src/com/hughes/android/dictionary/engine/WiktionarySplitter.java

   1 // Copyright 2011 Google Inc. All Rights Reserved.
   2 //
   3 // Licensed under the Apache License, Version 2.0 (the "License");
   4 // you may not use this file except in compliance with the License.
   5 // You may obtain a copy of the License at
   6 //
   7 //     http://www.apache.org/licenses/LICENSE-2.0
   8 //
   9 // Unless required by applicable law or agreed to in writing, software
  10 // distributed under the License is distributed on an "AS IS" BASIS,
  11 // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  12 // See the License for the specific language governing permissions and
  13 // limitations under the License.
  14
  15 package com.hughes.android.dictionary.engine;
  16
  17 import java.io.BufferedInputStream;
  18 import java.io.BufferedOutputStream;
  19 import java.io.DataOutputStream;
  20 import java.io.File;
  21 import java.io.FileInputStream;
  22 import java.io.FileOutputStream;
  23 import java.io.InputStream;
  24 import java.io.IOException;
  25 import java.io.OutputStream;
  26 import java.util.ArrayList;
  27 import java.util.LinkedHashMap;
  28 import java.util.List;
  29 import java.util.Map;
  30 import java.util.regex.Matcher;
  31 import java.util.regex.Pattern;
  32
  33 import javax.xml.parsers.ParserConfigurationException;
  34 import javax.xml.parsers.SAXParser;
  35
  36 import org.apache.xerces.jaxp.SAXParserFactoryImpl;
  37 import org.apache.commons.compress.compressors.CompressorStreamFactory;
  38 import org.xml.sax.Attributes;
  39 import org.xml.sax.SAXException;
  40
  41 import com.hughes.android.dictionary.parser.wiktionary.WiktionaryLangs;
  42
  43 public class WiktionarySplitter extends org.xml.sax.helpers.DefaultHandler {
  44
  45     // The matches the whole line, otherwise regexes don't work well on French:
  46     // {{=uk=}}
  47     // Spanish has no initial headings, tried to also detect {{ES as such
  48     // with "^(\\{\\{ES|(=+)[^=]).*$" but that broke English.
  49     static final Pattern headingStart = Pattern.compile("^(=+)[^=].*$", Pattern.MULTILINE);
  50
  51     final Map<String,List<Selector>> pathToSelectors = new LinkedHashMap<String, List<Selector>>();
  52     List<Selector> currentSelectors = null;
  53
  54     StringBuilder titleBuilder;
  55     StringBuilder textBuilder;
  56     StringBuilder currentBuilder = null;
  57
  58     public static void main(final String[] args) throws Exception {
  59         final WiktionarySplitter wiktionarySplitter = new WiktionarySplitter();
  60         wiktionarySplitter.go();
  61     }
  62
  63     private WiktionarySplitter() {
  64         List<Selector> selectors;
  65         for (final String code : WiktionaryLangs.wikiCodeToIsoCodeToWikiName.keySet()) {
  66             //if (!code.equals("fr")) {continue;}
  67             selectors = new ArrayList<WiktionarySplitter.Selector>();
  68             pathToSelectors.put(String.format("data/inputs/%swiktionary-pages-articles.xml", code), selectors);
  69             for (final Map.Entry<String, String> entry : WiktionaryLangs.wikiCodeToIsoCodeToWikiName.get(code).entrySet()) {
  70                 final String dir = String.format("data/inputs/wikiSplit/%s", code);
  71                 new File(dir).mkdirs();
  72                 selectors.add(new Selector(String.format("%s/%s.data", dir, entry.getKey()), entry.getValue()));
  73             }
  74         }
  75     }
  76
  77     private void go() throws Exception {
  78         final SAXParser parser = SAXParserFactoryImpl.newInstance().newSAXParser();
  79
  80         // Configure things.
  81         for (final Map.Entry<String, List<Selector>> pathToSelectorsEntry : pathToSelectors.entrySet()) {
  82
  83             currentSelectors = pathToSelectorsEntry.getValue();
  84
  85             for (final Selector selector : currentSelectors) {
  86                 OutputStream tmp = new FileOutputStream(selector.outFilename + ".gz");
  87                 tmp = new BufferedOutputStream(tmp);
  88                 tmp = new CompressorStreamFactory().createCompressorOutputStream(CompressorStreamFactory.GZIP, tmp);
  89                 tmp = new WriteBuffer(tmp, 20 * 1024 * 1024);
  90                 selector.out = new DataOutputStream(tmp);
  91             }
  92
  93             // Do it.
