Fix segfault in Damerau-Levenstein C code.

If one of the characters had a value of 128 or above, this would be
treated as a signed char and would result in an array lookup with a
negative index. The somewhat contrived test case given here --
comparing a space with a non-breaking space -- reproduces the
segmentation fault prior to the fix.

This also makes a Clang warning go away. Thanks, compiler! :-)

cjellyfish/damerau_levenshtein.c

@@ -40,7 +40,7 @@ int damerau_levenshtein_distance(const char *s1, const char *s2)
     for (i = 1; i <= len1; i++) {
         db = 0;
         for (j = 1; j <= len2; j++) {
-            i1 = da[(size_t)(s2[j-1])];
+            i1 = da[(unsigned char)s2[j-1]];
             j1 = db;
 
             if (s1[i - 1] == s2[j - 1]) {
@@ -58,7 +58,7 @@ int damerau_levenshtein_distance(const char *s1, const char *s2)
             dist[((i+1)*cols) + j + 1] = MIN(MIN(d1, d2), MIN(d3, d4));
         }
 
-        da[s1[i-1]] = i;
+        da[(unsigned char)s1[i-1]] = i;
     }
 
     result = dist[((len1+1) * cols) + len2 + 1];

jellyfish/test.py

@@ -59,7 +59,8 @@ class JellyfishTests(object):
                  ("abcd", "acb", 2),
                  ("cape sand recycling ", "edith ann graham", 17),
                  ("jellyifhs", "jellyfish", 2),
-                 ("ifhs", "fish", 2)]
+                 ("ifhs", "fish", 2),
+                 ("Hello, world!", "Hello,\xc2\xa0world!", 2)]
 
         for (s1, s2, value) in cases:
             self.assertEqual(self.jf.damerau_levenshtein_distance(s1, s2), value)