src/com/hughes/android/dictionary/parser/wiktionary/WholeSectionToHtmlParser.java

   1
   2 package com.hughes.android.dictionary.parser.wiktionary;
   3
   4 import java.net.URI;
   5 import java.util.ArrayList;
   6 import java.util.LinkedHashMap;
   7 import java.util.List;
   8 import java.util.Map;
   9 import java.util.regex.Pattern;
  10
  11 import org.apache.commons.text.StringEscapeUtils;
  12
  13 import com.hughes.android.dictionary.engine.EntryTypeName;
  14 import com.hughes.android.dictionary.engine.HtmlEntry;
  15 import com.hughes.android.dictionary.engine.IndexBuilder;
  16 import com.hughes.android.dictionary.engine.IndexBuilder.TokenData;
  17 import com.hughes.android.dictionary.engine.IndexedEntry;
  18 import com.hughes.android.dictionary.parser.WikiTokenizer;
  19 import com.hughes.util.StringUtil;
  20
  21 public class WholeSectionToHtmlParser extends AbstractWiktionaryParser {
  22
  23     public static final String NAME = "WholeSectionToHtmlParser";
  24
  25     interface LangConfig {
  26         boolean skipSection(final String name);
  27         EntryTypeName sectionNameToEntryType(String sectionName);
  28         boolean skipWikiLink(final WikiTokenizer wikiTokenizer);
  29         String adjustWikiLink(String wikiLinkDest, final String wikiLinkText);
  30         void addFunctionCallbacks(
  31             Map<String, FunctionCallback<WholeSectionToHtmlParser>> functionCallbacks);
  32     }
  33     static final Map<String,LangConfig> isoToLangConfig = new LinkedHashMap<>();
  34     static {
  35         final Pattern enSkipSections = Pattern.compile(".*(Translations|Anagrams|References).*");
  36         isoToLangConfig.put("EN", new LangConfig() {
  37             @Override
  38             public boolean skipSection(String headingText) {
  39                 return enSkipSections.matcher(headingText).matches();
  40             }
  41
  42             @Override
  43             public EntryTypeName sectionNameToEntryType(String sectionName) {
  44                 if (sectionName.equalsIgnoreCase("Synonyms")) {
  45                     return EntryTypeName.SYNONYM_MULTI;
  46                 }
  47                 if (sectionName.equalsIgnoreCase("Antonyms")) {
  48                     return EntryTypeName.ANTONYM_MULTI;
  49                 }
  50                 // We need to put it in the other index, too (probably) ?
  51                 // EnParser.partOfSpeechHeader.matcher(sectionName).matches()
  52
  53                 // Needs special handling?
  54                 // sectionName.equalsIgnoreCase("Derived Terms")
  55                 return null;
  56             }
  57
  58             @Override
  59             public boolean skipWikiLink(WikiTokenizer wikiTokenizer) {
  60                 final String wikiText = wikiTokenizer.wikiLinkText();
  61                 return wikiText.startsWith("Category:");
  62             }
  63             @Override
  64             public String adjustWikiLink(String wikiLinkDest, String wikiLinkText) {
  65                 if (wikiLinkDest.startsWith("w:") || wikiLinkDest.startsWith("Image:")) {
  66                     return null;
  67                 }
  68                 final int hashPos = wikiLinkDest.indexOf("#");
  69                 if (hashPos != -1) {
  70                     wikiLinkDest = wikiLinkDest.substring(0, hashPos);
  71                     if (wikiLinkDest.isEmpty()) {
  72                         wikiLinkDest = wikiLinkText;
  73                     }
  74                 }
  75                 return wikiLinkDest;
  76             }
  77
  78             @Override
  79             public void addFunctionCallbacks(
  80                 Map<String, FunctionCallback<WholeSectionToHtmlParser>> functionCallbacks) {
  81                 EnFunctionCallbacks.addGenericCallbacks(functionCallbacks);
  82             }
  83         });
  84
  85         final Pattern esSkipSections = Pattern.compile(".*(Traducciones|Locuciones).*");
  86         isoToLangConfig.put("ES", new LangConfig() {