  94             try {
  95                 File input = new File(pathToSelectorsEntry.getKey() + ".bz2");
  96                 if (!input.exists()) input = new File(pathToSelectorsEntry.getKey() + ".gz");
  97                 if (!input.exists()) input = new File(pathToSelectorsEntry.getKey() + ".xz");
  98                 if (!input.exists()) {
  99                     // Fallback to uncompressed file
 100                     parser.parse(new File(pathToSelectorsEntry.getKey()), this);
 101                 } else {
 102                     InputStream compressedIn = new BufferedInputStream(new FileInputStream(input));
 103                     InputStream in = new CompressorStreamFactory().createCompressorInputStream(compressedIn);
 104                     in = new ReadAheadBuffer(in, 20 * 1024 * 1024);
 105                     parser.parse(new BufferedInputStream(in), this);
 106                 }
 107             } catch (Exception e) {
 108                 System.err.println("Exception during parse, lastPageTitle=" + lastPageTitle + ", titleBuilder=" + titleBuilder.toString());
 109                 throw e;
 110             }
 111
 112             // Shutdown.
 113             for (final Selector selector : currentSelectors) {
 114                 selector.out.close();
 115             }
 116
 117         }
 118     }
 119
 120     String lastPageTitle = null;
 121     int pageCount = 0;
 122     Pattern endPatterns[] = new Pattern[100];
 123
 124     private Pattern getEndPattern(int depth) {
 125         if (endPatterns[depth] == null)
 126             endPatterns[depth] = Pattern.compile(String.format("^={1,%d}[^=].*$", depth), Pattern.MULTILINE);
 127         return endPatterns[depth];
 128     }
 129
 130     private void endPage() {
 131         final String title = titleBuilder.toString();
 132         lastPageTitle = title;
 133         if (++pageCount % 1000 == 0) {
 134             System.out.println("endPage: " + title + ", count=" + pageCount);
 135         }
 136         if (title.startsWith("Wiktionary:") ||
 137                 title.startsWith("Appendix:") ||
 138                 title.startsWith("Help:") ||
 139                 title.startsWith("Index:") ||
 140                 title.startsWith("MediaWiki:") ||
 141                 title.startsWith("Citations:") ||
 142                 title.startsWith("Concordance:") ||
 143                 title.startsWith("Glossary:") ||
 144                 title.startsWith("Rhymes:") ||
 145                 title.startsWith("Category:") ||
 146                 title.startsWith("Wikisaurus:") ||
 147                 title.startsWith("Unsupported titles/") ||
 148                 title.startsWith("Transwiki:") ||
 149                 title.startsWith("File:") ||
 150                 title.startsWith("Thread:") ||
 151                 title.startsWith("Template:") ||
 152                 title.startsWith("Summary:") ||
 153                 title.startsWith("Module:") ||
 154                 // DE
 155                 title.startsWith("Datei:") ||
 156                 title.startsWith("Verzeichnis:") ||
 157                 title.startsWith("Vorlage:") ||
 158                 title.startsWith("Thesaurus:") ||
 159                 title.startsWith("Kategorie:") ||
 160                 title.startsWith("Hilfe:") ||
 161                 title.startsWith("Reim:") ||
 162                 // FR:
 163                 title.startsWith("Annexe:") ||
 164                 title.startsWith("Catégori:") ||
 165                 title.startsWith("Modèle:") ||
 166                 title.startsWith("Thésaurus:") ||
 167                 title.startsWith("Projet:") ||
 168                 title.startsWith("Aide:") ||
 169                 title.startsWith("Fichier:") ||
 170                 title.startsWith("Wiktionnaire:") ||
 171                 title.startsWith("Catégorie:") ||
 172                 title.startsWith("Portail:") ||
 173                 title.startsWith("utiliusateur:") ||
 174                 title.startsWith("Kategorio:") ||
 175                 // IT
 176                 title.startsWith("Wikizionario:") ||
 177                 title.startsWith("Appendice:") ||
 178                 title.startsWith("Categoria:") ||
 179                 title.startsWith("Aiuto:") ||
 180                 title.startsWith("Portail:") ||
 181                 // ES
 182                 title.startsWith("Apéndice:") ||
 183                 title.startsWith("Archivo:") ||
 184                 title.startsWith("Ayuda:") ||
 185                 title.startsWith("Categoría:") ||
 186                 title.startsWith("Plantilla:") ||
 187                 title.startsWith("Wikcionario:") ||
 188
 189                 // sentinel
 190                 false
 191            ) {
 192             return;
 193         }
 194         if (title.contains(":")) {
 195             if (!title.startsWith("Sign gloss:")) {
 196                 System.err.println("title with colon: " + title);
 197             }
 198         }
 199
 200         String text = textBuilder.toString();
 201         String translingual = "";
 202
 203         while (text.length() > 0) {
 204             // Find start.
