Use KeyError to signal that the name does not resemble any unicode, this pattern is also used in the rest of pdfminer.six

pdfminer/encodingdb.py

@@ -10,12 +10,15 @@ from .psparser import PSLiteral
 HEXADECIMAL = re.compile(r'[0-9a-fA-F]+')
 
 
-def name2unicode(name: str):
+def name2unicode(name):
     """Converts Adobe glyph names to Unicode numbers.
 
+    In contrast to the specification, this raises a KeyError instead of return an empty string when the key is unknown.
+    This way the caller must explicitly define what to do when there is not a match.
+
     Reference: https://github.com/adobe-type-tools/agl-specification#2-the-mapping
 
-    :returns unicode character if name resembles something, empty string if not
+    :returns unicode character if name resembles something, otherwise a KeyError
     """
     full_stop = u'\u002E'
     name = name.split(full_stop)[0]
@@ -33,7 +36,7 @@ def name2unicode(name: str):
             if HEXADECIMAL.match(name_without_uni) and len(name_without_uni) % 4 == 0:
                 unicode_digits = [int(name_without_uni[i:i + 4], base=16) for i in range(0, len(name_without_uni), 4)]
                 if any([55295 < digit < 57344 for digit in unicode_digits]):
-                    return ''
+                    raise KeyError
                 characters = map(six.unichr, unicode_digits)
                 return ''.join(characters)
 
@@ -42,10 +45,10 @@ def name2unicode(name: str):
             if HEXADECIMAL.match(name_without_u) and 4 <= len(name_without_u) <= 6:
                 unicode_digit = int(name_without_u, base=16)
                 if 55295 < unicode_digit < 57344:
-                    return ''
+                    raise KeyError
                 return six.unichr(unicode_digit)
 
-    return ''
+    raise KeyError
 
 
 class EncodingDB(object):

tests/test_encodingdb.py

@@ -4,6 +4,8 @@ Tests based on the Adobe Glyph List Specification (https://github.com/adobe-type
 While not in the specification, lowercase unicode often occurs in pdf's. Therefore lowercase unittest variants are
 added.
 """
+from nose.tools import assert_raises
+
 from pdfminer.encodingdb import name2unicode
 
 
@@ -48,7 +50,7 @@ def test_name2unicode_uni_empty_string_long():
     expressed as D801 DC0C in UTF-16, specifically U+1040C. This character can be correctly mapped by using the
     glyph name "u1040C.
     """
-    assert u'' == name2unicode('uniD801DC0C')
+    assert_raises(KeyError, name2unicode, 'uniD801DC0C')
 
 
 def test_name2unicode_uni_empty_string_long_lowercase():
@@ -57,7 +59,7 @@ def test_name2unicode_uni_empty_string_long_lowercase():
     Neither D801 nor DC0C are in the appropriate set. This form cannot be used to map to the character which is
     expressed as D801 DC0C in UTF-16, specifically U+1040C. This character can be correctly mapped by using the
     glyph name "u1040C."""
-    assert u'' == name2unicode('uniD801DC0C')
+    assert_raises(KeyError, name2unicode, 'uniD801DC0C')
 
 
 def test_name2unicode_uni_pua():
@@ -102,12 +104,12 @@ def test_name2unicode_multiple_components_lowercase():
 
 def test_name2unicode_foo():
     """The name 'foo' maps to an empty string, because 'foo' is not in AGL, and because it does not start with a 'u.'"""
-    assert u'' == name2unicode('foo')
+    assert_raises(KeyError, name2unicode, 'foo')
 
 
 def test_name2unicode_notdef():
     """The name ".notdef" is reduced to an empty string (step 1) and mapped to an empty string (step 3)"""
-    assert u'' == name2unicode('.notdef')
+    assert_raises(KeyError, name2unicode, '.notdef')
 
 
 def test_name2unicode_pua_ogoneksmall():
@@ -116,4 +118,4 @@ def test_name2unicode_pua_ogoneksmall():
 
 
 def test_name2unicode_overflow_error():
-    name2unicode('226215240241240240240240')
+    assert_raises(KeyError, name2unicode, '226215240241240240240240')