Consider using cargo-util-schemas to parse Cargo.toml

[Rustin170506]

Consider using cargo-util-schemas from the cargo repo.

We are currently using https://github.com/LukeMathWalker/cargo-manifest to parse the Cargo.toml. The Cargo team is planning to release the cargo-util-schemas. Using the same library from Cargo directly would help us keep using the same Cargo.toml schema from Cargo.

If it's OK, I am willing to help verify that we can use it and migrate to using it.

[Turbo87]

Using the same logic as cargo generally makes sense to me.

cargo-manifest however does have some functionality to e.g. find bins from the filesystem that are not explicitly declared in the manifest. While we are not using this functionality yet, people have been asking us to display binary crates differently from library crates, and to implement that we will need this kind of binary detection. We could also build it on top of cargo-util-schemas but I'm a bit weary about reinventing and maintaining another wheel if cargo-manifest and cargo-toml already have done so 😅

[epage]

I'm a bit confused by this response as previously the Cargo team formerly requested that crates.io not use these packages and, based on the response, it sounded like there was support for something like cargo-util-schemas as a way to meet both teams interests. This was the motivation for creating cargo-util-schemas when we did.

cargo-manifest however does have some functionality to e.g. find bins from the filesystem that are not explicitly declared in the manifest

I can't remember if I brought this up in the past to the crates.io team but an idea I've been considering is explicitly listing all inferred build-targets in the published Cargo.toml. This would allow crates.io to extract this information with just cargo-util-schemas (among other benefits). This wouldn't allow for backfills but I feel that is less likely to be needed as this is for the UI (as compared to things like rust-version) and over time new releases will come out that will include this information.

I've picked this work back up in earnest last week and hope to have results soon (the Cargo blog post has sucked up my last couple of days, which has a section on this).