[Numpy-discussion] SWIG, numpy.i and errno: comments?

[Bill Spotz]

Egor,

This looks about right.  However, it is customary to invoke the SWIG  
macro "SWIG_fail;" instead of "break;".  (This translates into a  
"goto" to the failure label, and is better in case there is any other  
cleanup code to execute.)

On Aug 9, 2009, at 6:17 AM, Egor Zindy wrote:

> Hello list,
>
> this is my attempt at generating python exceptions in SWIG/C using the
> errno mechanism:
> http://www.scipy.org/Cookbook/SWIG_NumPy_examples#head-10f49a0f5ea6b313127d2ec5ffa1eaf1c133cb22
>
> Used together with numpy.i, this has been useful for notifying (in a
> pythonic way) memory allocation errors or array index problems.
>
> A change in the errno global variable is detected in the %exception
> part of the SWIG interface file, and Python exceptions are generated
> after $action depending on the errno error code value.
>
> %exception
> {
>    errno = 0;
>    $action
>
>    if (errno != 0)
>    {
>        switch(errno)
>        {
>            case EPERM:
>                PyErr_Format(PyExc_IndexError, "Index out of range");
>                break;
>            case ENOMEM:
>                PyErr_Format(PyExc_MemoryError, "Failed malloc()");
>                break;
>            default:
>                PyErr_Format(PyExc_Exception, "Unknown exception");
>        }
>        return NULL;
>    }
> }
>
> If there's a better way of doing this, I'll update the cookbook  
> recipe.
>
> Regards,
> Egor

** Bill Spotz                                              **
** Sandia National Laboratories  Voice: (505)845-0170      **
** P.O. Box 5800                 Fax:   (505)284-0154      **
** Albuquerque, NM 87185-0370    Email: wfspotz at sandia.gov **