Setting Exception data.

[Martin von Loewis]

Ivan Munoz <imunoz at eso.org> writes:

> When raising a foo.error exception In my C code I set the error object
> to some data (an object defined on my extension module, an integer and a
> string) as follows:
> 
>  v=Py_BuildValue("Ois", e, errorNumber, errStr);
>  if (v != NULL) {
>    PyErr_SetObject(ErrorObject, v);
>    Py_DECREF(v);
>  }

That may not exactly do what you expect. It first builds a tuple, and
then sets it as the exception. Python normalizes the exception,
essentially by doing

   exc = ErrorObject(e,errorNumber,errStr)

i.e. it creates an instance by calling the class with the tuple as
arguments.

> so far so good, in Python then I catch this exception as:
> 
> except foo.error, (object, errnum, errstr):

Here, you have an implicit instance-unpacking:

  object,errnum,errstr = exc

Now, since exc is a class instance, this accesses exc[0],exc[1],exc[2]
subsequently, which then invokes __getitem__ three times.

> Now what is the right way to acomplish the same from Python?, this does
> not seem to work ....
> 
>     raise foo.error MyObject, errnum, 'You are toasted'

Please have a look at the raise syntax. It allows
  raise Class
  raise Class, Value
  raise Value
  raise Class, Value, Traceback

Value can be an instance of Class, a single value, or a tuple; in the
latter cases, an instance is constructed by calling the class. So you
can write either

     raise foo.error(MyObject, errnum, 'You are toasted')

or

     raise foo.error,(MyObject, errnum, 'You are toasted')

Hope this helps,
Martin