilder indexBuilder, String field,
132 final EntryData entryData, final int subfieldIdx, final int numSubFields) {
133 // remove bracketed and parenthesized stuff.
134 final StringBuilder bracketed = new StringBuilder();
135 final StringBuilder parenthesized = new StringBuilder();
138 while ((matcher = BRACKETED.matcher(field)).find()) {
139 bracketed.append(matcher.group(1)).append(" ");
140 field = matcher.replaceFirst(" ");
143 while ((matcher = PARENTHESIZED.matcher(field)).find()) {
144 parenthesized.append(matcher.group(1)).append(" ");
145 field = matcher.replaceFirst(" ");
148 field = SPACES.matcher(field).replaceAll(" ").trim();
150 // split words on non -A-z0-9, do them.
151 final String[] tokens = NON_CHAR_DASH.split(field);
153 final EntryTypeName entryTypeName;
154 if (numSubFields == 1) {
155 assert subfieldIdx == 0;
156 if (tokens.length == 1) {
157 entryTypeName = EntryTypeName.ONE_WORD;
158 } else if (tokens.length == 2) {
159 entryTypeName = EntryTypeName.TWO_WORDS;
160 } else if (tokens.length == 3) {
161 entryTypeName = EntryTypeName.THREE_WORDS;
162 } else if (tokens.length == 4) {
163 entryTypeName = EntryTypeName.FOUR_WORDS;
165 entryTypeName = EntryTypeName.FIVE_OR_MORE_WORDS;
168 assert numSubFields > 1;
169 if (subfieldIdx == 0) {
170 if (tokens.length == 1) {
171 entryTypeName = EntryTypeName.MULTIROW_HEAD_ONE_WORD;
173 entryTypeName = EntryTypeName.MULTIROW_HEAD_MANY_WORDS;
176 assert subfieldIdx > 0;
177 if (tokens.length == 1) {
178 entryTypeName = EntryTypeName.MULTIROW_TAIL_ONE_WORD;
180 entryTypeName = EntryTypeName.MULTIROW_TAIL_MANY_WORDS;
185 for (String token : tokens) {
186 token = TRIM_PUNC.matcher(token).replaceAll("");
187 if (/*!alreadyDone.contains(token) && */token.length() > 0) {
188 final List<EntryData> entries = indexBuilder.getOrCreateEntries(token, entryTypeName);
189 entries.add(entryData);
190 // alreadyDone.add(token);
192 // also split words on dashes, do them, too.
193 if (token.contains("-")) {
194 final String[] dashed = token.split("-");
195 for (final String dashedToken : dashed) {
196 if (/*!alreadyDone.contains(dashedToken) && */dashedToken.length() > 0) {
197 final List<EntryData> dashEntries = indexBuilder.getOrCreateEntries(dashedToken, EntryTypeName.PART_OF_HYPHENATED);
198 dashEntries.add(entryData);
203 } // if (!alreadyDone.contains(token)) {
204 } // for (final String token : tokens) {
206 // process bracketed stuff (split on spaces and dashes always)
207 final String[] bracketedTokens = NON_CHAR.split(bracketed.toString());
208 for (final String token : bracketedTokens) {
209 assert !token.contains("-");
210 if (/*!alreadyDone.contains(token) && */token.length() > 0) {
211 final List<EntryData> entries = indexBuilder.getOrCreateEntries(token, EntryTypeName.BRACKETED);
212 entries.add(entryData);
216 // process paren stuff
217 final String[] parenTokens = NON_CHAR.split(parenthesized.toString());
218 for (final String token : parenTokens) {
219 assert !token.contains("-");
220 if (/*!alreadyDone.contains(token) && */token.length() > 0) {
221 final List<EntryData> entries = indexBuilder.getOrCreateEntries(token, EntryTypeName.PARENTHESIZED);
222 entries.add(entryData);
228 private String parseField_DE(final IndexBuilder indexBuilder, String field,
229 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);
242 // In English, curly braces are used for different tenses.
243 field = CURLY_BRACED.matcher(field).replaceAll(" ");
248 private String parseField_EN(final IndexBuilder indexBuilder, String field,
249 final EntryData entryData, final int subfieldIdx) {
250 if (field.startsWith("to ")) {
251 field = field.substring(3);