]> gitweb.fperrin.net Git - DictionaryPC.git/blobdiff - src/com/hughes/android/dictionary/engine/WiktionarySplitter.java
Move test data, fix DictFileParser, fix splitter, fix crash during
[DictionaryPC.git] / src / com / hughes / android / dictionary / engine / WiktionarySplitter.java
index 6dd043aa9ef69a8e38bd4a04e4984c858dfd9bd2..94f7e262b7421c7af26cf0710ee77cee8736c4b1 100644 (file)
@@ -79,50 +79,52 @@ public class WiktionarySplitter extends org.xml.sax.helpers.DefaultHandler {
 
     if (selectors.isEmpty()) {
       selectors.addAll(Arrays.asList(
-          new Selector("../DictionaryData/inputs/enWikiSplit/AF.data", ".*[Aa]frikaans.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/AR.data", ".*[Aa]rabic.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/HY.data", ".*[Aa]rmenian.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/HR.data", ".*[Cc]roation.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/CS.data", ".*[Cc]zech.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/ZH.data", ".*[Mm]andarin|[Cc]hinese.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/NL.data", ".*[Du]utch.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/EN.data", ".*[Ee]nglish.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/FI.data", ".*[Ff]inish.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/FR.data", ".*[Ff]rench.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/DE.data", ".*[Gg]erman.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/EL.data", ".*[Gg]reek.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/HE.data", ".*[Hh]ewbrew.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/HI.data", ".*[Hh]indi.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/IS.data", ".*[Ii]celandic.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/GA.data", ".*[Ii]rish.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/IT.data", ".*[Ii]talian.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/LT.data", ".*[Ll]ithuanian.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/JA.data", ".*[Jj]apanese.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/KO.data", ".*[Kk]orean.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/KU.data", ".*[Kk]urdish.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/MS.data", ".*[Mm]alay.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/MI.data", ".*[Mm]aori.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/MN.data", ".*[Mm]ongolian.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/NO.data", ".*[Nn]orwegian.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/FA.data", ".*[Pp]ersian.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/PT.data", ".*[Pp]ortuguese.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/RO.data", ".*[Rr]omanian.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/RU.data", ".*[Rr]ussian.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/SA.data", ".*[Ss]anskrit.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/SR.data", ".*[Ss]erbian.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/SO.data", ".*[Ss]omali.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/ES.data", ".*[Ss]panish.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/SU.data", ".*[Ss]udanese.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/SV.data", ".*[Ss]wedish.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/TG.data", ".*[Tt]ajik.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/TH.data", ".*[Tt]hai.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/BO.data", ".*[Tt]ibetan.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/TR.data", ".*[Tt]urkish.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/UK.data", ".*[Uu]kranian.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/VI.data", ".*[Vv]ietnamese.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/CI.data", ".*[Ww]elsh.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/YI.data", ".*[Yy]iddish.*"),
-          new Selector("../DictionaryData/inputs/enWikiSplit/ZU.data", ".*[Zz]ulu.*")
+//          new Selector("../DictionaryData/inputs/enWikiSplit/AF.data", ".*[Aa]frikaans.*"),
+//          new Selector("../DictionaryData/inputs/enWikiSplit/AR.data", ".*[Aa]rabic.*"),
+//          new Selector("../DictionaryData/inputs/enWikiSplit/HY.data", ".*[Aa]rmenian.*"),
+//          new Selector("../DictionaryData/inputs/enWikiSplit/HR.data", ".*[Cc]roation.*"),
