Impl bytemuck traits for GenericFamily
#213
Conversation
Probably not. It is already used by the font stack and the reason that I used it here is to provide a better implementation of something for I did an earlier version where I had a const and used that for We'll also use this once we look at FFIs (hopefully). |
waywardmonkeys
force-pushed
the
bytemuck-generic-family
branch
from
713ab68 to
53ceac3
Compare
nicoburns
dismissed
their stale review
Ok, I guess we can do unconditional then.
|
I'm a little uncomfortable with those unsafe impls. Specifically, it seems like it would be very easy for the MIN and MAX constants to get out of sync. I think we ought consider checking those in CI with a macro at some point. But they do seem to be correct now. |
|
They are tested in the PR already. |
|
We should probably have a comment in |
This can be used in the future as part of FFI work, but for now, the `Contiguous` trait is used to provide a `MAX_VALUE` used by `GenericFamilyMap`. This is a backport of part of linebender#209 where the separation between `GenericFamily` and `GenericFamilyMap` has grown to span crates. The testing here borrows from (copies) what is done for similar testing of `bytemuck` traits in the `color` crate. These crates already depend on `bytemuck`, so this isn't a new dependency. A decision is made here though to not use the `derive` feature as when we move this type to a vocabulary crate, we will want to avoid the `derive` feature there to avoid a dependency on proc macros.
waywardmonkeys
force-pushed
the
bytemuck-generic-family
branch
from
53ceac3 to
d9dbfc5
Compare
This can be used in the future as part of FFI work, but for now, the
Contiguoustrait is used to provide aMAX_VALUEused byGenericFamilyMap.This is a backport of part of #209 where the separation between
GenericFamilyandGenericFamilyMaphas grown to span crates.The testing here borrows from (copies) what is done for similar testing of
bytemucktraits in thecolorcrate.These crates already depend on
bytemuck, so this isn't a new dependency. A decision is made here though to not use thederivefeature as when we move this type to a vocabulary crate, we will want to avoid thederivefeature there to avoid a dependency on proc macros.