C API type issue

[Gabriel Genellina]

En Fri, 03 Oct 2008 12:27:50 -0300, Rich Henry <againstmethod at gmail.com>  
escribió:

> Made a simple little test program as im learning to embed python, have a
> simple script that just sets x=10.0 in test.py and prints type(x). Python
> prints that x is a float but PyFloat_Check() returns false. If i removed  
> the
> check and just force it to print the double value, its correct. Any ideas
> why PyFloat_Check() returns false on a float? Thanks in advance.

You got the reference count wrong. This is by far the most common source  
of errors when writting C code for Python (at least for beginners)

> #include "Python.h"
> #include <cstdio>
>
> int main(int argc, char** argv)
> {
>     FILE* fp = fopen("test.py", "r");
>     assert(fp);
>
>     Py_Initialize();
>
>     //
>     //    create a dictionary, load the builtins and execute our file
>     //
>     PyObject *d = PyDict_New();
>     assert(d);
>     PyDict_SetItemString(d, "__builtins__", PyEval_GetBuiltins());
>     PyRun_File(fp, "test.py", Py_file_input, d, NULL);

Instead of a bare dictionary, I'd use a module. Or the __main__ module.  
See how PyRun_SimpleFileExFlags handles this (or just use that function,  
or other variant).

>     //
>     //    get a variable that should be in our global namespace
>     //
>     PyObject *x = PyDict_GetItemString(d, "x");
>     assert(x);

This returns a borrowed reference, not a new one.

>     Py_DECREF(d);

This releases the only reference to d, and the dictionary is deleted. This  
in turn releases the last refrence to x, and it's deleted too.

>     //    determine its type and print it
>     //
>     if(PyFloat_Check(x))
>     {
>         printf("x = %lf\n", PyFloat_AS_DOUBLE(x));
>     }

At this stage x is an invalid object.

>     Py_DECREF(x);

And this is wrong now.

>
>     Py_Finalize();
>     fclose(fp);
>     return 0;
> }



-- 
Gabriel Genellina