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

   1
   2 package com.hughes.android.dictionary.parser.wiktionary;
   3
   4 import com.hughes.android.dictionary.engine.HtmlEntry;
   5 import com.hughes.android.dictionary.engine.IndexBuilder;
   6 import com.hughes.android.dictionary.engine.IndexBuilder.TokenData;
   7 import com.hughes.android.dictionary.engine.IndexedEntry;
   8 import com.hughes.android.dictionary.parser.WikiTokenizer;
   9 import com.hughes.util.StringUtil;
  10
  11 import org.apache.commons.lang3.StringEscapeUtils;
  12 import org.apache.commons.lang3.StringUtils;
  13
  14 import java.util.ArrayList;
  15 import java.util.LinkedHashMap;
  16 import java.util.List;
  17 import java.util.Map;
  18 import java.util.regex.Pattern;
  19
  20 public class WholeSectionToHtmlParser extends AbstractWiktionaryParser {
  21
  22     public static final String NAME = "WholeSectionToHtmlParser";
  23
  24     interface LangConfig {
  25         boolean skipSection(final String name);
  26         boolean skipWikiLink(final WikiTokenizer wikiTokenizer);
  27         String adjustWikiLink(String wikiLinkDest);
  28         void addFunctionCallbacks(
  29                 Map<String, FunctionCallback<WholeSectionToHtmlParser>> functionCallbacks);
  30     }
  31     static final Map<String,LangConfig> isoToLangConfig = new LinkedHashMap<String,LangConfig>();
  32     static {
  33         final Pattern enSkipSections = Pattern.compile(".*Translations|Anagrams|References.*");
  34         isoToLangConfig.put("EN", new LangConfig() {
  35             @Override
  36             public boolean skipSection(String headingText) {
  37                 return enSkipSections.matcher(headingText).matches();
  38             }
  39
  40             @Override
  41             public boolean skipWikiLink(WikiTokenizer wikiTokenizer) {
  42                 final String wikiText = wikiTokenizer.wikiLinkText();
  43                 if (wikiText.startsWith("Category:")) {
  44                     return true;
  45                 }
  46                 return false;
  47             }
  48             @Override
  49             public String adjustWikiLink(String wikiLinkDest) {
  50                 if (wikiLinkDest.startsWith("w:") || wikiLinkDest.startsWith("Image:")) {
  51                     return null;
  52                 }
  53                 return wikiLinkDest;
  54             }
  55
  56             @Override
  57             public void addFunctionCallbacks(
  58                     Map<String, FunctionCallback<WholeSectionToHtmlParser>> functionCallbacks) {
  59                 EnFunctionCallbacks.addGenericCallbacks(functionCallbacks);
  60             }});
  61     }
  62
  63     final IndexBuilder titleIndexBuilder;
  64     final String skipLangIso;
  65     final LangConfig langConfig;
  66
  67     public WholeSectionToHtmlParser(final IndexBuilder titleIndexBuilder, final String wiktionaryIso, final String skipLangIso) {
  68         this.titleIndexBuilder = titleIndexBuilder;
  69         assert isoToLangConfig.containsKey(wiktionaryIso): wiktionaryIso;
  70         this.langConfig = isoToLangConfig.get(wiktionaryIso);
  71         this.skipLangIso = skipLangIso;
  72     }
  73
  74     @Override
  75     void parseSection(String heading, String text) {
  76         HtmlEntry htmlEntry = new HtmlEntry(entrySource, StringEscapeUtils.escapeHtml3(title));
  77         IndexedEntry indexedEntry = new IndexedEntry(htmlEntry);
  78
  79         final AppendAndIndexWikiCallback<WholeSectionToHtmlParser> callback = new AppendCallback(
  80                 this);
  81         langConfig.addFunctionCallbacks(callback.functionCallbacks);
  82
  83         callback.builder = new StringBuilder();
  84         callback.indexedEntry = indexedEntry;
  85         callback.dispatch(text, null);
  86
  87         htmlEntry.html = callback.builder.toString();
  88         indexedEntry.isValid = true;
  89
  90         final TokenData tokenData = titleIndexBuilder.getOrCreateTokenData(title);
  91
  92         htmlEntry.addToDictionary(titleIndexBuilder.index.dict);
  93         tokenData.htmlEntries.add(htmlEntry);
  94         // titleIndexBuilder.addEntryWithString(indexedEntry, title,
  95         // EntryTypeName.WIKTIONARY_TITLE_MULTI_DETAIL);
  96     }
  97
  98     @Override
  99     void removeUselessArgs(Map<String, String> namedArgs) {
 100     }
 101
 102     static final Pattern ALL_ASCII = Pattern.compile("[\\p{ASCII}]*");
 103
 104     class AppendCallback extends AppendAndIndexWikiCallback<WholeSectionToHtmlParser> {
 105         public AppendCallback(WholeSectionToHtmlParser parser) {
 106             super(parser);
 107         }
 108
 109         @Override
 110         public void onPlainText(String plainText) {
 111             final String htmlEscaped = StringEscapeUtils.escapeHtml3(plainText);
 112             if (ALL_ASCII.matcher(htmlEscaped).matches()) {
 113                 super.onPlainText(htmlEscaped);
 114             } else {
 115                 super.onPlainText(StringUtil.escapeToPureHtmlUnicode(plainText));
 116             }
 117         }
 118
 119         @Override
 120         public void onWikiLink(WikiTokenizer wikiTokenizer) {
