refactor!: nuke top level await

[olehmisar]

Closes #9959

May be postponed if #9987 makes performance acceptable.

Removed top-level await. BarretenbergSync.getSingleton and all functions and methods that depend on it are now marked as async.

Changes

• FunctionSelector.equals cannot accept string and { name: string; parameters: ABIParameter[] } anymore. Otherwise it has to be async as well
• EntrypointPayload, AppEntrypointPayload, and FeeEntrypointPayload are worth taking closer look at as they were changed more than other classes
• Carefully review orchestrator.ts in prover-client because i have no idea what that code is
• await wallet.setPublicAuthWit({ caller, action }, true).send().wait() is now await (await wallet.setPublicAuthWit({ caller, action }, true).send()).wait()
• contract.methods.xxx.selector now returns a Promise<FunctionSelector>

There may be bugs introduced

• if functions that previously returned boolean were changed to return Promise and now they always return truthy values (because Promise is truthy)
• if functions that previously thrown an error synchronously were changed to async and now they throw asynchronously, so it breaks the flow

[olehmisar]

still countless errors to fix

[ludamad]

@charlielye any comments on the approach of mass-asyncifying bb?

[olehmisar]

everything except end-to-end typechecks. Waiting whether this PR is even gonna be merged before continuing. Also, see the PR description for substantial changes and potential bugs introduced.

[Groxan]

Why not to simply remove await BarretenbergSync.initSingleton(); and make users call it on their side instead? In this case you will give users more control and, the most important, won't break the API.

[olehmisar]

@Groxan it can be done as a temporary workaround. But this PR should be the end solution

[Groxan]

@olehmisar Why do you think that couldn't be an end solution? Allowing users to initialize the "environment" manually and, in particular, allowing them to choose the best moment for it is actually a better practice. At least, that's how it currently works for node devs.

[olehmisar]

@Groxan UI apps don’t have a top-to-bottom flow like cli apps. Imagine your app crashing randomly here and there because some wasm code was not initialised.

[Groxan]

@olehmisar Nothing will crash if you initialize it in advance

[Groxan]

Initialization on-demand, as you propose, is fine when the initialization itself takes <100ms. But when the initialization takes several seconds, the UX will leave much to be desired.

[olehmisar]

@olehmisar Nothing will crash if you initialize it in advance

how do you initialize in advance in a multipage app that has import {...} from "@aztec/aztec.js" scattered all around the place? It's not a top-to-bottom CLI application.

Initialization on-demand, as you propose, is fine when the initialization itself takes <100ms. But when the initialization takes several seconds, the UX will leave much to be desired.

I don't understand what you wanted to say with this sentence. You have to initialise wasm in any case. The UX will be better if you can do other things while waiting for wasm to be initialised, unlike how it currently halts the whole UI.

[Groxan]

It doesn't matter what type of app you have, there is always an entry point. But actually, "in advance" doesn't necessarily mean "at the entry point". "In advance" means "at ANY time before" calling aztec's components, requiring bb to be initialized.

So, I'm just saying that as a developer I'd like to have a choice - how, when and where to initialize bb, because in different use cases different approaches may be needed to give better UX.

[olehmisar]

This discussion leads to nowhere so I am not gonna entertain it anymore. Nuking top level await in favour of async initialisation does not take any control from the user. Requiring the user to initialize bb manually is a half measure that is gonna lead to a lot of headaches (look at Aleo)

[ludamad]

Sorry, this is my fault for letting this thread be followed as far as it did, but I think the consensus is we prefer to just drop the top level await, there is already an async API that can be used it preferred. Apologies again

[olehmisar]

Just want to say that I didn't intend my previous message to be offensive. Not the best choice of words, so I want to apologize to anyone that may be offended by it.

[rahul-kothari]

fwiw, I think there are some great alternate fixes and PRs for these issues coming down the line!
Definitely a HIGH priority for us. Will loop back with a list of upcoming PRs on this topic