X-Git-Url: http://gitweb.fperrin.net/?a=blobdiff_plain;f=src%2Fcom%2Fhughes%2Fandroid%2Fdictionary%2Fparser%2Fwiktionary%2FWholeSectionToHtmlParser.java;h=63e507ee9a97d9433efbc863f094480c5175dc96;hb=2182783b7ac6a22c23b37db4ba458ff12a6978dc;hp=c2ddd56d1d3991dc11a414c45adf0d6c10bd7d29;hpb=752c81f1f8ec816865611477896093d81a77f90a;p=DictionaryPC.git diff --git a/src/com/hughes/android/dictionary/parser/wiktionary/WholeSectionToHtmlParser.java b/src/com/hughes/android/dictionary/parser/wiktionary/WholeSectionToHtmlParser.java index c2ddd56..63e507e 100644 --- a/src/com/hughes/android/dictionary/parser/wiktionary/WholeSectionToHtmlParser.java +++ b/src/com/hughes/android/dictionary/parser/wiktionary/WholeSectionToHtmlParser.java @@ -8,10 +8,10 @@ import com.hughes.android.dictionary.engine.IndexBuilder.TokenData; import com.hughes.android.dictionary.engine.IndexedEntry; import com.hughes.android.dictionary.parser.WikiTokenizer; import com.hughes.util.StringUtil; -import com.sun.xml.internal.rngom.util.Uri; -import org.apache.commons.lang3.StringEscapeUtils; +import org.apache.commons.text.StringEscapeUtils; +import java.net.URI; import java.util.ArrayList; import java.util.LinkedHashMap; import java.util.List; @@ -28,17 +28,17 @@ public class WholeSectionToHtmlParser extends AbstractWiktionaryParser { boolean skipWikiLink(final WikiTokenizer wikiTokenizer); String adjustWikiLink(String wikiLinkDest, final String wikiLinkText); void addFunctionCallbacks( - Map> functionCallbacks); + Map> functionCallbacks); } static final Map isoToLangConfig = new LinkedHashMap(); static { - final Pattern enSkipSections = Pattern.compile(".*Translations|Anagrams|References.*"); + final Pattern enSkipSections = Pattern.compile(".*(Translations|Anagrams|References).*"); isoToLangConfig.put("EN", new LangConfig() { @Override public boolean skipSection(String headingText) { return enSkipSections.matcher(headingText).matches(); } - + @Override public EntryTypeName sectionNameToEntryType(String sectionName) { if (sectionName.equalsIgnoreCase("Synonyms")) { @@ -48,7 +48,7 @@ public class WholeSectionToHtmlParser extends AbstractWiktionaryParser { return EntryTypeName.ANTONYM_MULTI; } if (EnParser.partOfSpeechHeader.matcher(sectionName).matches()) { - // We need to put it in the other index, too. + // We need to put it in the other index, too (probably) return null; } if (sectionName.equalsIgnoreCase("Derived Terms")) { @@ -56,7 +56,7 @@ public class WholeSectionToHtmlParser extends AbstractWiktionaryParser { } return null; } - + @Override public boolean skipWikiLink(WikiTokenizer wikiTokenizer) { final String wikiText = wikiTokenizer.wikiLinkText(); @@ -82,41 +82,251 @@ public class WholeSectionToHtmlParser extends AbstractWiktionaryParser { @Override public void addFunctionCallbacks( - Map> functionCallbacks) { + Map> functionCallbacks) { EnFunctionCallbacks.addGenericCallbacks(functionCallbacks); } }); - - final LangConfig basicLangConfig = new LangConfig() { + + final Pattern esSkipSections = Pattern.compile(".*(Traducciones|Locuciones).*"); + isoToLangConfig.put("ES", new LangConfig() { @Override public boolean skipSection(String headingText) { + return esSkipSections.matcher(headingText).matches(); + } + + @Override + public EntryTypeName sectionNameToEntryType(String sectionName) { + if (sectionName.equalsIgnoreCase("sinónimo") || sectionName.equalsIgnoreCase("sinónimos")) { + return EntryTypeName.SYNONYM_MULTI; + } + if (sectionName.equalsIgnoreCase("antónimo") || sectionName.equalsIgnoreCase("antónimos")) { + return EntryTypeName.ANTONYM_MULTI; + } + return null; + } + + @Override + public boolean skipWikiLink(WikiTokenizer wikiTokenizer) { + final String wikiText = wikiTokenizer.wikiLinkText(); + if (wikiText.startsWith("Categoría:")) { + return true; + } return false; } + @Override + public String adjustWikiLink(String wikiLinkDest, String wikiLinkText) { + if (wikiLinkDest.startsWith("w:") || wikiLinkDest.startsWith("Image:")) { + return null; + } + final int hashPos = wikiLinkDest.indexOf("#"); + if (hashPos != -1) { + wikiLinkDest = wikiLinkDest.substring(0, hashPos); + if (wikiLinkDest.isEmpty()) { + wikiLinkDest = wikiLinkText; + } + } + return wikiLinkDest; + } + + @Override + public void addFunctionCallbacks( + Map> functionCallbacks) { + // TODO: need Spanish variant + } + }); + + final Pattern ptSkipSections = Pattern.compile(".*Tradução.*"); + isoToLangConfig.put("PT", new LangConfig() { + @Override + public boolean skipSection(String headingText) { + return esSkipSections.matcher(headingText).matches(); + } + @Override public EntryTypeName sectionNameToEntryType(String sectionName) { - return EntryTypeName.WIKTIONARY_MENTIONED; + if (sectionName.equalsIgnoreCase("Sinônimo") || sectionName.equalsIgnoreCase("Sinônimos")) { + return EntryTypeName.SYNONYM_MULTI; + } + if (sectionName.equalsIgnoreCase("Antônimo") || sectionName.equalsIgnoreCase("Antônimos")) { + return EntryTypeName.ANTONYM_MULTI; + } + return null; } + @Override public boolean skipWikiLink(WikiTokenizer wikiTokenizer) { final String wikiText = wikiTokenizer.wikiLinkText(); - if (wikiText.startsWith("Category:")) { + if (wikiText.startsWith("Categoria:")) { return true; } return false; } @Override - public String adjustWikiLink(String wikiLinkDest, final String wikiLinkText) { + public String adjustWikiLink(String wikiLinkDest, String wikiLinkText) { + if (wikiLinkDest.startsWith("w:") || wikiLinkDest.startsWith("Image:")) { + return null; + } + final int hashPos = wikiLinkDest.indexOf("#"); + if (hashPos != -1) { + wikiLinkDest = wikiLinkDest.substring(0, hashPos); + if (wikiLinkDest.isEmpty()) { + wikiLinkDest = wikiLinkText; + } + } return wikiLinkDest; } @Override public void addFunctionCallbacks( - Map> functionCallbacks) { + Map> functionCallbacks) { + // TODO: need Portuguese variant + } + }); + + final Pattern deSkipSections = Pattern.compile(".*(Übersetzungen|Referenzen|Quellen).*"); + isoToLangConfig.put("DE", new LangConfig() { + @Override + public boolean skipSection(String headingText) { + return deSkipSections.matcher(headingText).matches(); + } + + @Override + public EntryTypeName sectionNameToEntryType(String sectionName) { + if (sectionName.equalsIgnoreCase("Synonyme")) { + return EntryTypeName.SYNONYM_MULTI; + } + if (sectionName.equalsIgnoreCase("Gegenwörter")) { + return EntryTypeName.ANTONYM_MULTI; + } + return null; + } + + @Override + public boolean skipWikiLink(WikiTokenizer wikiTokenizer) { + final String wikiText = wikiTokenizer.wikiLinkText(); + if (wikiText.startsWith("Kategorie:")) { + return true; + } + return false; } - }; - isoToLangConfig.put("FR", basicLangConfig); - isoToLangConfig.put("DE", basicLangConfig); - isoToLangConfig.put("IT", basicLangConfig); + @Override + public String adjustWikiLink(String wikiLinkDest, String wikiLinkText) { + if (wikiLinkDest.startsWith("w:") || wikiLinkDest.startsWith("Image:")) { + return null; + } + final int hashPos = wikiLinkDest.indexOf("#"); + if (hashPos != -1) { + wikiLinkDest = wikiLinkDest.substring(0, hashPos); + if (wikiLinkDest.isEmpty()) { + wikiLinkDest = wikiLinkText; + } + } + return wikiLinkDest; + } + + @Override + public void addFunctionCallbacks( + Map> functionCallbacks) { + DeFunctionCallbacks.addGenericCallbacks(functionCallbacks); + } + }); + + final Pattern itSkipSections = Pattern.compile(".*(Traduzione|Note / Riferimenti).