Unstable Feature Usage Metrics

[yaahc]

Unstable Feature Usage Metrics

Track unstable feature usage trends by Rust users.

Motivation

• Support feature stabilization prioritization.
• Helping teams know which unstable features to invest their energy into.
• Evaluating how representative crates.io is as a sample of the rust ecosystem (will we see different patterns in crates.io feature usage vs private feature usage?)

Context

• research paper on unstable feature usage in the rust ecosystem

Steps / History (PROVISIONAL)

• define the format the metrics will be stored in
• add flag to dump unstable feature status metrics (what features exist, if they're stable or not, both for library and lang/compiler, tidy may already support this for lib features)
• add flag to enable unstable feature usage metrics in the compiler
• front end to display reports based on gathered usage and status metrics
• back end to save metrics data from published crates.io crates (private data should be anonymous and stored separately once we have established the ability to gather that in a statistically relevant volume)
• integrate with docs.rs to have it gather metrics and upload them to the backend while building newly uploaded crate versions

[yaahc]

As mentioned in rust-lang/goals#260 (comment), I need to have more detailed discussions with t-libs and t-lang to discuss specifics of unstable feature usage metrics and the precise questions t-libs and t-lang would likely like to use this data to answer.

@rustbot labels +I-libs-nominated +T-libs +I-lang-nominated +T-lang

Related conversation: https://rust-lang.zulipchat.com/#narrow/channel/131828-t-compiler/topic/unstable.20feature.20usage.20metrics