feat(cli): Try Prover.json if Prover.toml is not present

[aakoshh]

Description

Problem*

Extracted from AztecProtocol/aztec-packages#11294

Summary*

The debug and execute CLI commands accept --program-dir and a --prover-name argument to specify where to look for the inputs for the circuit. The prover name is Prover by default, but we can't specify e.g. Prover.json to make it take a JSON file, it always tries to use Format::Toml to parse it, which determines the .toml extension.

The PR changes commands that read Prover.toml to try both Format::Toml and Format::Json (in that order), so that if Prover.toml is not present but Prover.json is, then we use the latter.

Additional Context

Ostensibly we could have a --format argument, but I found this useful when capturing the input witnesses from the TypeScript integration tests like here, and wanted to use it as-is to replicate the error like this, without having to go through more hoops.

Documentation*

Check one:

• No documentation needed.
• Documentation included in this PR.
• [For Experimental Features] Documentation to be submitted in a separate PR.

PR Checklist*

• I have tested the changes locally.
• I have formatted the changes with Prettier and/or cargo fmt on default settings.

[vezenovm]

Let's add the change for the info command as well. We need it for profiling Brillig execution.

[TomAFrench]

This conflicts with nargo check which will ignore json files.

[TomAFrench]

If we're accepting this as a first class input format we then also need to support it in the profiler also, etc. etc.

[aakoshh]

I see that the check command can create an Prover.toml template file, but I don't see it reading the inputs, other than to verify they exist before accidentally overwriting them 👀

If that's all it does, then I don't think it necessarily conflicts, it just doesn't create a JSON template at the moment.

[TomAFrench]

I think it conflicts if we're expecting users to have a Prover.json file but that then prevents them from running nargo check without their inputs being soft-overwritten by an empty template file.

[aakoshh]

Maybe I'm misunderstanding something, but the check command at the moment is hardcoded to write a Prover.toml file, nothing else, it will ignore any JSON, won't write over it.

[TomAFrench]

Yeah but if you then try and execute your program afterwards then nargo will pick up the empty toml file over the json file you've been using previously.

[aakoshh]

Ah, by "soft overwrite" you meant shadowing, gotcha.

Thanks for pointing out the profiler, I updated that to handle both formats.

[aakoshh]

Updated check_cmd to check that any prover file with a format other than Format::Toml exists, and if so then ask for --overwrite to be used, even though it will not be deleted, just shadowed by the TOML template.

[aakoshh]

To highlight some of the differences with working with the different formats, I transformed this JSON witness to TOML with https://transform.tools/json-to-toml ; here's a part of it as it's opened in my editor:

Here's what a heavily hierarchical part looks like in JSON:

The same in TOML as it's formatted by the tool:

Luckily the tool indents the content to follow the nesting, but it's inconsistent: [previous_kernel_public_inputs.constants] is indented more than [previous_kernel_public_inputs.constants.historical_header.state.l1_to_l2_message_tree] below it.

When I save the TOML in VSCode it gets rid of all indentation. Tried a setting in the plugin but it didn't make a difference, but I don't expect a uniform setting for everyone.

Another thing I found work better in the JSON editing is the collapse of arrays. The TOML editor can collapse structs like thekey_validation_requests_and_generators here:

But there is no wrapper for the note_hashes array below it, so those entries can only be collapsed individually, not as a whole group.

In JSON this is easy, so helps to explore the document and delete regions that aren't interesting for the test:

To TOML's credit a collapsed record still shows its full path which is more useful than seeing those {} above note_hashes.

[github-actions]

⚠️ Performance Alert ⚠️

Possible performance regression was detected for benchmark 'Test Suite Duration'.
Benchmark result of this commit is worse than the previous benchmark result exceeding threshold 1.20.

Benchmark suite	Current: ce8ab2d	Previous: fc75298	Ratio
noir-lang_noir_sort_	1 s	0 s	+∞
noir-lang_mimc_	1 s	0 s	+∞
AztecProtocol_aztec-packages_noir-projects_noir-protocol-circuits_crates_blob	61 s	50 s	1.22

This comment was automatically generated by workflow using github-action-benchmark.

CC: @TomAFrench