*"); + isoToLangConfig.put("IT", new LangConfig() { + @Override + public boolean skipSection(String headingText) { + return itSkipSections.matcher(headingText).matches(); + } + + @Override + public EntryTypeName sectionNameToEntryType(String sectionName) { + if (sectionName.equalsIgnoreCase("Sinonimi")) { + return EntryTypeName.SYNONYM_MULTI; + } + if (sectionName.equalsIgnoreCase("Antonimi/Contrari")) { + return EntryTypeName.ANTONYM_MULTI; + } + return null; + } + + @Override + public boolean skipWikiLink(WikiTokenizer wikiTokenizer) { + final String wikiText = wikiTokenizer.wikiLinkText(); + if (wikiText.startsWith("Categoria:")) { + return true; + } + return false; + } + @Override + public String adjustWikiLink(String wikiLinkDest, String wikiLinkText) { + if (wikiLinkDest.startsWith("w:") || wikiLinkDest.startsWith("Image:")) { + return null; + } + final int hashPos = wikiLinkDest.indexOf("#"); + if (hashPos != -1) { + wikiLinkDest = wikiLinkDest.substring(0, hashPos); + if (wikiLinkDest.isEmpty()) { + wikiLinkDest = wikiLinkText; + } + } + return wikiLinkDest; + } + + @Override + public void addFunctionCallbacks( + Map> functionCallbacks) { + ItFunctionCallbacks.addGenericCallbacks(functionCallbacks); + } + }); + + + final Pattern frSkipSections = Pattern.compile(".*([Tt]raductions|[Aa]nagrammes).*"); + isoToLangConfig.put("FR", new LangConfig() { + @Override + public boolean skipSection(String headingText) { + return frSkipSections.matcher(headingText).matches(); + } + + @Override + public EntryTypeName sectionNameToEntryType(String sectionName) { + if (sectionName.equalsIgnoreCase("Synonymes")) { + return EntryTypeName.SYNONYM_MULTI; + } + if (sectionName.equalsIgnoreCase("Antonymes")) { + return EntryTypeName.ANTONYM_MULTI; + } + return null; + } + + @Override + public boolean skipWikiLink(WikiTokenizer wikiTokenizer) { + final String wikiText = wikiTokenizer.wikiLinkText(); + if (wikiText.startsWith("Catégorie:")) { + return true; + } + return false; + } + @Override + public String adjustWikiLink(String wikiLinkDest, String wikiLinkText) { + if (wikiLinkDest.startsWith("w:") || wikiLinkDest.startsWith("Image:")) { + return null; + } + final int hashPos = wikiLinkDest.indexOf("#"); + if (hashPos != -1) { + wikiLinkDest = wikiLinkDest.substring(0, hashPos); + if (wikiLinkDest.isEmpty()) { + wikiLinkDest = wikiLinkText; + } + } + return wikiLinkDest; + } + + @Override + public void addFunctionCallbacks( + Map> functionCallbacks) { + FrFunctionCallbacks.addGenericCallbacks(functionCallbacks); + } + }); } final IndexBuilder titleIndexBuilder; @@ -124,10 +334,10 @@ public class WholeSectionToHtmlParser extends AbstractWiktionaryParser { final String skipLangIso; final LangConfig langConfig; final String webUrlTemplate; - + public WholeSectionToHtmlParser(final IndexBuilder titleIndexBuilder, final IndexBuilder defIndexBuilder, final String wiktionaryIso, final String skipLangIso, - final String webUrlTemplate) { + final String webUrlTemplate) { this.titleIndexBuilder = titleIndexBuilder; this.defIndexBuilder = defIndexBuilder; assert isoToLangConfig.containsKey(wiktionaryIso): wiktionaryIso; @@ -135,7 +345,7 @@ public class WholeSectionToHtmlParser extends AbstractWiktionaryParser { this.skipLangIso = skipLangIso; this.webUrlTemplate = webUrlTemplate; } - + IndexedEntry indexedEntry = null; @Override @@ -145,7 +355,7 @@ public class WholeSectionToHtmlParser extends AbstractWiktionaryParser { indexedEntry = new IndexedEntry(htmlEntry); final AppendAndIndexWikiCallback callback = new AppendCallback( - this); + this); langConfig.addFunctionCallbacks(callback.functionCallbacks); callback.builder = new StringBuilder(); @@ -154,36 +364,43 @@ public class WholeSectionToHtmlParser extends AbstractWiktionaryParser { if (webUrlTemplate != null) { final String webUrl = String.format(webUrlTemplate, title); - callback.builder.append(String.format("