 205             final Matcher startMatcher = headingStart.matcher(text);
 206             if (!startMatcher.find()) {
 207                 return;
 208             }
 209             text = text.substring(startMatcher.end());
 210
 211             final String heading = startMatcher.group();
 212             for (final Selector selector : currentSelectors) {
 213                 if (heading.indexOf("Translingual") != -1) {
 214                     // Find end.
 215                     final int depth = startMatcher.group(1).length();
 216                     final Pattern endPattern = getEndPattern(depth);
 217
 218                     final Matcher endMatcher = endPattern.matcher(text);
 219                     if (endMatcher.find()) {
 220                         int end = endMatcher.start();
 221                         translingual = text.substring(0, endMatcher.start());
 222                         text = text.substring(end);
 223                         break;
 224                     }
 225                 }
 226                 if (selector.pattern.matcher(heading).find()) {
 227
 228                     // Find end.
 229                     final int depth = startMatcher.group(1).length();
 230                     final Pattern endPattern = getEndPattern(depth);
 231
 232                     final Matcher endMatcher = endPattern.matcher(text);
 233                     final int end;
 234                     if (endMatcher.find()) {
 235                         end = endMatcher.start();
 236                     } else {
 237                         end = text.length();
 238                     }
 239
 240                     String sectionText = text.substring(0, end);
 241                     // Hack to remove empty dummy section from French
 242                     if (sectionText.startsWith("\n=== {{S|étymologie}} ===\n: {{ébauche-étym")) {
 243                         int dummy_end = sectionText.indexOf("}}", 41) + 2;
 244                         while (dummy_end + 1 < sectionText.length() &&
 245                                 sectionText.charAt(dummy_end) == '\n' &&
 246                                 sectionText.charAt(dummy_end + 1) == '\n') ++dummy_end;
 247                         sectionText = sectionText.substring(dummy_end);
 248                     }
 249                     if (heading.indexOf("Japanese") == -1) sectionText += translingual;
 250                     final Section section = new Section(title, heading, sectionText);
 251
 252                     try {
 253                         selector.out.writeUTF(section.title);
 254                         selector.out.writeUTF(section.heading);
 255                         final byte[] bytes = section.text.getBytes("UTF8");
 256                         selector.out.writeInt(bytes.length);
 257                         selector.out.write(bytes);
 258                     } catch (IOException e) {
 259                         throw new RuntimeException(e);
 260                     }
 261
 262                     text = text.substring(end);
 263                     break;
 264                 }
 265             }
 266         }
 267
 268     }
 269
 270     // -----------------------------------------------------------------------
 271
 272     static class Section implements java.io.Serializable {
 273         private static final long serialVersionUID = -7676549898325856822L;
 274
 275         final String title;
 276         final String heading;
 277         final String text;
 278
 279         public Section(final String title, final String heading, final String text) {
 280             this.title = title;
 281             this.heading = heading;
 282             this.text = text;
 283
 284             //System.out.printf("TITLE:%s\nHEADING:%s\nTEXT:%s\n\n\n\n\n\n", title, heading, text);
 285         }
 286     }
 287
 288     static class Selector {
 289         final String outFilename;
 290         final Pattern pattern;
 291
 292         DataOutputStream out;
 293
 294         public Selector(final String filename, final String pattern) {
 295             this.outFilename = filename;
 296             this.pattern = Pattern.compile(pattern, Pattern.CASE_INSENSITIVE);
 297         }
 298     }
 299
 300     // -----------------------------------------------------------------------
 301
 302     @Override
 303     public void startElement(String uri, String localName, String qName,
 304                              Attributes attributes) {
 305         currentBuilder = null;
 306         if ("page".equals(qName)) {
 307             titleBuilder = new StringBuilder();
 308
 309             // Start with "\n" to better match certain strings.
 310             textBuilder = new StringBuilder("\n");
 311         } else if ("title".equals(qName)) {
 312             currentBuilder = titleBuilder;
 313         } else if ("text".equals(qName)) {
 314             currentBuilder = textBuilder;
 315         }
 316     }
 317
 318     @Override
 319     public void characters(char[] ch, int start, int length) throws SAXException {
 320         if (currentBuilder != null) {
 321             currentBuilder.append(ch, start, length);
 322         }
 323     }
 324
 325     @Override
 326     public void endElement(String uri, String localName, String qName)
 327     throws SAXException {
 328         currentBuilder = null;
 329         if ("page".equals(qName)) {
 330             endPage();
 331         }
 332     }
 333
 334     public void parse(final File file) throws ParserConfigurationException,
 335         SAXException, IOException {
 336         final SAXParser parser = SAXParserFactoryImpl.newInstance().newSAXParser();
 337         parser.parse(file, this);
 338     }
 339
 340 }