homepage

I was recently experimenting with the new PEP 3155 '__qualname__ implementation and noticed that '__qualname__' is not present on builtin methods and functions:
[meadori@motherbrain cpython]$ ./python 
Python 3.3.0a0 (default:aab45b904141+, Dec 10 2011, 14:53:54) 
[GCC 4.6.2 20111027 (Red Hat 4.6.2-1)] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> max.__qualname__
Traceback (most recent call last):
 File "<stdin>", line 1, in <module>
AttributeError: 'builtin_function_or_method' object has no attribute '__qualname__'
>>> str.strip.__qualname__
Traceback (most recent call last):
 File "<stdin>", line 1, in <module>
AttributeError: 'method_descriptor' object has no attribute '__qualname__'
I will work up a patch.

I already have a patch for the descriptor types which lazily calculates the __qualname__. However test.test_sys also needs fixing because it tests that these types have expected sizes.
I have not got round to builtin_function_or_method though.

Updated patch which fixes test.test_sys.SizeofTest. (It also adds __qualname__ to member descriptors and getset descriptors.)

The patch should add some tests for the added functionality (I'm not sure where, but perhaps test_descr is a good location).
Also, just a nitpick, you can use _PyObject_GetAttrId and the _Py_IDENTIFIER macro instead of interning the "__qualname__" string yourself.
Note that extension (non-builtin) types will need to have their __qualname__ fixed before their methods' __qualname__ is usable:
>>> collections.deque.__qualname__
'deque'

> Note that extension (non-builtin) types will need to have their 
> __qualname__ fixed before their methods' __qualname__ is usable:
>
> >>> collections.deque.__qualname__
> 'deque'
I'm confused. Isn't that the expected behaviour? Since the deque class is not nested inside another class or function, __qualname__ should be the same as __name__, shouldn't it?

> > Note that extension (non-builtin) types will need to have their 
> > __qualname__ fixed before their methods' __qualname__ is usable:
> >
> > >>> collections.deque.__qualname__
> > 'deque'
> 
> I'm confused. Isn't that the expected behaviour? Since the deque
> class is not nested inside another class or function, __qualname__
> should be the same as __name__, shouldn't it?
Uh, yes, my bad.

New version of the patch with tests and using _Py_IDENTIFIER.

Ok, a couple of further (minor) issues:
- I don't think AssertionError is the right exception type. TypeError should be used when a type mismatches (e.g. "not an unicode object");
- you don't need to check for d_type being NULL, since other methods don't;
- if type_qualname == NULL, the original error should be retained.
Otherwise, looks good, thank you.

Patch which add __qualname__ to builtin_function_or_method. Note that I had to make a builtin staticmethod have __self__ be the type instead of None.

> Ok, a couple of further (minor) issues:
> - I don't think AssertionError is the right exception type. TypeError 
> should be used when a type mismatches (e.g. "not an unicode object");
> - you don't need to check for d_type being NULL, since other methods don't;
> - if type_qualname == NULL, the original error should be retained.
Fixed. (I suspect, though, that caching the value of __qualname__ is an unnecessary optimisation.)

New changeset 24238e89f938 by Antoine Pitrou in branch 'default':
Issue #13577: various kinds of descriptors now have a __qualname__ attribute.
http://hg.python.org/cpython/rev/24238e89f938 

About method_qualname.patch:
- apparently you forgot to add BuiltinFunctionPropertiesTest in test_main()?
- a static method keeps a reference to the type: I think it's ok, although I'm not sure about the consequences (Guido, would you have an idea?)
- about "XXX Note that it is not possible to use __qualname__ to distinguish a builtin bound instance method from its unbound equivalent", I don't think it's a problem. People can e.g. check for __self__:
>>> list.append.__self__
Traceback (most recent call last):
 File "<stdin>", line 1, in <module>
AttributeError: 'method_descriptor' object has no attribute '__self__'
>>> list().append.__self__
[]

> - apparently you forgot to add BuiltinFunctionPropertiesTest in 
> test_main()?
Yes. Fixed.
> - a static method keeps a reference to the type: I think it's ok, although 
> I'm not sure about the consequences (Guido, would you have an idea?)
Interestingly, in the stdlib METH_STATIC is only used by (str|bytes|bytearray).maketrans and _multiprocessing.win32.*.

For the most part this looks OK, but I am not sure about this hunk:
diff --git a/Objects/typeobject.c b/Objects/typeobject.c
--- a/Objects/typeobject.c
+++ b/Objects/typeobject.c
@@ -3724,7 +3724,7 @@ add_methods(PyTypeObject *type, PyMethod
 descr = PyDescr_NewClassMethod(type, meth);
 }
 else if (meth->ml_flags & METH_STATIC) {
- PyObject *cfunc = PyCFunction_New(meth, NULL);
+ PyObject *cfunc = PyCFunction_New(meth, (PyObject*)type);
 if (cfunc == NULL)
 return -1;
 descr = PyStaticMethod_New(cfunc);
That may be a breaking change as existing code may rely on the 'None' behavior. In 
fact, this causes a unit test to fail with the patch applied:
[meadori@motherbrain cpython]$ ./python -m test test_descr
[1/1] test_descr
test test_descr failed -- Traceback (most recent call last):
 File "/home/meadori/src/python/cpython/Lib/test/test_descr.py", line 1485, in test_staticmethods_in_c
 self.assertEqual(x, None)
AssertionError: <class 'xxsubtype.spamlist'> != None
sbt, have you been running the test suite before submitting patches? If not, then
please do.

> sbt, have you been running the test suite before submitting patches? 
> If not, then please do.
I ran it after I submitted. Sorry.
Here is another patch. It also makes sure that __self__ is reported as None when METH_STATIC.

Thanks for the patch. Since you modified PyCFunction_GET_SELF to return NULL for static methods, why not simply use it instead of manually looking up the flags in several places?
I'm talking about the following changes:
- PyObject *self = PyCFunction_GET_SELF(func);
+ int flags = PyCFunction_GET_FLAGS(func);
+ PyObject *self = flags & METH_STATIC ? NULL :
+ _PyCFunction_GET_RAW_SELF(func);
[...]
 self = m->m_self;
- if (self == NULL)
+ if (self == NULL || PyCFunction_GET_FLAGS(m) & METH_STATIC)
 self = Py_None;
[...]
- PyObject *self = PyCFunction_GET_SELF(func);
+ PyObject *self = flags & METH_STATIC ? NULL :
+ _PyCFunction_GET_RAW_SELF(func);
Unless you demonstrate there's a significant performance improvement in this style, we should really favour the simpler style.

A simplified patch getting rid of _PyCFunction_GET_RAW_SELF().

New changeset 7a7b698f6f21 by Antoine Pitrou in branch 'default':
Issue #13577: Built-in methods and functions now have a __qualname__.
http://hg.python.org/cpython/rev/7a7b698f6f21 

The patch is committed, thank you!