From: Thad Hughes <thad.hughes@gmail.com>
Date: Wed, 4 Jan 2012 16:32:09 +0000 (-0800)
Subject: Example splitting fixes, tokenizer newline handling, Chinese
X-Git-Url: http://gitweb.fperrin.net/?p=DictionaryPC.git;a=commitdiff_plain;h=05a90f5458cdc0a72a8104fac86f1ba4b39a06f1

Example splitting fixes, tokenizer newline handling, Chinese
transliteration unit test.
---

diff --git a/src/com/hughes/android/dictionary/engine/DictionaryBuilderMain.java b/src/com/hughes/android/dictionary/engine/DictionaryBuilderMain.java
index 13ab972..a16fe91 100644
--- a/src/com/hughes/android/dictionary/engine/DictionaryBuilderMain.java
+++ b/src/com/hughes/android/dictionary/engine/DictionaryBuilderMain.java
@@ -70,7 +70,7 @@ public class DictionaryBuilderMain extends TestCase {
     //isoToWikiName.clear();
     boolean go = false;
     for (final String foreignIso : isoToWikiName.keySet()) {
-      if (foreignIso.equals("JA")) {
+      if (foreignIso.equals("SK")) {
         go = true;
       }
       if (!go) {
diff --git a/src/com/hughes/android/dictionary/engine/LanguageTest.java b/src/com/hughes/android/dictionary/engine/LanguageTest.java
index ea28e6f..2d9b6a0 100644
--- a/src/com/hughes/android/dictionary/engine/LanguageTest.java
+++ b/src/com/hughes/android/dictionary/engine/LanguageTest.java
@@ -142,15 +142,15 @@ public class LanguageTest extends TestCase {
     final Language zh = Language.lookup("zh");
     final Transliterator transliterator = Transliterator.createFromRules("", zh.getDefaultNormalizerRules(), Transliterator.FORWARD);
     
-    assertEquals("xie xie", transliterator.transliterate("謝謝"));
-    assertEquals("xie xie", transliterator.transliterate("谢谢"));
+    assertEquals("xiexie", transliterator.transliterate("謝謝"));
+    assertEquals("xiexie", transliterator.transliterate("谢谢"));
 
-    assertEquals("dian nao", transliterator.transliterate("電腦"));
-    assertEquals("dian nao", transliterator.transliterate("电脑"));
-    assertEquals("ji suan ji", transliterator.transliterate("計算機"));
-    assertEquals("ji suan ji", transliterator.transliterate("计算机"));
+    assertEquals("diannao", transliterator.transliterate("電腦"));
+    assertEquals("diannao", transliterator.transliterate("电脑"));
+    assertEquals("jisuanji", transliterator.transliterate("計算機"));
+    assertEquals("jisuanji", transliterator.transliterate("计算机"));
     
-    assertEquals("cheng jiu", transliterator.transliterate("成就"));
+    assertEquals("chengjiu", transliterator.transliterate("成就"));
     
   }
   
diff --git a/src/com/hughes/android/dictionary/parser/WikiTokenizer.java b/src/com/hughes/android/dictionary/parser/WikiTokenizer.java
index b79013d..5ac7d45 100644
--- a/src/com/hughes/android/dictionary/parser/WikiTokenizer.java
+++ b/src/com/hughes/android/dictionary/parser/WikiTokenizer.java
@@ -228,17 +228,17 @@ public final class WikiTokenizer {
     assert isWikiLink();
     // "[[.."
     if (lastUnescapedPipePos != -1) {
-      return wikiText.substring(lastUnescapedPipePos + 1, end - 2);
+      return trimNewlines(wikiText.substring(lastUnescapedPipePos + 1, end - 2));
     }
     assert start + 2 < wikiText.length() && end >= 2: wikiText;
-    return wikiText.substring(start + 2, end - 2);
+    return trimNewlines(wikiText.substring(start + 2, end - 2));
   }
 
