Add std::array support to BigInt::serialize function

[KaganCanSit]

This PR resolves the TODO comment in BigInt::serialize() by adding support for std::array<uint8_t, N> alongside existing std::vector and secure_vector support.

TODO Message:

// TODO this supports std::vector and secure_vector
// it would be nice if this also could work with std::array as in
// bn.serialize_to<std::array<uint8_t, 32>>(32);

[coveralls]

coverage: 90.337% (+0.001%) from 90.336%
when pulling 73d954c on KaganCanSit:bigint-array-serialization
into 90b85a8 on randombit:master.

[reneme]

Nice, thanks!

Interesting way to detect an array! However, IIUC this would work for other tuple-types as well, no? Like std::pair<int, std::string>. That would produce very strange compiler errors for sure... But then again, the API does also allow passing a std::vector<uint32_t>, which is just as awful I guess. You could look into using the statically_spanable_range<> concept to detect the array instead.

Anyway: Could you please add a few tests that serialize a BigInt into an array? src/tests/test_bigint.cpp has a function test_bigint_serialization, please add another CHECK that uses std::array. Also check that the exception is thrown as expected if the array is too short or too long.

[KaganCanSit]

Nice, thanks!

Interesting way to detect an array! However, IIUC this would work for other tuple-types as well, no? Like std::pair<int, std::string>. That would produce very strange compiler errors for sure... But then again, the API does also allow passing a std::vector<uint32_t>, which is just as awful I guess. You could look into using the statically_spanable_range<> concept to detect the array instead.

Anyway: Could you please add a few tests that serialize a BigInt into an array? src/tests/test_bigint.cpp has a function test_bigint_serialization, please add another CHECK that uses std::array. Also check that the exception is thrown as expected if the array is too short or too long.

You're right that it will provide general support for tuple types. I've made the changes as you suggested.

Additionally, for the situation you mentioned, the API does also allow passing a std::vector<uint32_t>, which is just as awful I guess. I think we could use requires std::same_as<typename T::value_type, uint8_t. There were no issues during compilation and testing.

However, since this is a general change to the API, I haven't added it to the branch. What do you think?

Update: 20.01.2026 - Rebase master branch and remove static_cast usage in big_int test code.

[reneme]

Sorry for changing my mind on this, but I think there's a better strategy to implement this TODO.

[KaganCanSit]

Sorry for changing my mind on this, but I think there's a better strategy to implement this TODO.

Thanks for the detailed feedback and this learning opportunity.

I've pushed the changes but since I worked on this after hours, I'd like to revisit and review them more thoroughly when I'm fresher (I'm not yet comfortable using concepts actively). I want to make sure I fully understand the reasoning behind each change. I will write experimental code blocks for exploration and possibly add more tests later.

But absolutily like this way, more readable and effective.

[reneme]

LGTM. @randombit I'm guessing you agree with adding a BigInt::serialize_to_array<32>() given that you left the TODO?

[KaganCanSit]

Hit #5376, #5328, #5334, #5341... Rebased master branch.