Incorrect usage of bytes constructor

[zilm13]

From the python spec:
A zero-filled bytes object of a specified length: bytes(10)
https://docs.python.org/3.10/library/stdtypes.html#binaryseq
So when initialized with int or something like 0x01 in https://github.com/protolambda/remerkleable/blob/master/remerkleable/byte_arrays.py#L28 you get 00-filled bytes of input length.
Broken usage is, for example, here: https://github.com/ethereum/consensus-specs/blame/dev/specs/eip4844/beacon-chain.md#L58

[zilm13]

I'm not sure, maybe consensus-specs input should be changed but it's easy repeatable error

[protolambda]

Nice catch! I think consensus-specs should specify it as string input for clarity, based on previous similar usage in the withdrawal prefix constants here: https://github.com/ethereum/consensus-specs/blob/dev/specs/phase0/beacon-chain.md#withdrawal-prefixes

That said, it's indeed an easy thing to get wrong, at least for the case when the value is equal to the bytevector length. (since it's checked here:

remerkleable/remerkleable/byte_arrays.py

Line 81 in 94ed7e2

if len(out) != byte_len:

And for a ByteList it makes sense to just follow the bytes behavior. For the fixed-size ByteVector it doesn't make sense, as one should just call the constructor without arguments to get the default value. So to make it less error-prone, I think we should add a check to the ByteVector constructor, before calling the super constructor, like if len(args) == 1 and isinstance(args[0], int): raise Exception(f"cannot init fixed-length ByteVector[{byte_len}] with dynamic-length {args[0]}")

What do you think?

[zilm13]

I agree with you as it's better to throw an exception as it could be not one byte integer but any size and even if we do conversion back it will be useless two-way conversion with new potential for error by filling leading zeros for example. I think we should go exception way