Refactor outside execution functions

[franciszekjob]

Feature Request

1. outside_execution_to_typed_data function should be a method of OutsideExecution, e.g.

@dataclass
class OutsideExecution:
    ...
    def to_typed_data(...) -> TypedData:

This requires resolving cycling imports errors, discussed here -> #1530 (comment)

2. OutsideExecution. to_abi_dict() should ideally use schema while creating dict.

[baitcode]

@franciszekjob I'm happy to tackle that issue, if you need help, of course. But would be nice to make a call to discuss.

[franciszekjob]

@franciszekjob I'm happy to tackle that issue, if you need help, of course. But would be nice to make a call to discuss.

Sure, you can address this once #1530 is done 😄

[kengoon]

I’d like to work on this.

[franciszekjob]

@kengoon sorry but @baitcode will be already tackling this issue.

[pheobeayo]

Can I jump on this task?

[nottherealalanturing]

Let me try this one!

[franciszekjob]

@baitcode would you like to work on this one as we discussed?

[kengoon]

@franciszekjob if his answer is negative, I'm still up for this.

[cryptogru812]

I’m interested in this one.

[jaykayudo]

Is this issue still available?

[kengoon]

@franciszekjob It's been 13 days with no response, can I take on this?

[baitcode]

@franciszekjob I'd love to but currently working on other issues. Can it wait for another week?

[nottherealalanturing]

@franciszekjob I can take it

[franciszekjob]

@franciszekjob I'd love to but currently working on other issues. Can it wait for another week?

@baitcode Yes

[BlessingEmejulu]

Can I start working on this?

[franciszekjob]

@BlessingEmejulu sorry, but another contributor will address it.

[emarc99]

May I try my hand at this?

[Nityam573]

Can I take this from here?

[uzochukwuV]

I’d love to work on this task.

[baitcode]

@franciszekjob Ready to start on 13th of Jan.

[franciszekjob]

@baitcode Ok, cool. Would you be also able to handle #1546?

[baitcode]

@franciszekjob Yes, sure, that is an easy one.

[ekumamatthew]

I’m interested in this one.

[baitcode]

@franciszekjob could you assign me to that one, please?

[franciszekjob]

@baitcode done

[baitcode]

@franciszekjob I did some research and found some problems.

I see that the main problem that produces loads of cyclic dependency issues is module: starknet_py.utils.typed_data
The main issue with this one - it's widely used and at the same time it imports a lot of deep living internals into itself.

I see 2 possible ways of approaching this

With minimal changes but increasing entropy.

Move OutsideExecution to separate module (need help with naming here) import typed_data there and implement discussed functionality. Will not produce any cyclic imports.

In several steps trying to improve module connectivity.

Lots of changes, probably will introduce some backward incompatible ones. I don't think I'll ever get approve for such a change. But at least I'll document my thoughts.
The idea is to make typed_data importable in starknet_py.net.client_models. I see several vectors of achieving that. Not sure all of those recommedations need to be implemented. But some of those definitely make sense and worth at least considering.

starknet_py/utils/typed_data.py simplification
The idea is to use Serialiser to replace the way _encode_data is preparing data for struct_hash().

split starknet_py.net.client_models

• move enums into subpackage importing indirectly
• mode schema dependent models

Import flow fix
starknet_py/utils internally use starknet_py/net/models
but judging from the structure, modules that are deeper in the tree should use parent utils and their own utils. but not the other way around, otherwise cyclic imports are hard to avoid.

I would propose to remove dependency:

• on _to_rpc_felt by moving method into utils.py and making it public - this will require a huge update as it is used a lot in the project and is imported indirectly
• from starknet_py.net.models.typed_data import DomainDict, Revision, TypedDataDict see point 2.
• from starknet_py.serialization.data_serializers import ByteArraySerializer - it would be absolutely awesome if data for struct_hash was generated using Serialiser, I'm happy to try to implement this one.

Merge starknet_py/utils/typed_data.py and starknet_py/net/models/typed_data.py

those files contain similar data structures, I haven't yet fully understood what are the responsibilities
Possible solutions:

• group all logic in starknet_py/utils/typed_data.py
• remove from_dict method on the starknet_py/utils/typed_data.py classes and and create to_typed_dict on the net/models classes and rewrite usage.

[franciszekjob]

Hi @baitcode 👋 , thanks for the deep dive into the problem! I think we should choose option 2, because these cyclic imports are really problematic in the long term, so it's best if we get rid of them. Will it brings lots of breaking changes? Would you like to start working on this?

[baitcode]

@franciszekjob Sure will do. But I wan't to approach it in several steps. Will outline those here a bit later.

[franciszekjob]

@baitcode sure, ping me whenever you need 👍