  87             @Override
  88             public boolean skipSection(String headingText) {
  89                 return esSkipSections.matcher(headingText).matches();
  90             }
  91
  92             @Override
  93             public EntryTypeName sectionNameToEntryType(String sectionName) {
  94                 if (sectionName.equalsIgnoreCase("sinónimo") || sectionName.equalsIgnoreCase("sinónimos")) {
  95                     return EntryTypeName.SYNONYM_MULTI;
  96                 }
  97                 if (sectionName.equalsIgnoreCase("antónimo") || sectionName.equalsIgnoreCase("antónimos")) {
  98                     return EntryTypeName.ANTONYM_MULTI;
  99                 }
 100                 return null;
 101             }
 102
 103             @Override
 104             public boolean skipWikiLink(WikiTokenizer wikiTokenizer) {
 105                 final String wikiText = wikiTokenizer.wikiLinkText();
 106                 return wikiText.startsWith("Categoría:");
 107             }
 108             @Override
 109             public String adjustWikiLink(String wikiLinkDest, String wikiLinkText) {
 110                 if (wikiLinkDest.startsWith("w:") || wikiLinkDest.startsWith("Image:")) {
 111                     return null;
 112                 }
 113                 final int hashPos = wikiLinkDest.indexOf("#");
 114                 if (hashPos != -1) {
 115                     wikiLinkDest = wikiLinkDest.substring(0, hashPos);
 116                     if (wikiLinkDest.isEmpty()) {
 117                         wikiLinkDest = wikiLinkText;
 118                     }
 119                 }
 120                 return wikiLinkDest;
 121             }
 122
 123             @Override
 124             public void addFunctionCallbacks(
 125                 Map<String, FunctionCallback<WholeSectionToHtmlParser>> functionCallbacks) {
 126                 // TODO: need Spanish variant
 127             }
 128         });
 129
 130         final Pattern ptSkipSections = Pattern.compile(".*Tradução.*");
 131         isoToLangConfig.put("PT", new LangConfig() {
 132             @Override
 133             public boolean skipSection(String headingText) {
 134                 return esSkipSections.matcher(headingText).matches();
 135             }
 136
 137             @Override
 138             public EntryTypeName sectionNameToEntryType(String sectionName) {
 139                 if (sectionName.equalsIgnoreCase("Sinônimo") || sectionName.equalsIgnoreCase("Sinônimos")) {
 140                     return EntryTypeName.SYNONYM_MULTI;
 141                 }
 142                 if (sectionName.equalsIgnoreCase("Antônimo") || sectionName.equalsIgnoreCase("Antônimos")) {
 143                     return EntryTypeName.ANTONYM_MULTI;
 144                 }
 145                 return null;
 146             }
 147
 148             @Override
 149             public boolean skipWikiLink(WikiTokenizer wikiTokenizer) {
 150                 final String wikiText = wikiTokenizer.wikiLinkText();
 151                 return wikiText.startsWith("Categoria:");
 152             }
 153             @Override
 154             public String adjustWikiLink(String wikiLinkDest, String wikiLinkText) {
 155                 if (wikiLinkDest.startsWith("w:") || wikiLinkDest.startsWith("Image:")) {
 156                     return null;
 157                 }
 158                 final int hashPos = wikiLinkDest.indexOf("#");
 159                 if (hashPos != -1) {
 160                     wikiLinkDest = wikiLinkDest.substring(0, hashPos);
 161                     if (wikiLinkDest.isEmpty()) {
 162                         wikiLinkDest = wikiLinkText;
 163                     }
 164                 }
 165                 return wikiLinkDest;
 166             }
 167
 168             @Override
 169             public void addFunctionCallbacks(
 170                 Map<String, FunctionCallback<WholeSectionToHtmlParser>> functionCallbacks) {
