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.EntryTypeName;
   5 import com.hughes.android.dictionary.engine.HtmlEntry;
   6 import com.hughes.android.dictionary.engine.IndexBuilder;
   7 import com.hughes.android.dictionary.engine.IndexBuilder.TokenData;
   8 import com.hughes.android.dictionary.engine.IndexedEntry;
   9 import com.hughes.android.dictionary.parser.WikiTokenizer;
  10 import com.hughes.util.StringUtil;
  11
  12 import org.apache.commons.lang3.StringEscapeUtils;
  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     IndexedEntry indexedEntry = null;
  75
  76     @Override
  77     public void parseSection(String heading, String text) {
  78         assert entrySource != null;
  79         final HtmlEntry htmlEntry = new HtmlEntry(entrySource, StringEscapeUtils.escapeHtml3(title));
  80         indexedEntry = new IndexedEntry(htmlEntry);
  81
  82         final AppendAndIndexWikiCallback<WholeSectionToHtmlParser> callback = new AppendCallback(
  83                 this);
  84         langConfig.addFunctionCallbacks(callback.functionCallbacks);
  85
  86         callback.builder = new StringBuilder();
  87         callback.indexedEntry = indexedEntry;
  88         callback.dispatch(text, null);
  89
  90         htmlEntry.html = callback.builder.toString();
  91         indexedEntry.isValid = true;
  92
  93         final TokenData tokenData = titleIndexBuilder.getOrCreateTokenData(title);
  94
  95         htmlEntry.addToDictionary(titleIndexBuilder.index.dict);
  96         tokenData.htmlEntries.add(htmlEntry);
  97         // titleIndexBuilder.addEntryWithString(indexedEntry, title,
  98         // EntryTypeName.WIKTIONARY_TITLE_MULTI_DETAIL);
  99
 100         indexedEntry = null;
 101     }
 102
 103     @Override
 104     void removeUselessArgs(Map<String, String> namedArgs) {
 105     }
 106
 107     @Override
 108     public void addLinkToCurrentEntry(String token, EntryTypeName entryTypeName) {
 109         titleIndexBuilder.addEntryWithString(indexedEntry, token, entryTypeName);
 110     }
 111
 112
 113
 114     static final Pattern ALL_ASCII = Pattern.compile("[\\p{ASCII}]*");
 115
 116     class AppendCallback extends AppendAndIndexWikiCallback<WholeSectionToHtmlParser> {
 117         public AppendCallback(WholeSectionToHtmlParser parser) {
 118             super(parser);
 119         }
 120
 121         @Override
 122         public void onPlainText(String plainText) {
 123             final String htmlEscaped = StringEscapeUtils.escapeHtml3(plainText);
 124             if (ALL_ASCII.matcher(htmlEscaped).matches()) {
 125                 super.onPlainText(htmlEscaped);
 126             } else {
 127                 super.onPlainText(StringUtil.escapeToPureHtmlUnicode(plainText));
 128             }
 129         }
 130
 131         @Override
 132         public void onWikiLink(WikiTokenizer wikiTokenizer) {
 133             if (wikiTokenizer.wikiLinkText().endsWith(":" + title)) {
 134                 // Skips wikilinks like: [[en::dick]]
 135                 return;
 136             }
 137             if (langConfig.skipWikiLink(wikiTokenizer)) {
 138                 return;
 139             }
 140             String linkDest;
 141             if (wikiTokenizer.wikiLinkDest() != null) {
 142                 linkDest = langConfig.adjustWikiLink(wikiTokenizer.wikiLinkDest());
 143             } else {
 144                 linkDest = wikiTokenizer.wikiLinkText();
 145             }
 146             if (linkDest != null) {
 147                 builder.append(String.format("<a href=\"%s\">", linkDest));
 148                 super.onWikiLink(wikiTokenizer);
 149                 builder.append(String.format("</a>"));
 150             } else {
 151                 super.onWikiLink(wikiTokenizer);
 152             }
 153         }
 154
 155         @Override
 156         public void onFunction(WikiTokenizer wikiTokenizer, String name,
 157                 List<String> args, Map<String, String> namedArgs) {
 158             if (skipLangIso.equalsIgnoreCase(namedArgs.get("lang"))) {
