Add the ByteSize measure

[nfrisby]

The new ByteSize measure should be used for all byte size summations relevant to the Cardano chain.

• It has checked arithmetic, with a pure representation of overflow.
• It still has a Monoid instance that will be useful for finger trees.

The other types defined in the module should be used for recording/transmitting the result of summations that did not overflow.

[nfrisby]

TODO test suite

[lehins]

I took a quick glance at it. Looks interesting.
I'll do a proper review after the Chang HF, if that is ok?

[lehins]

I wouldn't really call it partial, since partial functions in Haskell usually are the ones that produce exceptions, eg. error is a partial function. I get it that math has a different view on the subject, but in order to stick to Haskell nomenclature I'd call it something like toByteSizeMaybe

DefaultSignatures can reduce some boiler plate

Suggested change

	partialFrom :: a -> StrictMaybe ByteSize
	partialFrom :: a -> StrictMaybe ByteSize
	default partialFrom :: ByteSizeFrom a => a -> StrictMaybe ByteSize
	partialFrom = partialFromDefault

[lehins]

Not only to and from are very confusing and ambiguous names, but they are also highly likely to be used in user's codebase. Without qualification these will lead to clashes quite often (even lens/microlens has to)

I'd suggest better names that depict what is the actual conversion here. So in order to avoid ambiguity of what is being converted to what, I'd suggest something like this:

-- | A type that can be safely converted to `ByteSize`
class ToByteSize a where
    toByteSize :: a -> ByteSize

-- | A type that can be safely constructed from `ByteSize`
class FromByteSize a where
    fromByteSize :: a -> ByteSize

[nfrisby]

Without qualification ...

So far, I'm intending for this module to be imported qualified.

Example: ByteSize.from @Word32 1000

[coot]

Do we need to export SJust . from at all?

[lehins]

So far, I'm intending for this module to be imported qualified.

I appreciate your intention, but that should not be forced on the user. Moreover, qualification does not solve the ambiguity of naming from and to.

So, to put it bluntly, I will not let functions with names like from and to to be merged.

[lehins]

Why CPP here?

Suggested change

	#if __GLASGOW_HASKELL__ < 900
	-- Use the GHC version here because this is compiler dependent, and only indirectly lib dependent.
	import GHC.Natural (Natural)
	#endif
	import Numeric.Natural (Natural)

[nfrisby]

I don't know; I merely copied it from an existing module in the package.

Edit: seems like the CI does not like it 🤔

[coot]

This could overflow, isn't it? Shouldn't we do sth like this here:

 \a -> if a >= fromIntegral sentinel then SNothing else ...