Idiomatic is not pedantic-clean

Running clippy -W clippy::pedantic -W clippy::nursery on Spotuify produced 2014 warnings. The instinct is to treat that as 2014 problems. It's closer to 150 problems and 1850 lines of opinion you should switch off. Equating "idiomatic" with "zero pedantic warnings" is how you turn a clean codebase into a giant churny diff that catches no bugs.

The clippy groups, by whether to trust them

• correctness — deny. Curated to be false-positive-free. Non-negotiable.
• suspicious, perf — enforce, with a local #[allow] + comment when you genuinely mean it.
• complexity, style — mostly enforce; allow case-by-case where the "simpler" form is actually less clear.
• pedantic — opt-in, and the false positives are intentional. Turn it on to read the suggestions, then #[allow] liberally. Never blanket--D it.
• nursery — unstable, sometimes buggy. Cherry-pick individual lints, never enable the group.
• restriction — mutually contradictory by design. Cherry-pick a few good ones (unwrap_used, dbg_macro, todo). Enabling the group makes clippy itself warn you not to.

The noise, by name

About 90% of that 2014 was the lints every mature project allows:

• missing_errors_doc / missing_panics_doc (285) — wants a # Errors section on every fallible fn. Fine for a published library, pure ceremony for an internal crate.
• doc_markdown / missing-backticks (237) — flags your own product names and acronyms.
• must_use_candidate (179) — wants #[must_use] on basically every getter.
• cast_possible_truncation and friends (~165) — fires on intentional casts.
• missing_const_for_fn (118) — nursery, noisy.

And the one that looks scary but isn't: 287 unwrap() warnings, every single one inside #[cfg(test)] or a tests/ dir. unwrap in tests is correct — a failed unwrap is the test failure. The fix is not to remove them; it's to gate the lint by target.

Gate unwrap by target, don't ban it

The pattern that lets you forbid unwrap in production while keeping it in tests, straight from spotuify's CI:

# production code: strict
cargo clippy --workspace --locked -- -D warnings
# test code: same strictness, minus the test-friendly lints
cargo clippy --workspace --tests --locked -- -D warnings \
  -A clippy::panic -A clippy::unwrap_used

Two passes. The -A flags on the second are why the 287 test unwraps aren't failures. A rubric that counts them as violations is measuring the wrong thing.

The signal that's left

Strip the noise and ~150 warnings remain that are worth fixing: redundant_clone, redundant_closure, map_unwrap_or (prefer map_or), uninlined_format_args, needless_pass_by_value (especially &PathBuf where &Path belongs), match_same_arms, unnecessary_wraps, unused_async. These are mechanical, behaviour-preserving, and clippy will often --fix them for you. That's the list a rubric should enforce — and "enforce" means ratchet each one into [workspace.lints] so it stays fixed, not just fix it once (Ratchet Lints Into the Build). The same sweep on Mxr (4640 warnings, 26 crates) boiled down to 11 lints worth gating CI on. See Idiomatic Rust Rubric for where this sits in the bigger picture.

See also

• Idiomatic Rust Rubric — the rubric this lint policy plugs into
• Ratchet Lints Into the Build — how to make the ~150 real lints stay fixed
• Rust Project Conventions — where [workspace.lints] levels vs clippy.toml parameters are explained
• The clippy lint list: https://rust-lang.github.io/rust-clippy/