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

From: Peter Dimov (pdimov_at_[hidden])
Date: 2024-12-05 11:52:21

Alexander Grund wrote:
> If the hash-algorithms would accept `span<const void>` (or <const char>)
> instead as the only `update` method (no need for overloads) we could have
> 
>      hash_append( span{v} );
>      hash_append( span{st} );
> 
> by accepting a templated span and casting it to the span-type expected by the
> hash-algorithms.
> And/or a `hash2::span` that accepts any range-like in the ctor and doing the
> cast.

Yes, obviously.

However, there's an additional concern here that people perhaps don't
appreciate. The hash algorithm requirements govern user-defined algorithms
as well. N3980 wasn't supposed to be a library of hash algorithms; it was
supposed to be a framework into which any user-defined hash algorithm
can plug into, and be used for hashing C++ objects.

That's why the hash algorithm interface is kept minimal and uncomplicated
(which is at odds with making it as ergonomic as possible.)

I can obviously support any form of `update` by making it templated and do
the appropriate enable_if incantations, but I don't want to impose this
requirement on hash algorithm authors.

And I don't want to force hash algorithm authors to have to include
<boost/hash2/span.hpp>, because this would mean a physical dependency
on Hash2. At the moment, you can define a hash algorithm without including
anything from Hash2, or anything from Boost. And that's how it has to be.