Antonio Cuni

Some random considerations: 1. as a general rule, we need to keep in mind that we want to make migration to HPy easy: so, I'm +1 to clean up the...

> @mattip kindly grep'ed pandas, scipy and numpy for us to see which calling conventions they use. The full results are at https://paste.ubuntu.com/p/3xs6dC5Gy9/. The summary is that the most frequent...

Update as of 2021-11-29. In the meantime, we merged #147 and we added `HPy_CallTupleDict` as the most generic calling API, and basically the equivalent of the old `PyObject_Call`. In the...

> I'm not sure what postponing the decision about `PY_VECTORCALL_ARGUMENTS_OFFSET` means concretely. Does it mean that initially `HPy_CallMethod` and `HPy_CallMethod_s` won't support the offset flag? Or that they initially will...

> Was the idea for HPy to only support heap types if we can since non-heap types are global state we'd like to avoid? in an ideal world: yes. However,...

Here is a possible way to augment `exec` in such a way that it remembers the source code: ```python import inspect import linecache def smart_exec(source, *args): if isinstance(source, str): n...

> Numba functions can be pickled (cloudpickle, to be precise) for remote execution (most commonly, with Dask), and will be re-compiled on the target system in case it does not...

> I don't know if anyone is applying the Numba decorator to functions after unpickling on the destination. I think that this case is also covered, because pickled functions are...

@timfel > The problem is that the bitcode doesn't keep any information about the cast around. So we just get an "access this handle at offset such and such" Note...

@arigo > For now I'll just point out that "obj" is automatically closed both makes the API more irregular, and is wrong in your example, because a function should not...