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.IOException;
  24 import java.io.InputStream;
  25 import java.io.OutputStream;
  26 import java.nio.charset.StandardCharsets;
  27 import java.util.ArrayList;
  28 import java.util.LinkedHashMap;
  29 import java.util.List;
  30 import java.util.Map;
  31 import java.util.regex.Matcher;
  32 import java.util.regex.Pattern;
  33
  34 import javax.xml.parsers.ParserConfigurationException;
  35 import javax.xml.parsers.SAXParser;
  36 import javax.xml.parsers.SAXParserFactory;
  37
  38 import org.apache.commons.compress.compressors.CompressorStreamFactory;
  39 import org.xml.sax.Attributes;
  40 import org.xml.sax.SAXException;
  41
  42 import com.hughes.android.dictionary.parser.wiktionary.WiktionaryLangs;
  43
  44 public class WiktionarySplitter extends org.xml.sax.helpers.DefaultHandler {
  45
  46     // The matches the whole line, otherwise regexes don't work well on French:
  47     // {{=uk=}}
  48     // Spanish has no initial headings, tried to also detect {{ES as such
  49     // with "^(\\{\\{ES|(=+)[^=]).*$" but that broke English.
  50     static final Matcher headingStart = Pattern.compile("^(=+)[^=].*$", Pattern.MULTILINE).matcher("");
  51     static final Matcher startSpanish = Pattern.compile("\\{\\{ES(\\|[^{}=]*)?}}").matcher("");
  52
  53     final Map<String,List<Selector>> pathToSelectors = new LinkedHashMap<>();
  54     List<Selector> currentSelectors = null;
  55
  56     StringBuilder titleBuilder;
  57     StringBuilder textBuilder;
  58     StringBuilder currentBuilder = null;
  59
  60     public static void main(final String[] args) throws Exception {
  61         final WiktionarySplitter wiktionarySplitter = new WiktionarySplitter();
  62         wiktionarySplitter.go();
  63     }
  64
  65     private WiktionarySplitter() {
  66         List<Selector> selectors;
  67         for (final String code : WiktionaryLangs.wikiCodeToIsoCodeToWikiName.keySet()) {
  68             //if (!code.equals("fr")) {continue;}
  69             selectors = new ArrayList<>();
  70             pathToSelectors.put(String.format("data/inputs/%swiktionary-pages-articles.xml", code), selectors);
  71             for (final Map.Entry<String, String> entry : WiktionaryLangs.wikiCodeToIsoCodeToWikiName.get(code).entrySet()) {
  72                 final String dir = String.format("data/inputs/wikiSplit/%s", code);
  73                 new File(dir).mkdirs();
  74                 selectors.add(new Selector(String.format("%s/%s.data", dir, entry.getKey()), entry.getValue()));
  75             }
  76         }
  77     }
  78
  79     private void go() throws Exception {
  80         final SAXParser parser = SAXParserFactory.newInstance().newSAXParser();
  81
  82         // Configure things.
  83         for (final Map.Entry<String, List<Selector>> pathToSelectorsEntry : pathToSelectors.entrySet()) {
  84
  85             currentSelectors = pathToSelectorsEntry.getValue();
  86
  87             for (final Selector selector : currentSelectors) {
  88                 OutputStream tmp = new FileOutputStream(selector.outFilename + ".gz");
  89                 tmp = new BufferedOutputStream(tmp);
  90                 tmp = new CompressorStreamFactory().createCompressorOutputStream(CompressorStreamFactory.GZIP, tmp);
  91                 tmp = new WriteBuffer(tmp, 20 * 1024 * 1024);
  92                 selector.out = new DataOutputStream(tmp);
  93             }
  94
  95             // Do it.
  96             try {
  97                 File input = new File(pathToSelectorsEntry.getKey() + ".bz2");
  98                 if (!input.exists()) input = new File(pathToSelectorsEntry.getKey() + ".gz");
  99                 if (!input.exists()) input = new File(pathToSelectorsEntry.getKey() + ".xz");
 100                 if (!input.exists()) {
 101                     // Fallback to uncompressed file
 102                     parser.parse(new File(pathToSelectorsEntry.getKey()), this);
 103                 } else {
 104                     InputStream compressedIn = new BufferedInputStream(new FileInputStream(input));
 105                     InputStream in = new CompressorStreamFactory().createCompressorInputStream(compressedIn);
 106                     in = new ReadAheadBuffer(in, 20 * 1024 * 1024);
 107                     parser.parse(new BufferedInputStream(in), this);
 108                 }
 109             } catch (Exception e) {
 110                 System.err.println("Exception during parse, lastPageTitle=" + lastPageTitle + ", titleBuilder=" + titleBuilder + " of file " + pathToSelectorsEntry.getKey());
 111                 throw e;
 112             }
 113
 114             // Shutdown.