 171                 // TODO: need Portuguese variant
 172             }
 173         });
 174
 175         final Pattern deSkipSections = Pattern.compile(".*(Übersetzungen|Referenzen|Quellen).*");
 176         isoToLangConfig.put("DE", new LangConfig() {
 177             @Override
 178             public boolean skipSection(String headingText) {
 179                 return deSkipSections.matcher(headingText).matches();
 180             }
 181
 182             @Override
 183             public EntryTypeName sectionNameToEntryType(String sectionName) {
 184                 if (sectionName.equalsIgnoreCase("Synonyme")) {
 185                     return EntryTypeName.SYNONYM_MULTI;
 186                 }
 187                 if (sectionName.equalsIgnoreCase("Gegenwörter")) {
 188                     return EntryTypeName.ANTONYM_MULTI;
 189                 }
 190                 return null;
 191             }
 192
 193             @Override
 194             public boolean skipWikiLink(WikiTokenizer wikiTokenizer) {
 195                 final String wikiText = wikiTokenizer.wikiLinkText();
 196                 return wikiText.startsWith("Kategorie:");
 197             }
 198             @Override
 199             public String adjustWikiLink(String wikiLinkDest, String wikiLinkText) {
 200                 if (wikiLinkDest.startsWith("w:") || wikiLinkDest.startsWith("Image:")) {
 201                     return null;
 202                 }
 203                 final int hashPos = wikiLinkDest.indexOf("#");
 204                 if (hashPos != -1) {
 205                     wikiLinkDest = wikiLinkDest.substring(0, hashPos);
 206                     if (wikiLinkDest.isEmpty()) {
 207                         wikiLinkDest = wikiLinkText;
 208                     }
 209                 }
 210                 return wikiLinkDest;
 211             }
 212
 213             @Override
 214             public void addFunctionCallbacks(
 215                 Map<String, FunctionCallback<WholeSectionToHtmlParser>> functionCallbacks) {
 216                 DeFunctionCallbacks.addGenericCallbacks(functionCallbacks);
 217             }
 218         });
 219
 220         final Pattern itSkipSections = Pattern.compile(".*(Traduzione|Note / Riferimenti).*");
 221         isoToLangConfig.put("IT", new LangConfig() {
 222             @Override
 223             public boolean skipSection(String headingText) {
 224                 return itSkipSections.matcher(headingText).matches();
 225             }
 226
 227             @Override
 228             public EntryTypeName sectionNameToEntryType(String sectionName) {
 229                 if (sectionName.equalsIgnoreCase("Sinonimi")) {
 230                     return EntryTypeName.SYNONYM_MULTI;
 231                 }
 232                 if (sectionName.equalsIgnoreCase("Antonimi/Contrari")) {
 233                     return EntryTypeName.ANTONYM_MULTI;
 234                 }
 235                 return null;
 236             }
 237
 238             @Override
 239             public boolean skipWikiLink(WikiTokenizer wikiTokenizer) {
 240                 final String wikiText = wikiTokenizer.wikiLinkText();
 241                 return wikiText.startsWith("Categoria:");
 242             }
 243             @Override
 244             public String adjustWikiLink(String wikiLinkDest, String wikiLinkText) {
 245                 if (wikiLinkDest.startsWith("w:") || wikiLinkDest.startsWith("Image:")) {
 246                     return null;
 247                 }
 248                 final int hashPos = wikiLinkDest.indexOf("#");
 249                 if (hashPos != -1) {
 250                     wikiLinkDest = wikiLinkDest.substring(0, hashPos);
 251                     if (wikiLinkDest.isEmpty()) {
 252                         wikiLinkDest = wikiLinkText;
 253                     }
 254                 }
 255                 return wikiLinkDest;
 256             }
 257
 258             @Override
 259             public void addFunctionCallbacks(