   public String wikiLinkDest() {
     assert isWikiLink();
     // "[[.."
     if (firstUnescapedPipePos != -1) {
-      return wikiText.substring(start + 2, firstUnescapedPipePos);
+      return trimNewlines(wikiText.substring(start + 2, firstUnescapedPipePos));
     }
     return null;
   }
diff --git a/src/com/hughes/android/dictionary/parser/enwiktionary/EnWiktionaryXmlParser.java b/src/com/hughes/android/dictionary/parser/enwiktionary/EnWiktionaryXmlParser.java
index c49b1b6..c09bd4f 100644
--- a/src/com/hughes/android/dictionary/parser/enwiktionary/EnWiktionaryXmlParser.java
+++ b/src/com/hughes/android/dictionary/parser/enwiktionary/EnWiktionaryXmlParser.java
@@ -50,6 +50,7 @@ public class EnWiktionaryXmlParser {
       "Preposition|Proper noun|Article|Prepositional phrase|Acronym|" +
       "Abbreviation|Initialism|Contraction|Prefix|Suffix|Symbol|Letter|" +
       "Ligature|Idiom|Phrase|\\{\\{acronym\\}\\}|\\{\\{initialism\\}\\}|" +
+      "\\{\\{abbreviation\\}\\}|" +
       // These are @deprecated:
       "Noun form|Verb form|Adjective form|Nominal phrase|Noun phrase|" +
       "Verb phrase|Transitive verb|Intransitive verb|Reflexive verb|" +
@@ -161,7 +162,7 @@ public class EnWiktionaryXmlParser {
           
         } else if (headerName.equals("Translations")) {
           if (pos == null) {
-            LOG.warning("Translations without POS: " + title);
+            LOG.info("Translations without POS (but using anyway): " + title);
           }
           doTranslations(wikiTokenizer, pos);
         } else if (headerName.equals("Pronunciation")) {
@@ -518,6 +519,8 @@ public class EnWiktionaryXmlParser {
     for (int i = 0; i < listSection.nextPrefixes.size(); ++i) {
       final String nextPrefix = listSection.nextPrefixes.get(i);
       final String nextLine = listSection.nextLines.get(i);
+
+      // TODO: This splitting is not sensitive to wiki code.
       int dash = nextLine.indexOf("&mdash;");
       int mdashLen = 7;
       if (dash == -1) {
@@ -585,7 +588,12 @@ public class EnWiktionaryXmlParser {
     appendAndIndexWikiCallback.reset(builder, indexedEntry);
     appendAndIndexWikiCallback.entryTypeName = EntryTypeName.WIKTIONARY_EXAMPLE;
     appendAndIndexWikiCallback.entryTypeNameSticks = true;
-    appendAndIndexWikiCallback.dispatch(example, indexBuilder, EntryTypeName.WIKTIONARY_EXAMPLE);
+    try {
+      // TODO: this is a hack needed because we don't safely split on the dash.
+      appendAndIndexWikiCallback.dispatch(example, indexBuilder, EntryTypeName.WIKTIONARY_EXAMPLE);
+    } catch (AssertionError e) {
+      return "--";
+    }
     final String result = trim(builder.toString());
     return result.length() > 0 ? result : "--";
   }
diff --git a/src/com/hughes/android/dictionary/parser/enwiktionary/FunctionCallbacksDefault.java b/src/com/hughes/android/dictionary/parser/enwiktionary/FunctionCallbacksDefault.java
index 7452645..7941a97 100644
--- a/src/com/hughes/android/dictionary/parser/enwiktionary/FunctionCallbacksDefault.java
+++ b/src/com/hughes/android/dictionary/parser/enwiktionary/FunctionCallbacksDefault.java
@@ -256,7 +256,11 @@ public final class FunctionCallbacksDefault {
         displayText = ListUtil.get(args, 1, null);
       }
       
-      appendAndIndexWikiCallback.dispatch(displayText, indexBuilder, entryTypeName);
+      if (displayText != null) {
+        appendAndIndexWikiCallback.dispatch(displayText, indexBuilder, entryTypeName);
+      } else {
+        LOG.warning("no display text: " + wikiTokenizer.token());
+      }
       
       final String tr = namedArgs.remove("tr");
       if (tr != null) {
diff --git a/todo.txt b/todo.txt
index 8f7e76b..2016c55 100644
--- a/todo.txt
+++ b/todo.txt
@@ -1,11 +1,9 @@
 For next release:
-refactor wiki parsing.
 "form of" to bottom
 handle examples like "asdf (asdf)"
 random word jump
 multiword find.
 dictionary update.
-{{Arab}}
 ???italian verbs
 
 pronunciation
@@ -60,13 +58,6 @@ Bad filing: under Arab?
   fare {{it-verb}} {{transitive}} :: To do
 
 
-
-
-**** Wiktionary:
-
-in wiktionary
-  futurismo :: futurism () (noun)
-
   
 done:
 {infl}
@@ -80,4 +71,6 @@ always put defs in list...
 ! Check analytics
 ! Upload dics
 font size
+refactor wiki parsing.
+{{Arab}}
   
\ No newline at end of file