+//          new Selector("../DictionaryData/inputs/enWikiSplit/CS.data", ".*[Cc]zech.*"),
+          new Selector("../DictionaryData/inputs/enWikiSplit/ZH.data", ".*[Cc]hinese.*|.*[Mm]andarin.*|.*Cantonese.*"),
+          new Selector("../DictionaryData/inputs/enWikiSplit/DA.data", ".*[Dd]anish.*")
+//          new Selector("../DictionaryData/inputs/enWikiSplit/NL.data", ".*[Dd]utch.*"),
+//          new Selector("../DictionaryData/inputs/enWikiSplit/EN.data", ".*[Ee]nglish.*"),
+//          new Selector("../DictionaryData/inputs/enWikiSplit/FI.data", ".*[Ff]innish.*"),
+//          new Selector("../DictionaryData/inputs/enWikiSplit/FR.data", ".*[Ff]rench.*"),
+//          new Selector("../DictionaryData/inputs/enWikiSplit/DE.data", ".*[Gg]erman.*"),
+//          new Selector("../DictionaryData/inputs/enWikiSplit/EL.data", ".*[Gg]reek.*"),
+//          new Selector("../DictionaryData/inputs/enWikiSplit/haw.data", ".*[Hh]awaiian.*"),
+//          new Selector("../DictionaryData/inputs/enWikiSplit/HE.data", ".*[Hh]ebrew.*"),
+//          new Selector("../DictionaryData/inputs/enWikiSplit/HI.data", ".*[Hh]indi.*"),
+//          new Selector("../DictionaryData/inputs/enWikiSplit/IS.data", ".*[Ii]celandic.*"),
+//          new Selector("../DictionaryData/inputs/enWikiSplit/GA.data", ".*[Ii]rish.*"),
+//          new Selector("../DictionaryData/inputs/enWikiSplit/IT.data", ".*[Ii]talian.*"),
+//          new Selector("../DictionaryData/inputs/enWikiSplit/LT.data", ".*[Ll]ithuanian.*"),
+//          new Selector("../DictionaryData/inputs/enWikiSplit/JA.data", ".*[Jj]apanese.*"),
+//          new Selector("../DictionaryData/inputs/enWikiSplit/KO.data", ".*[Kk]orean.*"),
+//          new Selector("../DictionaryData/inputs/enWikiSplit/KU.data", ".*[Kk]urdish.*"),
+//          new Selector("../DictionaryData/inputs/enWikiSplit/MS.data", ".*[Mm]alay.*"),
+//          new Selector("../DictionaryData/inputs/enWikiSplit/MI.data", ".*[Mm]aori.*"),
+//          new Selector("../DictionaryData/inputs/enWikiSplit/MN.data", ".*[Mm]ongolian.*"),
+//          new Selector("../DictionaryData/inputs/enWikiSplit/NO.data", ".*[Nn]orwegian.*"),
+//          new Selector("../DictionaryData/inputs/enWikiSplit/FA.data", ".*[Pp]ersian.*"),
+//          new Selector("../DictionaryData/inputs/enWikiSplit/PT.data", ".*[Pp]ortuguese.*"),
+//          new Selector("../DictionaryData/inputs/enWikiSplit/PL.data", ".*[Pp]olish.*")
+//          new Selector("../DictionaryData/inputs/enWikiSplit/RO.data", ".*[Rr]omanian.*"),
+//          new Selector("../DictionaryData/inputs/enWikiSplit/RU.data", ".*[Rr]ussian.*"),
+//          new Selector("../DictionaryData/inputs/enWikiSplit/SA.data", ".*[Ss]anskrit.*"),
+//          new Selector("../DictionaryData/inputs/enWikiSplit/SR.data", ".*[Ss]erbian.*"),
+//          new Selector("../DictionaryData/inputs/enWikiSplit/SO.data", ".*[Ss]omali.*"),
+//          new Selector("../DictionaryData/inputs/enWikiSplit/ES.data", ".*[Ss]panish.*"),
+//          new Selector("../DictionaryData/inputs/enWikiSplit/SV.data", ".*[Ss]wedish.*"),
+//          new Selector("../DictionaryData/inputs/enWikiSplit/TG.data", ".*[Tt]ajik.*"),
+//          new Selector("../DictionaryData/inputs/enWikiSplit/TH.data", ".*[Tt]hai.*"),
+//          new Selector("../DictionaryData/inputs/enWikiSplit/BO.data", ".*[Tt]ibetan.*"),
+//          new Selector("../DictionaryData/inputs/enWikiSplit/TR.data", ".*[Tt]urkish.*"),
+//          new Selector("../DictionaryData/inputs/enWikiSplit/UK.data", ".*[Uu]krainian.*")
+//          new Selector("../DictionaryData/inputs/enWikiSplit/VI.data", ".*[Vv]ietnamese.*"),
+//          new Selector("../DictionaryData/inputs/enWikiSplit/CI.data", ".*[Ww]elsh.*"),
+//          new Selector("../DictionaryData/inputs/enWikiSplit/YI.data", ".*[Yy]iddish.*"),
+//          new Selector("../DictionaryData/inputs/enWikiSplit/ZU.data", ".*[Zz]ulu.*")
           ));
     }
     
@@ -137,9 +139,12 @@ public class WiktionarySplitter extends org.xml.sax.helpers.DefaultHandler {
 
   static final Pattern headingStart = Pattern.compile("^(=+)[^=]+=+", Pattern.MULTILINE);
   
+  int pageCount = 0;
   private void endPage() {
     final String title = titleBuilder.toString();
-    System.out.println("endPage: " + title);
+    if (++pageCount % 1000 == 0) {
+      System.out.println("endPage: " + title + ", count=" + pageCount);
+    }
     
     String text = textBuilder.toString();