 159                 namedArgs.remove("lang");
 160             }
 161             super.onFunction(wikiTokenizer, name, args, namedArgs);
 162         }
 163
 164         @Override
 165         public void onHtml(WikiTokenizer wikiTokenizer) {
 166             super.onHtml(wikiTokenizer);
 167         }
 168
 169         @Override
 170         public void onNewline(WikiTokenizer wikiTokenizer) {
 171         }
 172
 173         @Override
 174         public void onHeading(WikiTokenizer wikiTokenizer) {
 175             final String headingText = wikiTokenizer.headingWikiText();
 176             final int depth = wikiTokenizer.headingDepth();
 177             if (langConfig.skipSection(headingText)) {
 178                 while ((wikiTokenizer = wikiTokenizer.nextToken()) != null) {
 179                     if (wikiTokenizer.isHeading() && wikiTokenizer.headingDepth() <= depth) {
 180                         wikiTokenizer.returnToLineStart();
 181                         return;
 182                     }
 183                 }
 184                 return;
 185             }
 186             builder.append(String.format("\n<h%d>", depth));
 187             dispatch(headingText, null);
 188             builder.append(String.format("</h%d>\n", depth));
 189         }
 190
 191         final List<Character> listPrefixStack = new ArrayList<Character>();
 192
 193         @Override
 194         public void onListItem(WikiTokenizer wikiTokenizer) {
 195             if (builder.length() != 0 && builder.charAt(builder.length() - 1) != '\n') {
 196                 builder.append("\n");
 197             }
 198             final String prefix = wikiTokenizer.listItemPrefix();
 199             while (listPrefixStack.size() < prefix.length()) {
 200                 builder.append(String.format("<%s>",
 201                         WikiTokenizer.getListTag(prefix.charAt(listPrefixStack.size()))));
 202                 listPrefixStack.add(prefix.charAt(listPrefixStack.size()));
 203             }
 204             builder.append("<li>");
 205             dispatch(wikiTokenizer.listItemWikiText(), null);
 206             builder.append("</li>\n");
 207
 208             WikiTokenizer nextToken = wikiTokenizer.nextToken();
 209             boolean returnToLineStart = false;
 210             if (nextToken != null && nextToken.isNewline()) {
 211                 nextToken = nextToken.nextToken();
 212                 returnToLineStart = true;
 213             }
 214             final String nextListHeader;
 215             if (nextToken == null || !nextToken.isListItem()) {
 216                 nextListHeader = "";
 217             } else {
 218                 nextListHeader = nextToken.listItemPrefix();
 219             }
 220             if (returnToLineStart) {
 221                 wikiTokenizer.returnToLineStart();
 222             }
 223             while (listPrefixStack.size() > nextListHeader.length()) {
 224                 final char prefixChar = listPrefixStack.remove(listPrefixStack.size() - 1);
 225                 builder.append(String.format("</%s>\n", WikiTokenizer.getListTag(prefixChar)));
 226             }
 227         }
 228
 229         boolean boldOn = false;
 230         boolean italicOn = false;
 231
 232         @Override
 233         public void onMarkup(WikiTokenizer wikiTokenizer) {
 234             if ("'''".equals(wikiTokenizer.token())) {
 235                 if (!boldOn) {
 236                     builder.append("<b>");
 237                 } else {
 238                     builder.append("</b>");
 239                 }
 240                 boldOn = !boldOn;
 241             } else if ("''".equals(wikiTokenizer.token())) {
 242                 if (!italicOn) {
 243                     builder.append("<em>");
 244                 } else {
 245                     builder.append("</em>");
 246                 }
 247                 italicOn = !italicOn;
 248             } else {
 249                 assert false;
 250             }
 251         }
 252
 253     }
 254
 255 }