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