]> gitweb.fperrin.net Git - DictionaryPC.git/blobdiff - src/com/hughes/android/dictionary/parser/wiktionary/WholeSectionToHtmlParser.java
Added simple parsing logic for DE and IT wiktionaries.
[DictionaryPC.git] / src / com / hughes / android / dictionary / parser / wiktionary / WholeSectionToHtmlParser.java
index 8cc1ad2fd178248f53402427af71df35e773557d..b3249d12f395a81fb0393f4319b5d0a26104426f 100644 (file)
@@ -1,11 +1,14 @@
 
 package com.hughes.android.dictionary.parser.wiktionary;
 
+import com.hughes.android.dictionary.engine.EntryTypeName;
 import com.hughes.android.dictionary.engine.HtmlEntry;
 import com.hughes.android.dictionary.engine.IndexBuilder;
 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;
 
@@ -16,20 +19,181 @@ import java.util.Map;
 import java.util.regex.Pattern;
 
 public class WholeSectionToHtmlParser extends AbstractWiktionaryParser {
-    
+
+    public static final String NAME = "WholeSectionToHtmlParser";
+
     interface LangConfig {
         boolean skipSection(final String name);
+        EntryTypeName sectionNameToEntryType(String sectionName);
         boolean skipWikiLink(final WikiTokenizer wikiTokenizer);
+        String adjustWikiLink(String wikiLinkDest, final String wikiLinkText);
+        void addFunctionCallbacks(
+                Map<String, FunctionCallback<WholeSectionToHtmlParser>> functionCallbacks);
     }
     static final Map<String,LangConfig> isoToLangConfig = new LinkedHashMap<String,LangConfig>();
     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")) {
+                    return EntryTypeName.SYNONYM_MULTI;
+                }
+                if (sectionName.equalsIgnoreCase("Antonyms")) {
+                    return EntryTypeName.ANTONYM_MULTI;
+                }
+                if (EnParser.partOfSpeechHeader.matcher(sectionName).matches()) {
+                    // We need to put it in the other index, too (probably)
+                    return null;
+                }
+                if (sectionName.equalsIgnoreCase("Derived Terms")) {
+                    return null;
+                }
+                return null;
+            }
+            
+            @Override
+            public boolean skipWikiLink(WikiTokenizer wikiTokenizer) {
+                final String wikiText = wikiTokenizer.wikiLinkText();
+                if (wikiText.startsWith("Category:")) {
+                    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<String, FunctionCallback<WholeSectionToHtmlParser>> functionCallbacks) {
+                EnFunctionCallbacks.addGenericCallbacks(functionCallbacks);
+            }
+        });
+        
+        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("???Category:")) {
+                    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<String, FunctionCallback<WholeSectionToHtmlParser>> 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("???Category:")) {
+                    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<String, FunctionCallback<WholeSectionToHtmlParser>> functionCallbacks) {
+                ItFunctionCallbacks.addGenericCallbacks(functionCallbacks);
+            }
+        });
+
+
+        
+        final LangConfig basicLangConfig = new LangConfig() {
+            @Override
+            public boolean skipSection(String headingText) {
+                return false;
+            }
+            @Override
+            public EntryTypeName sectionNameToEntryType(String sectionName) {
+                return EntryTypeName.WIKTIONARY_MENTIONED;
+            }
             @Override
             public boolean skipWikiLink(WikiTokenizer wikiTokenizer) {
                 final String wikiText = wikiTokenizer.wikiLinkText();
@@ -37,46 +201,93 @@ public class WholeSectionToHtmlParser extends AbstractWiktionaryParser {
                     return true;
                 }
                 return false;
-            }});
-    }
+            }
+            @Override
+            public String adjustWikiLink(String wikiLinkDest, final String wikiLinkText) {
+                return wikiLinkDest;
+            }
 
-    public static final String NAME = "WholeSectionToHtmlParser";
+            @Override
+            public void addFunctionCallbacks(
+                    Map<String, FunctionCallback<WholeSectionToHtmlParser>> functionCallbacks) {
+            }
+        };
+        isoToLangConfig.put("FR", basicLangConfig);
+    }
 
     final IndexBuilder titleIndexBuilder;
+    final IndexBuilder defIndexBuilder;
+    final String skipLangIso;
     final LangConfig langConfig;
+    final String webUrlTemplate;
+    
 