 260                 Map<String, FunctionCallback<WholeSectionToHtmlParser>> functionCallbacks) {
 261                 ItFunctionCallbacks.addGenericCallbacks(functionCallbacks);
 262             }
 263         });
 264
 265
 266         final Pattern frSkipSections = Pattern.compile(".*([Tt]raductions|[Aa]nagrammes).*");
 267         isoToLangConfig.put("FR", new LangConfig() {
 268             @Override
 269             public boolean skipSection(String headingText) {
 270                 return frSkipSections.matcher(headingText).matches();
 271             }
 272
 273             @Override
 274             public EntryTypeName sectionNameToEntryType(String sectionName) {
 275                 if (sectionName.equalsIgnoreCase("Synonymes")) {
 276                     return EntryTypeName.SYNONYM_MULTI;
 277                 }
 278                 if (sectionName.equalsIgnoreCase("Antonymes")) {
 279                     return EntryTypeName.ANTONYM_MULTI;
 280                 }
 281                 return null;
 282             }
 283
 284             @Override
 285             public boolean skipWikiLink(WikiTokenizer wikiTokenizer) {
 286                 final String wikiText = wikiTokenizer.wikiLinkText();
 287                 return wikiText.startsWith("Catégorie:");
 288             }
 289             @Override
 290             public String adjustWikiLink(String wikiLinkDest, String wikiLinkText) {
 291                 if (wikiLinkDest.startsWith("w:") || wikiLinkDest.startsWith("Image:")) {
 292                     return null;
 293                 }
 294                 final int hashPos = wikiLinkDest.indexOf("#");
 295                 if (hashPos != -1) {
 296                     wikiLinkDest = wikiLinkDest.substring(0, hashPos);
 297                     if (wikiLinkDest.isEmpty()) {
 298                         wikiLinkDest = wikiLinkText;
 299                     }
 300                 }
 301                 return wikiLinkDest;
 302             }
 303
 304             @Override
 305             public void addFunctionCallbacks(
 306                 Map<String, FunctionCallback<WholeSectionToHtmlParser>> functionCallbacks) {
 307                 FrFunctionCallbacks.addGenericCallbacks(functionCallbacks);
 308             }
 309         });
 310     }
 311
 312     final IndexBuilder titleIndexBuilder;
 313     final IndexBuilder defIndexBuilder;
 314     final String skipLangIso;
 315     final LangConfig langConfig;
 316     final String webUrlTemplate;
 317
 318
 319     public WholeSectionToHtmlParser(final IndexBuilder titleIndexBuilder, final IndexBuilder defIndexBuilder, final String wiktionaryIso, final String skipLangIso,
 320                                     final String webUrlTemplate) {
 321         this.titleIndexBuilder = titleIndexBuilder;
 322         this.defIndexBuilder = defIndexBuilder;
 323         assert isoToLangConfig.containsKey(wiktionaryIso): wiktionaryIso;
 324         this.langConfig = isoToLangConfig.get(wiktionaryIso);
 325         this.skipLangIso = skipLangIso;
 326         this.webUrlTemplate = webUrlTemplate;
 327     }
 328
 329     IndexedEntry indexedEntry = null;
 330
 331     @Override
 332     public void parseSection(String heading, String text) {
 333         assert entrySource != null;
 334         final HtmlEntry htmlEntry = new HtmlEntry(entrySource, title);
 335         indexedEntry = new IndexedEntry(htmlEntry);
 336
 337         final AppendAndIndexWikiCallback<WholeSectionToHtmlParser> callback = new AppendCallback(
 338             this);
 339         langConfig.addFunctionCallbacks(callback.functionCallbacks);
 340
 341         callback.builder = new StringBuilder();
 342         callback.indexedEntry = indexedEntry;
 343         callback.dispatch(text, null);
 344
 345         if (webUrlTemplate != null) {
 346             final String webUrl = String.format(webUrlTemplate, title);
 347             String asciiWebUrl = null;
 348             // URI.create can raise an exception e.g. if webUrl contains %, just ignore those cases.
