X-Git-Url: http://git.veekun.com/zzz-pokedex.git/blobdiff_plain/e50aca7ae94c04b5cec1baacef6e0a115a0305e0..9a3152efb04b1c90fcbe786882775ed5984f24ce:/pokedex/roomaji.py?ds=sidebyside

diff --git a/pokedex/roomaji.py b/pokedex/roomaji.py
index 9db6b07..e91c960 100644
--- a/pokedex/roomaji.py
+++ b/pokedex/roomaji.py
@@ -60,7 +60,7 @@ _roomaji_small_kana_combos = {
     # particularly weird, but it seems to be what GF intends
 
     # Simple vowel replacement
-                    u'ウィ': 'wi',  u'ウゥ': 'wu',  u'ウェ': 'we',
+                    u'ウィ': 'wi',  u'ウゥ': 'wu',  u'ウェ': 'we',  u'ウォ': 'wo',
     u'ヴァ': 'va',  u'ヴィ': 'vi',                  u'ヴェ': 've',  u'ヴォ': 'vo',
                                                     u'チェ': 'che',
                                                     u'シェ': 'she',
@@ -71,6 +71,7 @@ _roomaji_small_kana_combos = {
 
     # Not so much
     u'シィ': 'sy',
+    u'ミィ': 'my',
     u'ビィ': 'by',
     u'ピィ': 'py',
 }
@@ -85,11 +86,12 @@ def romanize(string):
     last_char = None  # Used for small kana combos
     for char in string:
         # Full-width Latin
-        if ord(char) >= 0xff11 and ord(char) <= 0xff5e:
+        if 0xff01 <= ord(char) <= 0xff5e:
             if last_kana == 'sokuon':
                 raise ValueError("Sokuon cannot precede Latin characters.")
 
-            char = chr(ord(char) - 0xff11 + 0x31)
+            # XXX Real Unicode decomposition would be nicer
+            char = chr(ord(char) - 0xff01 + 0x21)
             characters.append(char)
 
             last_kana = None