In EC keys store the data as EC_Scalar / EC_AffinePoint

[randombit]

In the short term we still have to retain everything in the old types since there are getters that return a const reference to those values.

#4027

[coveralls]

coverage: 91.657% (-0.01%) from 91.67%
when pulling 317ee07 on jack/ecc-key-struct
into 3e2a7ce on master.

[randombit]

@reneme Ping on this one - this is blocking other work that I'm loathe to start knowing the merge conflicts will get messy so I'd like to get this merged within ~~ the next week if possible.

[reneme]

Looks good to me. Several std::move() related nits in the constructors. I realize that the inner structure of the affected structures doesn't benefit a lot from moving. I find it good measure, nevertheless.

I tried the suggestions locally, so here's a patch, if you just want to apply it: 220ee66

[reneme]

Moving BigInt down the from_bigint() line would also make sense. Eventually it would end up in EC_Scalar_Data_BN and moved in place. That would require further adaptions down this line, though.

[randombit]

I'm not sure if this one makes sense in the long run - in fairly short order the vast majority of EC_Scalar::from_bigint is going to throw away the BigInt argument because it's converting the value to a PrimeOrderCurve::Scalar instead. So it seems to me better to avoid the copy; the move optimizations will only be helpful for obscure/deprecated curves.

[reneme]

Perhaps that's a good vehicle to try and use std::ranges::views (and perhaps even integrate this with concat()), to end up with a construction like so:

namespace v = std::ranges::views;
auto encoding =
   concat<std::vector<uint8_t>>(
      '\x04',
      v::reverse(std::span{bits}.first(part_size)),
      v::reverse(std::span{bits}.last(part_size)));

... certainly not here, though.

[reneme]

If you wanted to avoid a few heap allocations (not sure this is worth it, though):

Suggested change

	hash.update(group.get_a().serialize(p_bytes));
	hash.update(group.get_b().serialize(p_bytes));
	hash.update(group.get_g_x().serialize(p_bytes));
	hash.update(group.get_g_y().serialize(p_bytes));
	auto update_with_scalar = [bytes = std::vector<uint8_t>(group.get_p_bytes())](HashFunction& h,
	const BigInt& scalar) mutable {
	scalar.serialize_to(bytes);
	h.update(bytes);
	};
	update_with_scalar(hash, group.get_a());
	update_with_scalar(hash, group.get_b());
	update_with_scalar(hash, group.get_g_x());
	update_with_scalar(hash, group.get_g_y());