 115             for (final Selector selector : currentSelectors) {
 116                 selector.out.close();
 117             }
 118
 119         }
 120     }
 121
 122     String lastPageTitle = null;
 123     int pageCount = 0;
 124     final Matcher[] endPatterns = new Matcher[100];
 125
 126     private Matcher getEndPattern(int depth) {
 127         if (endPatterns[depth] == null)
 128             endPatterns[depth] = Pattern.compile(String.format("^={1,%d}[^=].*$", depth), Pattern.MULTILINE).matcher("");
 129         return endPatterns[depth];
 130     }
 131
 132     private void endPage() {
 133         final String title = titleBuilder.toString();
 134         lastPageTitle = title;
 135         if (++pageCount % 100000 == 0) {
 136             System.out.println("endPage: " + title + ", count=" + pageCount);
 137         }
 138         if (title.startsWith("Unsupported titles/")) return;
 139         if (title.contains(":")) {
 140             if (title.startsWith("Wiktionary:") ||
 141                 title.startsWith("Appendix:") ||
 142                 title.startsWith("Help:") ||
 143                 title.startsWith("Index:") ||
 144                 title.startsWith("MediaWiki:") ||
 145                 title.startsWith("Citations:") ||
 146                 title.startsWith("Concordance:") ||
 147                 title.startsWith("Glossary:") ||
 148                 title.startsWith("Rhymes:") ||
 149                 title.startsWith("Category:") ||
 150                 title.startsWith("Wikisaurus:") ||
 151                 title.startsWith("Transwiki:") ||
 152                 title.startsWith("File:") ||
 153                 title.startsWith("Thread:") ||
 154                 title.startsWith("Template:") ||
 155                 title.startsWith("Summary:") ||
 156                 title.startsWith("Module:") ||
 157                 title.startsWith("Reconstruction:") ||
 158                 // DE
 159                 title.startsWith("Datei:") ||
 160                 title.startsWith("Verzeichnis:") ||
 161                 title.startsWith("Vorlage:") ||
 162                 title.startsWith("Thesaurus:") ||
 163                 title.startsWith("Kategorie:") ||
 164                 title.startsWith("Hilfe:") ||
 165                 title.startsWith("Reim:") ||
 166                 title.startsWith("Modul:") ||
 167                 // FR:
 168                 title.startsWith("Annexe:") ||
 169                 title.startsWith("Catégori:") ||
 170                 title.startsWith("Modèle:") ||
 171                 title.startsWith("Thésaurus:") ||
 172                 title.startsWith("Projet:") ||
 173                 title.startsWith("Aide:") ||
 174                 title.startsWith("Fichier:") ||
 175                 title.startsWith("Wiktionnaire:") ||
 176                 title.startsWith("Translations:Wiktionnaire:") ||
 177                 title.startsWith("Translations:Projet:") ||
 178                 title.startsWith("Catégorie:") ||
 179                 title.startsWith("Portail:") ||
 180                 title.startsWith("utiliusateur:") ||
 181                 title.startsWith("Kategorio:") ||
 182                 title.startsWith("Tutoriel:") ||
 183                 // IT
 184                 title.startsWith("Wikizionario:") ||
 185                 title.startsWith("Appendice:") ||
 186                 title.startsWith("Categoria:") ||
 187                 title.startsWith("Aiuto:") ||
 188                 title.startsWith("Portail:") ||
 189                 title.startsWith("Modulo:") ||
 190                 // ES
 191                 title.startsWith("Apéndice:") ||
 192                 title.startsWith("Archivo:") ||
 193                 title.startsWith("Ayuda:") ||
 194                 title.startsWith("Categoría:") ||
 195                 title.startsWith("Plantilla:") ||
 196                 title.startsWith("Wikcionario:") ||
 197
 198                 // PT
 199                 title.startsWith("Ajuda:") ||
 200                 title.startsWith("Apêndice:") ||
 201                 title.startsWith("Citações:") ||
 202                 title.startsWith("Portal:") ||
 203                 title.startsWith("Predefinição:") ||
 204                 title.startsWith("Vocabulário:") ||
 205                 title.startsWith("Wikcionário:") ||
 206                 title.startsWith("Módulo:") ||
 207
 208                 // sentinel
 209                 false
 210                ) return;
 211             // leave the Flexion: pages in for now and do not warn about them
 212             if (!title.startsWith("Sign gloss:") && !title.startsWith("Flexion:")) {
 213                 System.err.println("title with colon: " + title);
 214             }
 215         }
 216
 217         String text = textBuilder.toString();
 218         // Workaround for Spanish wiktionary {{ES}} and {{ES|word}} patterns
 219         text = startSpanish.reset(text).replaceAll("== {{lengua|es}} ==");
 220         String translingual = "";
 221         int start = 0;
 222         headingStart.reset(text);
 223
 224         while (start < text.length()) {
 225             // Find start.