 121             if (wikiTokenizer.wikiLinkText().endsWith(":" + title)) {
 122                 // Skips wikilinks like: [[en::dick]]
 123                 return;
 124             }
 125             if (langConfig.skipWikiLink(wikiTokenizer)) {
 126                 return;
 127             }
 128             String linkDest;
 129             if (wikiTokenizer.wikiLinkDest() != null) {
 130                 linkDest = langConfig.adjustWikiLink(wikiTokenizer.wikiLinkDest());
 131             } else {
 132                 linkDest = wikiTokenizer.wikiLinkText();
 133             }
 134             if (linkDest != null) {
 135                 builder.append(String.format("<a href=\"%s\">", linkDest));
 136                 super.onWikiLink(wikiTokenizer);
 137                 builder.append(String.format("</a>"));
 138             } else {
 139                 super.onWikiLink(wikiTokenizer);
 140             }
 141         }
 142
 143         @Override
 144         public void onFunction(WikiTokenizer wikiTokenizer, String name,
 145                 List<String> args, Map<String, String> namedArgs) {
 146             if (skipLangIso.equalsIgnoreCase(namedArgs.get("lang"))) {
 147                 namedArgs.remove("lang");
 148             }
 149             super.onFunction(wikiTokenizer, name, args, namedArgs);
 150         }
 151
 152         @Override
 153         public void onHtml(WikiTokenizer wikiTokenizer) {
 154             super.onHtml(wikiTokenizer);
 155         }
 156
 157         @Override
 158         public void onNewline(WikiTokenizer wikiTokenizer) {
 159         }
 160
 161         @Override
 162         public void onHeading(WikiTokenizer wikiTokenizer) {
 163             final String headingText = wikiTokenizer.headingWikiText();
 164             final int depth = wikiTokenizer.headingDepth();
 165             if (langConfig.skipSection(headingText)) {
 166                 while ((wikiTokenizer = wikiTokenizer.nextToken()) != null) {
 167                     if (wikiTokenizer.isHeading() && wikiTokenizer.headingDepth() <= depth) {
 168                         wikiTokenizer.returnToLineStart();
 169                         return;
 170                     }
 171                 }
 172                 return;
 173             }
 174             builder.append(String.format("\n<h%d>", depth));
 175             dispatch(headingText, null);
 176             builder.append(String.format("</h%d>\n", depth));
 177         }
 178
 179         final List<Character> listPrefixStack = new ArrayList<Character>();
 180
 181         @Override
 182         public void onListItem(WikiTokenizer wikiTokenizer) {
 183             if (builder.length() != 0 && builder.charAt(builder.length() - 1) != '\n') {
 184                 builder.append("\n");
 185             }
 186             final String prefix = wikiTokenizer.listItemPrefix();
 187             while (listPrefixStack.size() < prefix.length()) {
 188                 builder.append(String.format("<%s>",
 189                         WikiTokenizer.getListTag(prefix.charAt(listPrefixStack.size()))));
 190                 listPrefixStack.add(prefix.charAt(listPrefixStack.size()));
 191             }
 192             builder.append("<li>");
 193             dispatch(wikiTokenizer.listItemWikiText(), null);
 194             builder.append("</li>\n");
 195
 196             WikiTokenizer nextToken = wikiTokenizer.nextToken();
 197             boolean returnToLineStart = false;
 198             if (nextToken != null && nextToken.isNewline()) {
 199                 nextToken = nextToken.nextToken();
 200                 returnToLineStart = true;
 201             }
 202             final String nextListHeader;
 203             if (nextToken == null || !nextToken.isListItem()) {
 204                 nextListHeader = "";
 205             } else {
 206                 nextListHeader = nextToken.listItemPrefix();
 207             }
 208             if (returnToLineStart) {
 209                 wikiTokenizer.returnToLineStart();
 210             }
 211             while (listPrefixStack.size() > nextListHeader.length()) {
 212                 final char prefixChar = listPrefixStack.remove(listPrefixStack.size() - 1);
 213                 builder.append(String.format("</%s>\n", WikiTokenizer.getListTag(prefixChar)));
 214             }
 215         }
 216
 217         boolean boldOn = false;
 218         boolean italicOn = false;
 219
 220         @Override
 221         public void onMarkup(WikiTokenizer wikiTokenizer) {
 222             if ("'''".equals(wikiTokenizer.token())) {
 223                 if (!boldOn) {
 224                     builder.append("<b>");
 225                 } else {
 226                     builder.append("</b>");
 227                 }
 228                 boldOn = !boldOn;
 229             } else if ("''".equals(wikiTokenizer.token())) {
 230                 if (!italicOn) {
 231                     builder.append("<em>");
 232                 } else {
 233                     builder.append("</em>");
 234                 }
 235                 italicOn = !italicOn;
 236             } else {
 237                 assert false;
 238             }
 239         }
 240
 241     }
 242
 243 }