Merge pull request #72 from gaoflow/fix-parse-nth-invalid-crash

Return None instead of crashing on invalid An+B input

Author: liZe

tests/test_tinycss2.py

@@ -153,6 +153,20 @@ def test_nth(input):
     return parse_nth(input)
 
 
+@pytest.mark.parametrize('invalid', ['+', '+/**/', 'n+', 'n +', '-n-', '2n +'])
+def test_nth_invalid_does_not_crash(invalid):
+    # Truncated/invalid An+B fragments must return None per parse_nth's
+    # documented contract, not raise StopIteration or AttributeError.
+    assert parse_nth(invalid) is None
+
+
+def test_nth_leading_plus_whitespace_still_invalid():
+    # The fix for the above must not make '+ n' (whitespace after a leading
+    # '+') accidentally valid: only '+n' is a valid nth expression.
+    assert parse_nth('+n') == (1, 0)
+    assert parse_nth('+ n') is None
+
+
 def _number(value):
     if value is None:
         return 'none'

tinycss2/nth.py

@@ -59,8 +59,10 @@ def parse_nth(input):
             if match:
                 return parse_end(tokens, 1, int(match.group(1)))
     elif token == '+':
-        token = next(tokens)  # Whitespace after an initial '+' is invalid.
-        if token.type == 'ident':
+        # Whitespace after an initial '+' is invalid, so the next token is read
+        # without skipping it. ``None`` is used when the iterator is exhausted.
+        token = next(tokens, None)
+        if token is not None and token.type == 'ident':
             ident = token.lower_value
             if ident == 'n':
                 return parse_b(tokens, 1)
@@ -87,7 +89,7 @@ def parse_b(tokens, a):
 
 def parse_signless_b(tokens, a, b_sign):
     token = _next_significant(tokens)
-    if (token.type == 'number' and token.is_integer and
+    if (token is not None and token.type == 'number' and token.is_integer and
             token.representation[0] not in '-+'):
         return parse_end(tokens, a, b_sign * token.int_value)