On 14 June 2018 at 01:15, Jeroen Demeyer <J.Demeyer@ugent.be> wrote:

I have finished my "meta-PEP" for issues with built-in (implemented in C) functions and methods. This is meant to become an "informational" (not standards track) PEP for other PEPs to refer to.

You can read the full text at https://github.com/jdemeyer/PEP-functions-meta

I also give brief ideas of solutions for the various issues. The main idea is a new PyTypeObject field tp_ccalloffset giving an offset in the object structure for a new PyCCallDef struct. This new struct replaces PyMethodDef for calling functions/methods and defines a new "C call" protocol. Comparing with PEP 575, one could say that the base_function class has been replaced by PyCCallDef. This is even more general than PEP 575 and it should be easier to support this new protocol in existing classes.

I plan to submit this as PEP in the next days, but I wanted to check for some early feedback first.

This looks like a nice overview of the problem space to me, thanks for putting it together! It will probably make sense to publish it as a PEP just before you publish the first draft of your CCall protocol PEP, so that the two PEPs get assigned consecutive numbers (it doesn't really matter if they're non-consecutive, but at the same time, I don't think there's any specific urgency in getting this one published before the CCall PEP needs to reference it as background information). Cheers, Nick. -- Nick Coghlan | ncoghlan@gmail.com | Brisbane, Australia

On 14 June 2018 at 01:15, Jeroen Demeyer <J.Demeyer@ugent.be> wrote:

I have finished my "meta-PEP" for issues with built-in (implemented in C) functions and methods. This is meant to become an "informational" (not standards track) PEP for other PEPs to refer to.

You can read the full text at https://github.com/jdemeyer/PEP-functions-meta

I also give brief ideas of solutions for the various issues. The main idea is a new PyTypeObject field tp_ccalloffset giving an offset in the object structure for a new PyCCallDef struct. This new struct replaces PyMethodDef for calling functions/methods and defines a new "C call" protocol. Comparing with PEP 575, one could say that the base_function class has been replaced by PyCCallDef. This is even more general than PEP 575 and it should be easier to support this new protocol in existing classes.

I plan to submit this as PEP in the next days, but I wanted to check for some early feedback first.

This looks like a nice overview of the problem space to me, thanks for putting it together! It will probably make sense to publish it as a PEP just before you publish the first draft of your CCall protocol PEP, so that the two PEPs get assigned consecutive numbers (it doesn't really matter if they're non-consecutive, but at the same time, I don't think there's any specific urgency in getting this one published before the CCall PEP needs to reference it as background information). Cheers, Nick. -- Nick Coghlan | ncoghlan@gmail.com | Brisbane, Australia