Handle independent H-W VOICED SOUND MARK

When independent half-width voiced sound mark exist,
ex. U+FF9E HALFWIDTH KATAKANA VOICED SOUND MARK
It become full-width voiced sound mark
(U+309B) or U+0022 double-quote.

Fixes #115 infinite loop

Signed-off-by: Hiroshi Miura <[email protected]>

src/data/halfkana.utf8

@@ -91,4 +91,5 @@
 「 ｢
 」 ｣
 。 ｡
-、 ､
+、 ､
+\u309B \uFF9E

src/data/hepburnhira.utf8

@@ -248,3 +248,4 @@ wi \U0001b150
 we \U0001b151
 wo \U0001b152
 ;;
+" ゛

src/pykakasi/scripts.py

@@ -152,7 +152,13 @@ def convert_h(self, text):
                 if length > 0:
                     max_len += length
                     x += length
-                    Hstr = Hstr + chr(ord(kstr) - self._diff)
+                    if ord(kstr) == 0x309B:
+                        Hstr = Hstr + kstr
+                    else:
+                        Hstr = Hstr + chr(ord(kstr) - self._diff)
+                else:
+                    max_len += 1
+                    x += 1  # skip unknown character(issue #115)
             else:  # pragma: no cover
                 break
         return (Hstr, max_len)

tests/test_pykakasi_issues.py

@@ -139,3 +139,11 @@ def test_issue114():
     assert result[0]['hepburn'] == 'omotta'
     assert result[2]['hepburn'] == 'itta'
     assert result[4]['hepburn'] == 'itta'
+
+
+def test_issue115():
+    kks = pykakasi.kakasi()
+    result = kks.convert('ﾞっ、')  # \uFF9E
+    assert result[0]['hira'] == '\u309Bっ、'
+    assert result[0]['kana'] == '\uFF9Eッ、'
+    assert result[0]['hepburn'] == '"tsu,'