when indexing 'str', it's safe to infer 'str' for the result

diff --git a/Cython/Compiler/ExprNodes.py b/Cython/Compiler/ExprNodes.py
index 228a5b2fae93e650a9dd72ff44311964c6e9f286..07b0db220fadc1672e25c3390f8300e32e8c690a 100755 (executable)
--- a/Cython/Compiler/ExprNodes.py
+++ b/Cython/Compiler/ExprNodes.py
@@ -2074,6 +2074,9 @@ class IndexNode(ExprNode):
                 # to receive it, throw it away, and potentially rebuild it
                 # on a subsequent PyObject coercion.
                 return PyrexTypes.c_py_ucs4_type
+            elif base_type is str_type:
+                # always returns str - Py2: bytes, Py3: unicode
+                return base_type
             elif isinstance(self.base, BytesNode):
                 #if env.global_scope().context.language_level >= 3:
                 #    # infering 'char' can be made to work in Python 3 mode

diff --git a/tests/run/type_inference.pyx b/tests/run/type_inference.pyx
index e51e43e464b6d55f3c5360ae11914c19c0ce4c96..845bb146bdf46a3a840ea2fad2077a8bf48d32fa 100644 (file)
--- a/tests/run/type_inference.pyx
+++ b/tests/run/type_inference.pyx
@@ -94,7 +94,7 @@ def indexing():
     b = b"abc"
     assert typeof(b) == "bytes object", typeof(b)
     b1 = b[1]
-    assert typeof(b1) == "Python object", typeof(b1)
+    assert typeof(b1) == "Python object", typeof(b1) # Py2: bytes, Py3: int
     u = u"xyz"
     assert typeof(u) == "unicode object", typeof(u)
     u1 = u[1]
@@ -102,7 +102,7 @@ def indexing():
     s = "xyz"
     assert typeof(s) == "str object", typeof(s)
     s1 = s[1]
-    assert typeof(s1) == "Python object", typeof(s1)
+    assert typeof(s1) == "str object", typeof(s1)
     L = [1,2,3]
     assert typeof(L) == "list object", typeof(L)
     L1 = L[1]
@@ -296,6 +296,7 @@ def loop_over_bytes():
     Python object
     """
     cdef bytes bytes_string = b'abcdefg'
+    # bytes in Py2, int in Py3
     for c in bytes_string:
         pass
     return typeof(c)
@@ -303,9 +304,10 @@ def loop_over_bytes():
 def loop_over_str():
     """
     >>> print( loop_over_str() )
-    Python object
+    str object
     """
     cdef str string = 'abcdefg'
+    # str (bytes) in Py2, str (unicode) in Py3
     for c in string:
         pass
     return typeof(c)
@@ -316,6 +318,7 @@ def loop_over_unicode():
     Py_UCS4
     """
     cdef unicode ustring = u'abcdefg'
+    # Py_UCS4 can represent any Unicode character
     for uchar in ustring:
         pass
     return typeof(uchar)