-    public WholeSectionToHtmlParser(final IndexBuilder titleIndexBuilder, final String wiktionaryIso) {
+    public WholeSectionToHtmlParser(final IndexBuilder titleIndexBuilder, final IndexBuilder defIndexBuilder, final String wiktionaryIso, final String skipLangIso,
+            final String webUrlTemplate) {
         this.titleIndexBuilder = titleIndexBuilder;
+        this.defIndexBuilder = defIndexBuilder;
         assert isoToLangConfig.containsKey(wiktionaryIso): wiktionaryIso;
         this.langConfig = isoToLangConfig.get(wiktionaryIso);
+        this.skipLangIso = skipLangIso;
+        this.webUrlTemplate = webUrlTemplate;
     }
+    
+    IndexedEntry indexedEntry = null;
 
     @Override
-    void parseSection(String heading, String text) {
-        HtmlEntry htmlEntry = new HtmlEntry(entrySource, StringEscapeUtils.escapeHtml3(title));
-        IndexedEntry indexedEntry = new IndexedEntry(htmlEntry);
+    public void parseSection(String heading, String text) {
+        assert entrySource != null;
+        final HtmlEntry htmlEntry = new HtmlEntry(entrySource, title);
+        indexedEntry = new IndexedEntry(htmlEntry);
 
         final AppendAndIndexWikiCallback<WholeSectionToHtmlParser> callback = new AppendCallback(
                 this);
+        langConfig.addFunctionCallbacks(callback.functionCallbacks);
 
         callback.builder = new StringBuilder();
         callback.indexedEntry = indexedEntry;
         callback.dispatch(text, null);
 
+        if (webUrlTemplate != null) {
+            final String webUrl = String.format(webUrlTemplate, title);
+            callback.builder.append(String.format("<p> <a href=\"%s\">%s</a>", Uri.escapeDisallowedChars(webUrl), escapeHtmlLiteral(webUrl)));
+        }
         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<String, String> namedArgs) {
     }
+    
+    @Override
+    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);
+        }
+
+    }
+
+
 
     class AppendCallback extends AppendAndIndexWikiCallback<WholeSectionToHtmlParser> {
         public AppendCallback(WholeSectionToHtmlParser parser) {
@@ -85,7 +296,7 @@ public class WholeSectionToHtmlParser extends AbstractWiktionaryParser {
 
         @Override
         public void onPlainText(String plainText) {
-            super.onPlainText(StringEscapeUtils.escapeHtml3(plainText));
+            super.onPlainText(escapeHtmlLiteral(plainText));
         }
 
         @Override
@@ -97,12 +308,31 @@ public class WholeSectionToHtmlParser extends AbstractWiktionaryParser {
             if (langConfig.skipWikiLink(wikiTokenizer)) {
                 return;
             }
-            super.onWikiLink(wikiTokenizer);
+            String linkDest;
+            if (wikiTokenizer.wikiLinkDest() != null) {
+                linkDest = langConfig.adjustWikiLink(wikiTokenizer.wikiLinkDest(), wikiTokenizer.wikiLinkText());
+            } else {
+                linkDest = wikiTokenizer.wikiLinkText();
+            }
+            if (sectionEntryTypeName != null) {
+                // TODO: inside a definition, this could be the wrong language.
+                titleIndexBuilder.addEntryWithString(indexedEntry, wikiTokenizer.wikiLinkText(), sectionEntryTypeName);
+            }
+            if (!StringUtil.isNullOrEmpty(linkDest)) {
+                builder.append(String.format("<a href=\"%s\">", HtmlEntry.formatQuickdicUrl("", linkDest)));
+                super.onWikiLink(wikiTokenizer);
+                builder.append(String.format("</a>"));
+            } else {
+                super.onWikiLink(wikiTokenizer);
+            }
         }
 
         @Override
         public void onFunction(WikiTokenizer wikiTokenizer, String name,
                 List<String> args, Map<String, String> namedArgs) {
+            if (skipLangIso.equalsIgnoreCase(namedArgs.get("lang"))) {
+                namedArgs.remove("lang");
+            }
             super.onFunction(wikiTokenizer, name, args, namedArgs);
         }
 
@@ -114,10 +344,14 @@ public class WholeSectionToHtmlParser extends AbstractWiktionaryParser {
         @Override
         public void onNewline(WikiTokenizer wikiTokenizer) {
         }
+        
+        EntryTypeName sectionEntryTypeName;
+        IndexBuilder currentIndexBuilder;
 
         @Override
         public void onHeading(WikiTokenizer wikiTokenizer) {
             final String headingText = wikiTokenizer.headingWikiText();
+            sectionEntryTypeName = langConfig.sectionNameToEntryType(headingText);
             final int depth = wikiTokenizer.headingDepth();
             if (langConfig.skipSection(headingText)) {
                 while ((wikiTokenizer = wikiTokenizer.nextToken()) != null) {