Re: [boost] [hash2][review] Early review (due to holidays)

On Thu, Dec 5, 2024 at 10:47 AM Peter Dimov via Boost wrote:

Zach Laine wrote:

...

On Thu, Dec 5, 2024 at 7:31 AM Peter Dimov via Boost wrote:

...

Alexander Grund wrote:

...

So span<void const> then or as that isn't accepted span<char const> assuming sizeof(char)==sizeof(byte)

span<void const> doesn't exist, though. I was giving it as an example of something that I would have added to the hypothetical hash2::span, but no existing span library has it.

If we limit ourselves to std::span, I suppose we can just add three (or four) overloads, one per byte type.

(Four because of char8_t.)

This part confuses me. It's always safe to cast to void const *; what not just have a single template that takes a std::span<T>, and cast it's .begin() and .end() to void const *?

If you mean this

template<class T> void update( std::span<T> sp );

then the reason to not use it is that you can't pass things like `string` and `vector<unsigned char>` and `char[65536]` directly because deduction will not work.

I'm not particularly fond of the user-facing syntax being

hash.update( std::span{buffer} );

instead of

hash.update( buffer );

Thanks, that makes sense. Looking at the docs, I think I get it now -- this is a user-provided API that your lib requires, right? If that's the case, I think I'd also require what you have, since there's no single concrete span type you could write there. An alternative might be to simply change the API depending on the value of __cplusplus -- void* + size_t before C++20, and std::spanstd::byte in C++20 and later. I realize some people might hate this. :) Zach