%s", Uri.escapeDisallowedChars(webUrl), escapeHtmlLiteral(webUrl))); + // URI.create can raise an exception e.g. if webUrl contains %, just ignore those cases. + try { + callback.builder.append(String.format("

%s", URI.create(webUrl).toASCIIString(), escapeHtmlLiteral(webUrl))); + } catch (Exception e) { + } } htmlEntry.html = callback.builder.toString(); indexedEntry.isValid = true; final TokenData tokenData = titleIndexBuilder.getOrCreateTokenData(title); + tokenData.hasMainEntry = true; htmlEntry.addToDictionary(titleIndexBuilder.index.dict); tokenData.htmlEntries.add(htmlEntry); // titleIndexBuilder.addEntryWithString(indexedEntry, title, // EntryTypeName.WIKTIONARY_TITLE_MULTI_DETAIL); - + indexedEntry = null; } @Override void removeUselessArgs(Map namedArgs) { } - + @Override - public void addLinkToCurrentEntry(String token, EntryTypeName entryTypeName) { - titleIndexBuilder.addEntryWithString(indexedEntry, token, entryTypeName); + public void addLinkToCurrentEntry(String token, final String lang, EntryTypeName entryTypeName) { + if (lang == null || lang.equals(skipLangIso)) { + titleIndexBuilder.addEntryWithString(indexedEntry, token, entryTypeName); + } } - + public static String escapeHtmlLiteral(final String plainText) { final String htmlEscaped = StringEscapeUtils.escapeHtml3(plainText); if (StringUtil.isAscii(htmlEscaped)) { return htmlEscaped; - } else { - return StringUtil.escapeToPureHtmlUnicode(plainText); + } else { + return StringUtil.escapeUnicodeToPureHtml(plainText); } } @@ -219,7 +436,7 @@ public class WholeSectionToHtmlParser extends AbstractWiktionaryParser { // TODO: inside a definition, this could be the wrong language. titleIndexBuilder.addEntryWithString(indexedEntry, wikiTokenizer.wikiLinkText(), sectionEntryTypeName); } - if (linkDest != null) { + if (!StringUtil.isNullOrEmpty(linkDest)) { builder.append(String.format("", HtmlEntry.formatQuickdicUrl("", linkDest))); super.onWikiLink(wikiTokenizer); builder.append(String.format("")); @@ -230,7 +447,7 @@ public class WholeSectionToHtmlParser extends AbstractWiktionaryParser { @Override public void onFunction(WikiTokenizer wikiTokenizer, String name, - List args, Map namedArgs) { + List args, Map namedArgs) { if (skipLangIso.equalsIgnoreCase(namedArgs.get("lang"))) { namedArgs.remove("lang"); } @@ -245,7 +462,7 @@ public class WholeSectionToHtmlParser extends AbstractWiktionaryParser { @Override public void onNewline(WikiTokenizer wikiTokenizer) { } - + EntryTypeName sectionEntryTypeName; IndexBuilder currentIndexBuilder; @@ -255,10 +472,14 @@ public class WholeSectionToHtmlParser extends AbstractWiktionaryParser { sectionEntryTypeName = langConfig.sectionNameToEntryType(headingText); final int depth = wikiTokenizer.headingDepth(); if (langConfig.skipSection(headingText)) { + //System.out.println("Skipping section:" + headingText); while ((wikiTokenizer = wikiTokenizer.nextToken()) != null) { if (wikiTokenizer.isHeading() && wikiTokenizer.headingDepth() <= depth) { + // System.out.println("Resume on: " + wikiTokenizer.token()); wikiTokenizer.returnToLineStart(); return; + } else { + // System.out.println("Skipped: " + wikiTokenizer.token()); } } return; @@ -278,7 +499,7 @@ public class WholeSectionToHtmlParser extends AbstractWiktionaryParser { final String prefix = wikiTokenizer.listItemPrefix(); while (listPrefixStack.size() < prefix.length()) { builder.append(String.format("<%s>", - WikiTokenizer.getListTag(prefix.charAt(listPrefixStack.size())))); + WikiTokenizer.getListTag(prefix.charAt(listPrefixStack.size())))); listPrefixStack.add(prefix.charAt(listPrefixStack.size())); } builder.append("

  • ");