Lawrence D'Oliveiro <l...@geek-central.gen.new_zealand> writes: >> AddrObj = PyTuple_GetItem(TheBufferInfo, 0); >> LenObj = PyTuple_GetItem(TheBufferInfo, 1); >> >> the first PyTuple_GetItem succeeds and the second one fails. > > Admittedly, I did take a shortcut here: array.buffer_info returns a tuple of > two items, so I’m not expecting one GetItem to succeed and the other to > fail.
FromArray is a parameter to the function, with no type check to make sure it's really an array. In fact your code allows for the possibility that it doesn't support the buffer_info operation (if I understand the purpose of the null return check after the PyObject_CallMethod) which means it's prepared for the argument to -not- be an array. In that case maybe it's some other object with a "buffer_info" operation that returns a 1-element tuple. If the function is callable from Python code, then that arg type is completely out of the C code's control. Even if it's only callable from C, you're still depending on not one but two different invariants (that the arg is an array, and that array.buffer_info returns a 2-tuple) that are undocumented and unchecked in the function. I cannot agree with your claim that the approach scales. -- http://mail.python.org/mailman/listinfo/python-list