 349             try {
 350                 asciiWebUrl = URI.create(webUrl).toASCIIString();
 351             } catch (Exception e) {
 352             }
 353             if (asciiWebUrl != null) {
 354                 callback.builder.append("<p> <a href=\"");
 355                 callback.builder.append(asciiWebUrl);
 356                 callback.builder.append("\">");
 357                 callback.builder.append(escapeHtmlLiteral(webUrl));
 358                 callback.builder.append("</a>");
 359             }
 360         }
 361         htmlEntry.html = callback.builder.toString();
 362         indexedEntry.isValid = true;
 363
 364         final TokenData tokenData = titleIndexBuilder.getOrCreateTokenData(title);
 365         tokenData.hasMainEntry = true;
 366
 367         htmlEntry.addToDictionary(titleIndexBuilder.index.dict);
 368         tokenData.htmlEntries.add(htmlEntry);
 369         // titleIndexBuilder.addEntryWithString(indexedEntry, title,
 370         // EntryTypeName.WIKTIONARY_TITLE_MULTI_DETAIL);
 371
 372         indexedEntry = null;
 373     }
 374
 375     @Override
 376     void removeUselessArgs(Map<String, String> namedArgs) {
 377     }
 378
 379     @Override
 380     public void addLinkToCurrentEntry(String token, final String lang, EntryTypeName entryTypeName) {
 381         if (lang == null || lang.equals(skipLangIso)) {
 382             titleIndexBuilder.addEntryWithString(indexedEntry, token, entryTypeName);
 383         }
 384     }
 385
 386     public static String escapeHtmlLiteral(final String plainText) {
 387         final String htmlEscaped = StringEscapeUtils.escapeHtml3(plainText);
 388         if (StringUtil.isAscii(htmlEscaped)) {
 389             return htmlEscaped;
 390         } else {
 391             return StringUtil.escapeUnicodeToPureHtml(plainText);
 392         }
 393
 394     }
 395
 396
 397
 398     class AppendCallback extends AppendAndIndexWikiCallback<WholeSectionToHtmlParser> {
 399         public AppendCallback(WholeSectionToHtmlParser parser) {
 400             super(parser);
 401         }
 402
 403         @Override
 404         public void onPlainText(String plainText) {
 405             super.onPlainText(escapeHtmlLiteral(plainText));
 406         }
 407
 408         @Override
 409         public void onWikiLink(WikiTokenizer wikiTokenizer) {
 410             if (wikiTokenizer.wikiLinkText().endsWith(":" + title)) {
 411                 // Skips wikilinks like: [[en::dick]]
 412                 return;
 413             }
 414             if (langConfig.skipWikiLink(wikiTokenizer)) {
 415                 return;
 416             }
 417             String linkDest;
 418             if (wikiTokenizer.wikiLinkDest() != null) {
 419                 linkDest = langConfig.adjustWikiLink(wikiTokenizer.wikiLinkDest(), wikiTokenizer.wikiLinkText());
 420             } else {
 421                 linkDest = wikiTokenizer.wikiLinkText();
 422             }
 423             if (sectionEntryTypeName != null) {
 424                 // TODO: inside a definition, this could be the wrong language.
