pycuba userdata problems

[johannesbuchner]

pycuba userdata problems

[JohannesBuchner]

Hi Stevens,

try making it a pointer. In the cuba manual (cuba.pdf) it is defined as a (void *)
For instance

n = 0.5
p = ctypes.POINTER(c_int)(n)

Avoid this trap: #5 and this is the appropriate manual for ctypes: http://python.net/crew/theller/ctypes/tutorial.html
For debugging it can also useful to set the environment variable CUBACORES=1 disabling parallel execution.

Regarding separating pycuba out, I don't really see a benefit. Its a tiny package, and separation would triple the maintenance effort (e.g. creating the documentation) although the package might be more visible. I don't see the harm in having the other packages installed. Combining the packages also means that people might like to try out the other algorithms or switch between them. I see the package similar to scipy.integrate or a kind of scikit-integrate.

Cheers, Johannes

[steven-murray]

Hi Johannes,

Sorry it has been a while for me to get back to this project. I have tried what you suggest (except that I use c_double rather than c_int), and neither does it work. I think the problem is that when the userdata variable is passed to the Integrand function, in which it should be of type c_void_p, it is actually of type "int". I'm assuming this int is the address of the memory at which the userdata is stored, but I can't quite figure out more than that.

Cheers,
Steven

[JohannesBuchner]

That is correct. Hopefully I will have time to look into the problem. I usually put my extra variables into a global variable and access them from the function. Maybe you can do that in the meantime?

[steven-murray]

Yes if you have time it would be nice to have this sorted. In the meantime I have been doing as you suggest and putting my extra variables as globals.

Cheers,
Steven

[JohannesBuchner]

If you or someone else seriously needs this functionality in python, please re-open.

[farnikn]

Hi Johannes,

I'm dealing with the same problem by putting my extra variables into a global variable, but I was wondering if there is any update for this issue and passing the userdata as a ctype variable?

Best,
Farnik

[JohannesBuchner]

I don't really see the need to mangle it into C and then back out. You don't even need a global variable either, you can use a class with a method that accesses the classes attributes, or you could create a function that returns a function.

[steven-murray]

I agree with the approach @JohannesBuchner suggests -- wrapped functions or classes would be much easier.

[farnikn]

Yeah I got it, thanks a lot!