feat: slim wallet

[Thunkar]

Reduces the surface of our Wallet API, focusing on removing the extension of PXE. This went downhill fast, but I think that ended up better than it was:

/**
 * The wallet interface.
 */
export type Wallet = AccountInterface &
  Pick<
    PXE,
    // Simulation
    | 'simulateTx'
    | 'simulateUnconstrained'
    | 'profileTx'
    // Sending
    | 'sendTx'
    // Contract management (will probably be collapsed in the future to avoid instance and class versions)
    | 'getContractClassMetadata'
    | 'getContractMetadata'
    | 'registerContract'
    | 'registerContractClass'
    // Likely to be removed
    | 'proveTx'
    // Will probably be collapsed 
    | 'getNodeInfo'
    | 'getPXEInfo'
    // Fee info
    | 'getCurrentBaseFees'
    // Unclear 
    | 'updateContract'
    // Sender management
    | 'registerSender'
    | 'getSenders'
    | 'removeSender'
    // Tx status
    | 'getTxReceipt'
    // Events. Kept bc events are going to be reworked and removing them now increases the already huge diff
    | 'getPrivateEvents'
    | 'getPublicEvents'
  > & {
    createAuthWit(intent: IntentInnerHash | IntentAction): Promise<AuthWitness>;
  };

• Authwits are not longer stored in PXE. It really didn't make sense: either they were used and "consumed" (or at least they should in most use cases), or the txs failed because they were invalid and as such didn't make sense to keep them around. Removing them from PXE's DB and forcing them to be provided externally cleans up the wallet's interface considerably, and delegates authwit management on the wallet software (which I think it's good). Closes: account.createAuthWit should not add the witness to PXE #12394
• The above meant supporting transient authwits in unconstrained simulation, which has been done.
• Removes a bunch of debug utilities from wallet (namely debug option from .send and proven: true) In the real world it doesn't make sense to wait for several epochs so txs are proven (a utility is kept for e2e testing) and all the info that debug retrieved can be easily obtained via a single query to the node. This also cleans up the interface nicely.
• Cleans up the @aztec/entrypoints package, attempting to reduce the interdependency with aztec.js
• Starts addressing the insanity that is ContractFunctionInteraction. There's a new object called ExecutionPayload that represents the "minimal unit of execution" in Aztec. It consists of contract calls, authwits and capsules, and every single higher abstraction object in our codebase generates one. After that, they can easily be composed to create complex interactions, such as account entrypoints or batched calls. It builds up on the system @LeilaWang designed, but starts a little bit earlier in the chain and I think ends up looking cleaner.
• As such, a lot of the complexity of our DeployMethod/DeployAccountMethod classes have been delegated on these objects and the factories that build them.

[netlify]

❌ Deploy Preview for barretenberg failed. Why did it fail? →

Name	Link
🔨 Latest commit	2656753
🔍 Latest deploy log	https://app.netlify.com/sites/barretenberg/deploys/67d96978b95d030008bf5a35

[benesjan]

Looks like a step in the right direction. Just needs a lot of docs fixes

Making authwits and transient capsules stateless is very cute

[benesjan]

It's no longer supported to get the function via the methods thing?

This function is hard to read as we have unconstrainedFunctionArtifact and functionArtifact in the same scope and it's not clear what is what.

[Thunkar]

I've tried to clarify names, but this is necessary until it's further refactored. It's possible to use contracfunctioninteraction.request() to get a building block to assemble a tx, but I've removed the ability to inject capsules and authwitnesses after the fact, in an effort to improve the interface. This and register_class.ts are the only places where this happens and I'll try to improve in next PR

[benesjan]

Throw when options are populated?

[Thunkar]

This is going to be heavily refactored, entrypoints are a mess

[benesjan]

Suggested change

	/** Represents the ExecutionPayload after encoding for the entrypint to execute */
	/** Represents the ExecutionPayload after encoding for the entrypont to execute */

Could you expand on what does it mean to encode for an entrypoint here?

[Thunkar]

Unfortunately it's a sad situation. Entrypoints require function calls to be encoded in noir format (snake_case, args hash, etc). The whole payload shouldn't have to be encoded, but changing the model turned out to be extremely complicated in this PR. Will change soon to something that's not completely unreadable.

[benesjan]

Happy merging!

[sklppy88]

a great change !

[sklppy88]

This is interesting to me, a contract function interaction having authWitnesses as well as the ones being passed in. When is this useful ? or is it on the list to be reworked ?

[Thunkar]

In the particular case of contractfunctioninteraction it was required for them to have capsules so you could create a batch call of deployments, for example, but I don't love it. Hopefully it's part of what's going to be refactored