See also python/cpython#108797.

Do you need a different review? Or are you waiting for something?

Do you need a different review? Or are you waiting for something?

I think @encukou wants to wait until the sprint before any C API guidelines are updated in any kind of way. Correct me if I'm wrong, Petr.

I think @encukou wants to wait until the sprint before any C API guidelines are updated in any kind of way. Correct me if I'm wrong, Petr.

Oh, I wasn't aware of that.

Some converters, like _PyEval_SliceIndex(), are designed so that they do not set the output parameter if the argument is None. It allows to set the default value to what you want: 0, -1, PY_SSIZE_T_MAX or Py_SIZE(self). It is the only way, because the signature of such converters is fixed, and they cannot take other arguments.

Some converters, like _PyEval_SliceIndex(), are designed so that they do not set the output parameter if the argument is None. It allows to set the default value to what you want: 0, -1, PY_SSIZE_T_MAX or Py_SIZE(self). It is the only way, because the signature of such converters is fixed, and they cannot take other arguments.

As I commented on python/cpython#108797, I think the guideline should care only about the general case, not the special case. There will be deviant APIs once in a while; that's ok.

BTW, _PyEval_SliceIndex is exposed through Python.h, but I guess it is considered an internal and/or private API since it is not documented and it is prefixed with an underscore. The guidelines are for public API.

I think the guideline should care only about the general case, not the special case.

I concur with @erlend-aasland on that.

FYI there is one public converter :-) PyUnicode_FSConverter().

I tried to move other private ones to the internal C API, but it's quite complicated: python/cpython#106320 (comment) I decided to give up in Python 3.13. I prefer to wait to see what's going on with the limited C API.

I'll leave this to the C API workgroup.