]> gitweb.fperrin.net Git - DictionaryPC.git/blobdiff - src/com/hughes/android/dictionary/engine/WiktionarySplitter.java
Add script to download data.
[DictionaryPC.git] / src / com / hughes / android / dictionary / engine / WiktionarySplitter.java
index 6dd043aa9ef69a8e38bd4a04e4984c858dfd9bd2..277dac18419eac3893861428e07e55ba17a76da1 100644 (file)
@@ -56,7 +56,7 @@ public class WiktionarySplitter extends org.xml.sax.helpers.DefaultHandler {
     
     public Selector(final String filename, final String pattern) throws IOException {
       this.out = new DataOutputStream(new BufferedOutputStream(new FileOutputStream(filename)));
-      this.pattern = Pattern.compile(pattern);
+      this.pattern = Pattern.compile(pattern, Pattern.CASE_INSENSITIVE);
     }
   }
 
@@ -82,16 +82,18 @@ public class WiktionarySplitter extends org.xml.sax.helpers.DefaultHandler {
           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/HR.data", ".*[Cc]roatian.*"),
           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/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]inish.*"),
+          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/HE.data", ".*[Hh]ewbrew.*"),
+          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.*"),
@@ -106,19 +108,19 @@ public class WiktionarySplitter extends org.xml.sax.helpers.DefaultHandler {
           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/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/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.*"),
@@ -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();