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));
68 while ((line = reader.readLine()) != null) {
73 private void parseLine(final String line) {
74 if (line.startsWith("#") || line.length() == 0) {
75 logger.info("Skipping comment line: " + line);
78 final String[] fields = fieldSplit.split(line);
79 if (fields.length != 2) {
80 logger.warning("Malformed line: " + line);
84 fields[0] = SPACES.matcher(fields[0]).replaceAll(" ").trim();
85 fields[1] = SPACES.matcher(fields[1]).replaceAll(" ").trim();
87 final String temp = fields[0];
88 fields[0] = fields[1];
92 final String[][] subfields = new String[2][];
93 if (subfieldSplit != null) {
94 subfields[0] = subfieldSplit.split(fields[0]);
95 subfields[1] = subfieldSplit.split(fields[1]);
96 if (subfields[0].length != subfields[1].length) {
97 logger.warning("Number of subfields doesn't match: " + line);
101 subfields[0] = new String[] { fields[0] };
102 subfields[1] = new String[] { fields[1] };
105 final Pair[] pairs = new Pair[subfields[0].length];
106 for (int i = 0; i < pairs.length; ++i) {
107 pairs[i] = new Pair(subfields[0][i], subfields[1][i]);
109 final PairEntry pairEntry = new PairEntry(pairs);
110 final EntryData entryData = new EntryData(dictBuilder.dictionary.pairEntries.size(), pairEntry);
111 dictBuilder.dictionary.pairEntries.add(pairEntry);
112 dictBuilder.entryDatas.add(entryData); // TODO: delete me.
114 for (int l = 0; l < 2; ++l) {
115 // alreadyDone.clear();
117 for (int j = 0; j < subfields[l].length; ++j) {
118 String subfield = subfields[l][j];
119 final IndexBuilder indexBuilder = langIndexBuilders[l];
120 if (indexBuilder.index.sortLanguage == Language.de) {
121 subfield = parseField_DE(indexBuilder, subfield, entryData, j);
122 } else if (indexBuilder.index.sortLanguage == Language.en) {
123 subfield = parseField_EN(indexBuilder, subfield, entryData, j);
125 parseFieldGeneric(indexBuilder, subfield, entryData, j, subfields[l].length);
130 private void parseFieldGeneric(final IndexBuilder indexBuilder, String field,
131 final EntryData entryData, final int subfieldIdx, final int numSubFields) {
132 // remove bracketed and parenthesized stuff.
133 final StringBuilder bracketed = new StringBuilder();
134 final StringBuilder parenthesized = new StringBuilder();
137 while ((matcher = BRACKETED.matcher(field)).find()) {
138 bracketed.append(matcher.group(1)).append(" ");
139 field = matcher.replaceFirst(" ");
142 while ((matcher = PARENTHESIZED.matcher(field)).find()) {
143 parenthesized.append(matcher.group(1)).append(" ");
144 field = matcher.replaceFirst(" ");
147 field = SPACES.matcher(field).replaceAll(" ").trim();
149 // split words on non -A-z0-9, do them.
150 final String[] tokens = NON_CHAR_DASH.split(field);
152 final EntryTypeName entryTypeName;
153 if (numSubFields == 1) {
154 assert subfieldIdx == 0;
155 if (tokens.length == 1) {
156 entryTypeName = EntryTypeName.ONE_WORD;
157 } else if (tokens.length == 2) {
158 entryTypeName = EntryTypeName.TWO_WORDS;
159 } else if (tokens.length == 3) {
160 entryTypeName = EntryTypeName.THREE_WORDS;
161 } else if (tokens.length == 4) {
162 entryTypeName = EntryTypeName.FOUR_WORDS;
164 entryTypeName = EntryTypeName.FIVE_OR_MORE_WORDS;
167 assert numSubFields > 1;
168 if (subfieldIdx == 0) {
169 if (tokens.length == 1) {
170 entryTypeName = EntryTypeName.MULTIROW_HEAD_ONE_WORD;
172 entryTypeName = EntryTypeName.MULTIROW_HEAD_MANY_WORDS;
175 assert subfieldIdx > 0;
176 if (tokens.length == 1) {
177 entryTypeName = EntryTypeName.MULTIROW_TAIL_ONE_WORD;
179 entryTypeName = EntryTypeName.MULTIROW_TAIL_MANY_WORDS;
184 for (String token : tokens) {
185 token = TRIM_PUNC.matcher(token).replaceAll("");
186 if (/*!alreadyDone.contains(token) && */token.length() > 0) {
187 final List<EntryData> entries = indexBuilder.getOrCreateEntries(token, entryTypeName);
188 entries.add(entryData);
189 // alreadyDone.add(token);
191 // also split words on dashes, do them, too.
192 if (token.contains("-")) {
193 final String[] dashed = token.split("-");
194 for (final String dashedToken : dashed) {
195 if (/*!alreadyDone.contains(dashedToken) && */dashedToken.length() > 0) {
196 final List<EntryData> dashEntries = indexBuilder.getOrCreateEntries(dashedToken, EntryTypeName.PART_OF_HYPHENATED);
197 dashEntries.add(entryData);
202 } // if (!alreadyDone.contains(token)) {
203 } // for (final String token : tokens) {
205 // process bracketed stuff (split on spaces and dashes always)
206 final String[] bracketedTokens = NON_CHAR.split(bracketed.toString());
207 for (final String token : bracketedTokens) {
208 assert !token.contains("-");
209 if (/*!alreadyDone.contains(token) && */token.length() > 0) {
210 final List<EntryData> entries = indexBuilder.getOrCreateEntries(token, EntryTypeName.BRACKETED);
211 entries.add(entryData);
215 // process paren stuff
216 final String[] parenTokens = NON_CHAR.split(parenthesized.toString());
217 for (final String token : parenTokens) {
218 assert !token.contains("-");
219 if (/*!alreadyDone.contains(token) && */token.length() > 0) {
220 final List<EntryData> entries = indexBuilder.getOrCreateEntries(token, EntryTypeName.PARENTHESIZED);
221 entries.add(entryData);
227 private String parseField_DE(final IndexBuilder indexBuilder, String field,
228 final EntryData entryData, final int subfieldIdx) {
230 // final Matcher matcher = DE_NOUN.matcher(field);
231 // while (matcher.find()) {
232 // final String noun = matcher.group(1);
233 //final String gender = matcher.group(2);
234 // if (alreadyDone.add(noun)) {
235 // System.out.println("Found DE noun " + noun + ", " + gender);
236 // final List<EntryData> entries = indexBuilder.getOrCreateEntries(noun, EntryTypeName.NOUN);
237 // entries.add(entryData);
241 // In English, curly braces are used for different tenses.
242 field = CURLY_BRACED.matcher(field).replaceAll(" ");
247 private String parseField_EN(final IndexBuilder indexBuilder, String field,
248 final EntryData entryData, final int subfieldIdx) {
249 if (field.startsWith("to ")) {
250 field = field.substring(3);