 425                 titleIndexBuilder.addEntryWithString(indexedEntry, wikiTokenizer.wikiLinkText(), sectionEntryTypeName);
 426             }
 427             if (!StringUtil.isNullOrEmpty(linkDest)) {
 428                 builder.append("<a href=\"");
 429                 builder.append(HtmlEntry.formatQuickdicUrl("", linkDest));
 430                 builder.append("\">");
 431                 super.onWikiLink(wikiTokenizer);
 432                 builder.append("</a>");
 433             } else {
 434                 super.onWikiLink(wikiTokenizer);
 435             }
 436         }
 437
 438         @Override
 439         public void onFunction(WikiTokenizer wikiTokenizer, String name,
 440                                List<String> args, Map<String, String> namedArgs) {
 441             if (skipLangIso.equalsIgnoreCase(namedArgs.get("lang"))) {
 442                 namedArgs.remove("lang");
 443             }
 444             super.onFunction(wikiTokenizer, name, args, namedArgs);
 445         }
 446
 447         @Override
 448         public void onHtml(WikiTokenizer wikiTokenizer) {
 449             super.onHtml(wikiTokenizer);
 450         }
 451
 452         @Override
 453         public void onNewline(WikiTokenizer wikiTokenizer) {
 454         }
 455
 456         EntryTypeName sectionEntryTypeName;
 457         IndexBuilder currentIndexBuilder;
 458
 459         @Override
 460         public void onHeading(WikiTokenizer wikiTokenizer) {
 461             final String headingText = wikiTokenizer.headingWikiText();
 462             sectionEntryTypeName = langConfig.sectionNameToEntryType(headingText);
 463             final int depth = wikiTokenizer.headingDepth();
 464             if (langConfig.skipSection(headingText)) {
 465                 //System.out.println("Skipping section:" + headingText);
 466                 while ((wikiTokenizer = wikiTokenizer.nextToken()) != null) {
 467                     if (wikiTokenizer.isHeading() && wikiTokenizer.headingDepth() <= depth) {
 468                         // System.out.println("Resume on: " + wikiTokenizer.token());
 469                         wikiTokenizer.returnToLineStart();
 470                         return;
 471                     } else {
 472                         // System.out.println("Skipped: " + wikiTokenizer.token());
 473                     }
 474                 }
 475                 return;
 476             }
 477             builder.append("\n<h");
 478             builder.append(depth);
 479             builder.append('>');
 480             dispatch(headingText, null);
 481             builder.append("</h");
 482             builder.append(depth);
 483             builder.append(">\n");
 484         }
 485
 486         final List<Character> listPrefixStack = new ArrayList<>();
 487
 488         @Override
 489         public void onListItem(WikiTokenizer wikiTokenizer) {
 490             if (builder.length() != 0 && builder.charAt(builder.length() - 1) != '\n') {
 491                 builder.append("\n");
 492             }
 493             final String prefix = wikiTokenizer.listItemPrefix();
 494             while (listPrefixStack.size() < prefix.length()) {
 495                 builder.append('<');
 496                 builder.append(WikiTokenizer.getListTag(prefix.charAt(listPrefixStack.size())));
 497                 builder.append('>');
 498                 listPrefixStack.add(prefix.charAt(listPrefixStack.size()));
 499             }
 500             builder.append("<li>");
 501             dispatch(wikiTokenizer.listItemWikiText(), null);
 502             builder.append("</li>\n");
 503
 504             WikiTokenizer nextToken = wikiTokenizer.nextToken();
 505             boolean returnToLineStart = false;
 506             if (nextToken != null && nextToken.isNewline()) {
 507                 nextToken = nextToken.nextToken();
 508                 returnToLineStart = true;
 509             }
 510             final String nextListHeader;
 511             if (nextToken == null || !nextToken.isListItem()) {
 512                 nextListHeader = "";
 513             } else {
 514                 nextListHeader = nextToken.listItemPrefix();
 515             }
 516             if (returnToLineStart) {
 517                 wikiTokenizer.returnToLineStart();
 518             }
 519             while (listPrefixStack.size() > nextListHeader.length()) {
 520                 final char prefixChar = listPrefixStack.remove(listPrefixStack.size() - 1);
 521                 builder.append("</");
 522                 builder.append(WikiTokenizer.getListTag(prefixChar));
 523                 builder.append(">\n");
 524             }
 525         }
 526
 527         boolean boldOn = false;
 528         boolean italicOn = false;
 529
 530         @Override
 531         public void onMarkup(WikiTokenizer wikiTokenizer) {
 532             if ("'''".equals(wikiTokenizer.token())) {
 533                 if (!boldOn) {
 534                     builder.append("<b>");
 535                 } else {
 536                     builder.append("</b>");
 537                 }
 538                 boldOn = !boldOn;
 539             } else if ("''".equals(wikiTokenizer.token())) {
 540                 if (!italicOn) {
 541                     builder.append("<em>");
 542                 } else {
 543                     builder.append("</em>");
 544                 }
 545                 italicOn = !italicOn;
 546             } else {
 547                 assert false;
 548             }
 549         }
 550
 551     }
 552
 553 }