[castrog]

Summary

We use proto in our projects, and supporting Chrome for Testing's version feed would make a ChromeDriver TOML plugin possible for browser testing. Its known-good-versions.json wraps versions in a versions array, while schema plugins currently expect the response itself to be an array.

Add manifest-array-key (v1) and index_array_key (v2) to select the array:

[resolve]
manifest-url = "https://googlechromelabs.github.io/chrome-for-testing/known-good-versions.json"
manifest-array-key = "versions"
manifest-version-key = "version"

Top-level arrays continue to work. Missing keys and values that are not arrays return clear errors. The v2 setting also works in resolve overrides.

Addresses moonrepo/proto#948.

Verification

• cargo build -p schema_tool --target wasm32-wasip1 --release --locked — passed.
• cargo nextest run -p schema_tool --no-default-features --no-fail-fast --no-tests=pass --locked — 85 passed.
• cargo clippy -p schema_tool --target wasm32-wasip1 --release --locked -- -D warnings — passed.
• cargo fmt --all --check — passed.

Thanks for maintaining proto and its plugins—we use them regularly.

[milesj]

@castrog What if we supported xpath-like notation instead? index-version-key = 'versions[0].version'?

[castrog]

That makes sense. Were you thinking JSONPath, like $.versions[*].version to get every version, or a simpler path without the $? I can adjust the PR either way.

[milesj]

Maybe whatever jq users to keep the syntax familiar?

[castrog]

Thinking more about your suggestion, there are two options for a version query format. Either would replace the array-key setting in this PR, so we’d ship one way to extract versions from the ChromeDriver JSON index:

Option	Query	Benefit	Cost
jq-style syntax with jaq	.versions[].version	Familiar to jq users	Adds ~1.18 MB to schema_tool.wasm in our local build
JMESPath syntax	versions[*].version	Adds only ~183 KB	Less familiar syntax for jq users

Both handle this lookup. On a synthetic index of 2,590 versions, both ran in under 2 ms, so performance doesn’t seem to decide it. I lean toward JMESPath for this narrow version lookup because of the size difference, but jq familiarity is a real benefit. Which tradeoff makes more sense for schema_tool?

[milesj]

@castrog JMESPath seems fine to me. In this example, does [*] represent the 0 index?

[castrog]

@milesj it represents all, versions[0].version would represent the first entry.

[milesj]

Sure, works for me