enhancement: factor out a type from access lists

[richardgreg]

What was wrong?

Because it's repeated in several different transaction types, the Tuple[Address, Tuple[Bytes32, ...]] portion should be promoted to its own type.

Related to Issue #
Fixes #947

How was it fixed?

This PR refactors the access list type used in various transaction classes to reduce redundancy and improve maintainability. Specifically, it introduces a new Access class to represent the access_list structure, which is used in multiple transaction types.

1. Created a new Access class to encapsulate the account and slots structure.

@slotted_freezable
@dataclass
class Access:
    account: Address
    slots: Tuple[Bytes32, ...]

2. Updated Transaction Classes

3. RLP Encoding and Decoding Adjustments:

• Introduced helper functions encode_access_list and decode_access_list to handle the conversion between the new Access type and the previous tuple structure for RLP encoding and decoding.
• Updated encode_transaction and decode_transaction functions to use these helpers.

Cute Animal Picture

[SamWilsn]

This looks like pretty much exactly what I had in mind!

[richardgreg]

If this is good, I can do it for Prague if need be

[SamWilsn]

I've been playing around with this PR a bit, and it's looking great!

I think it should be rebased on top of #968. That branch can handle nested dataclasses better than master, and you shouldn't need the encode_access_list and decode_access_list helpers there.

That said, there's going to be a bit more work involved in this refactor. You'll need to use the Access type in tests near, for example, here:

execution-specs/tests/shanghai/test_rlp.py

Line 62 in c854868

((address1, (hash1, hash2)), (address2, tuple())),

There are a few other required changes, like dealing with unpacking here:

execution-specs/src/ethereum/shanghai/fork.py

Lines 620 to 624 in c854868

	if isinstance(tx, (AccessListTransaction, FeeMarketTransaction)):
	for address, keys in tx.access_list:
	preaccessed_addresses.add(address)
	for key in keys:
	preaccessed_storage_keys.add((address, key))

mypy is your best friend when doing refactoring. The easiest way to run it in this project is:

$ tox -e static

[SamWilsn]

Suggested change

# Helper function to handle the RLP encoding of the Access class instances.

[SamWilsn]

Suggested change

# Helper function to handle the RLP decoding of the Access class instances.

[richardgreg]

I'm back to continue working on this

[SamWilsn]

Took the liberty of rebasing and running isort/black. You'll need to force pull/reset your local branch.

[SamWilsn]

Was removing these intentional?

[richardgreg]

I don't think so. Fixed it 🙂

[richardgreg]

Took the liberty of rebasing and running isort/black. You'll need to force pull/reset your local branch.

Thanks. Running tox -e static breaks due to some issue with rust on my system that I'm still trying to fix, but running the individual linting/formatting commands works. I will be doing that from now on. I also noticed running mypy src tests setup.py --exclude "tests/fixtures/*" --namespace-packages returns many errors. Should that be an issue on its own?

[SamWilsn]

Running tox -e static breaks due to some issue with rust on my system that I'm still trying to fix

What seems to be the problem?

I also noticed running mypy src tests setup.py --exclude "tests/fixtures/*" --namespace-packages returns many errors. Should that be an issue on its own?

Does it return errors on an unmodified master checkout, or just on your branch? master passes mypy for me without any errors. If both branches are erroring, it's likely that you're missing some packages in your install. Inside your virtual environment, try pip install -e .[lint,test].