It was specific to Python 3.7, which is already EOL'd, so not much point
in investigating further. I won't deliberately make it broken there
though, so the order will stay maintained like this as much as possible.
They're unpacked to full floats on element access and packed back from
full floats on mutable access, which makes this all very nice and
transparent. Yay Python!
That's it for now, I'll postpone half-float and matrix types for later
when these are actually needed, as it needs extra testing for the
aligned variants too.
To not have to duplicate these for each and every case, enlarging the
surface for potential bugs. Also changing the signatures to number +
identifier, instead of identifier repeated number of times. Means the
compiler won't be able to deduplicate / overlap the literals in the
binary, but is much more maintainable.
For Numpy this causes the slices to be reported as actual typed Numpy
arrays instead of typeless tuples, which is good I guess? Not sure why
that wasn't the case before, or what is the difference, and how it will
behave for sparse types such as aligned matrices.
It was casting in the wrong direction, causing an unbound type to be
returned and also accessing the data totally wrong. Should have tested
this properly in the first place.
Otherwise people wouldn't be able to conveniently use os.path, which is
an undesirable pain point. And do this for all currently exposed APIs,
not just the AbstractImporter that caused a problem.
This reverts commit 28497ec2ca.
Because for some reason pybind11 on *some platforms* (or in some
configurations?) picks the integer overload even for enum arguments,
only because it's first in the overload set.
This is extremely brittle, but for now I don't know any better solution.
In concrete types. Yes. Well, some of them for now, I still need to
sleep on normalized and matrix types, how it makes the most sense to
expose those. Oh, and array attributes as well.
Because making & testing my own string caster is not something I'd want
to do right now (seems complicated by looking at what's all done for
std::string), the functions currently all go through a std::string
instead.
Not ArrayView yet, and also no documentation on this whatsoever. That'll
come next. Also not everything works with arbitrary types yet,
converting from buffer protocol doesn't remember the format and
conversion to bytes doesn't take the actual type size into account
either.
Pybind changed py::module to py::module_ in order to support C++
modules, adapting to that change. It still builds on older versions,
but we're using only the new APIs.
That makes it much easier to handle in that case (because that's the
whole point of static builds) and also nicely circumvents any issues
with duplicated global data when a static lib would be linked to
multiple dynamic libraries.
Vladimír Vondruš