when I build a PyObject with C API of Python3, its ob_type is NULL

Question

I ran such code in C++11:

PyObject* aa = PyLong_FromLong(7L);

Then I checked the value of aa, it's not NULL, but aa->ob_type was NULL.

However, when I ran:

PyObject* aa = PyLong_FromLong(257L);

aa->ob_type was not NULL anymore. I read the document of PyLong_FromLong, and found this:

PyObject* PyLong_FromLong(long v)
Return value: New reference. 
Return a new PyLongObject object from v, or NULL on failure.

The current implementation keeps an array of integer objects for all integers between -5 and 256, when you create an int in that range you actually just get back a reference to the existing object. So it should be possible to change the value of 1. I suspect the behaviour of Python in this case is undefined. :-)

Seems that when building PyLongObject between -5 and 256 will meet this problem. But I don't understand the reason.

What's more, this problem doesn't appear in Python2. That's incredible!

Related: [“is” operator behaves unexpectedly with integers](http://stackoverflow.com/questions/306313/is-operator-behaves-unexpectedly-with-integers) — Ignacio Vazquez-Abrams, Mar 21 '17 at 05:55
In Python2 there's two integer types. `PyInt` is the smaller type, so I think -5 -> 256 will be defined there and not for `PyLong`. — DavidW, Mar 21 '17 at 10:36

score 1 · Accepted Answer · answered Mar 21 '17 at 12:17

You haven't initialized Python. Since these small objects are special-cased they are set up when Python is initialized.

#include <Python.h>
#include <stdio.h>

int main(int argc, char *argv[])
{
  Py_Initialize();
  PyObject* aa = PyLong_FromLong(7L);
  printf("aa=%d\naa->ob_type=%p\n",PyLong_AsLong(aa),aa->ob_type);
  Py_Finalize();
  return 0;
}

correctly prints

aa=7

aa->ob_type=0x7f380d6b5800 (note that this will vary from run to run)

If I comment out Py_Initialize() and Py_Finalize() then I get a segmentation fault, but if I don't try reading the value with PyLong_AsLong then I get a null pointer for ob_type.

The documentation does tell you to initialize the interpreter.

With respect to Python2 it has two integer types PyInt and PyLong with PyInt handling the small values and so having the special-cased table. You'll probably see the same issue if you use PyInt in Python2. Anything you do with Python before calling Py_Initialize is undefined though, so it could fail in different exciting ways.

when I build a PyObject with C API of Python3, its ob_type is NULL

1 Answers1