[C++-sig] Re: [BPL] python::interpreter proposal

[Joerg Sauer]

Hi,
up to know we have calling scripts, evaluations and files. 
What about calling python functions directly and pass it arguments.

Maybe I'm wrong, cause I'm new to all this, but I thought I could
build my extension modules, some in c++ using BPL and some as Python
code. In the Python modules I have declared some functions (in my case
script hooks to extend /change the default behaviour of my app).
Then I would like to call one of the functions and pass it their
params directly. As far as I see I would use one of
PyObject_CallObject()
PyObject_CallFunction()
PyObject_CallMethod()

As far as I can see using PyObject_CallObject() would have the
advantage to pass the parameters firectly. And with the to_python
conversions I could pass my apps c++ objects (wrapper classes).

This would have the advantage that the script (function) can
manipulate the data directly.

Using the proposed PyRun* calls has the disadvantage that I don't get
my objects to deal with only indirectly, or did I miss something?


Another thing is sync and async execution. For me it would be nice to
have the possibility to run a python script in both states. If it is
running async I would like to have 
a) a state function to get execution state (busy, finished)
b) registered callback to be triggered when execution is finished.

This could be done directly in the interpreter class or detaching a
new thread for async execution by myself. So what do you think, better
leaf async execution alone or include it as option in the interpreter?

Joerg