 226             if (!headingStart.find(start)) {
 227                 return;
 228             }
 229             start = headingStart.end();
 230
 231             final String heading = headingStart.group();
 232
 233             // For Translingual entries just store the text for later
 234             // use in the per-language sections
 235             if (heading.contains("Translingual")) {
 236                 // Find end.
 237                 final int depth = headingStart.group(1).length();
 238                 final Matcher endMatcher = getEndPattern(depth).reset(text);
 239
 240                 if (endMatcher.find(start)) {
 241                     int end = endMatcher.start();
 242                     translingual = text.substring(start, end);
 243                     start = end;
 244                     continue;
 245                 }
 246             }
 247
 248             for (final Selector selector : currentSelectors) {
 249                 if (selector.pattern.reset(heading).find()) {
 250                     // Find end.
 251                     final int depth = headingStart.group(1).length();
 252                     final Matcher endMatcher = getEndPattern(depth).reset(text);
 253
 254                     final int end;
 255                     if (endMatcher.find(start)) {
 256                         end = endMatcher.start();
 257                     } else {
 258                         end = text.length();
 259                     }
 260
 261                     String sectionText = text.substring(start, end);
 262                     // Hack to remove empty dummy section from French
 263                     if (sectionText.startsWith("\n=== {{S|étymologie}} ===\n: {{ébauche-étym")) {
 264                         int dummy_end = sectionText.indexOf("}}", 41) + 2;
 265                         while (dummy_end + 1 < sectionText.length() &&
 266                                 sectionText.charAt(dummy_end) == '\n' &&
 267                                 sectionText.charAt(dummy_end + 1) == '\n') ++dummy_end;
 268                         sectionText = sectionText.substring(dummy_end);
 269                     }
 270                     if (!heading.contains("Japanese")) sectionText += translingual;
 271                     final Section section = new Section(title, heading, sectionText);
 272
 273                     try {
 274                         selector.out.writeUTF(section.title);
 275                         selector.out.writeUTF(section.heading);
 276                         final byte[] bytes = section.text.getBytes(StandardCharsets.UTF_8);
 277                         selector.out.writeInt(bytes.length);
 278                         selector.out.write(bytes);
 279                     } catch (IOException e) {
 280                         throw new RuntimeException(e);
 281                     }
 282
 283                     start = end;
 284                     break;
 285                 }
 286             }
 287         }
 288
 289     }
 290
 291     // -----------------------------------------------------------------------
 292
 293     static class Section implements java.io.Serializable {
 294         private static final long serialVersionUID = -7676549898325856822L;
 295
 296         final String title;
 297         final String heading;
 298         final String text;
 299
 300         public Section(final String title, final String heading, final String text) {
 301             this.title = title;
 302             this.heading = heading;
 303             this.text = text;
 304
 305             //System.out.printf("TITLE:%s\nHEADING:%s\nTEXT:%s\n\n\n\n\n\n", title, heading, text);
 306         }
 307     }
 308
 309     static class Selector {
 310         final String outFilename;
 311         final Matcher pattern;
 312
 313         DataOutputStream out;
 314
 315         public Selector(final String filename, final String pattern) {
 316             this.outFilename = filename;
 317             this.pattern = Pattern.compile(pattern, Pattern.CASE_INSENSITIVE).matcher("");
 318         }
 319     }
 320
 321     // -----------------------------------------------------------------------
 322
 323     @Override
 324     public void startElement(String uri, String localName, String qName,
 325                              Attributes attributes) {
 326         currentBuilder = null;
 327         if ("page".equals(qName)) {
 328             titleBuilder = new StringBuilder();
 329
 330             // Start with "\n" to better match certain strings.
 331             textBuilder = new StringBuilder("\n");
 332         } else if ("title".equals(qName)) {
 333             currentBuilder = titleBuilder;
 334         } else if ("text".equals(qName)) {
 335             currentBuilder = textBuilder;
 336         }
 337     }
 338
 339     @Override
 340     public void characters(char[] ch, int start, int length) {
 341         if (currentBuilder != null) {
 342             currentBuilder.append(ch, start, length);
 343         }
 344     }
 345
 346     @Override
 347     public void endElement(String uri, String localName, String qName) {
 348         currentBuilder = null;
 349         if ("page".equals(qName)) {
 350             endPage();
 351         }
 352     }
 353
 354     public void parse(final File file) throws ParserConfigurationException,
 355         SAXException, IOException {
 356         final SAXParser parser = SAXParserFactory.newInstance().newSAXParser();
 357         parser.parse(file, this);
 358     }
 359
 360 }