refactor: Introduce CacheKey trait (#3323)

This PR introduces a new `CacheKey` trait for types that can be used as a cache key.

I'm not entirely sure if this is worth the "overhead", but I was surprised to find `HashableHashSet` and got scared when I looked at the time complexity of the `hash` function. These implementations must be extremely slow in hashed collections.

I then searched for usages and quickly realized that only the cache uses these `Hash` implementations, where performance is less sensitive.

This PR introduces a new `CacheKey` trait to communicate the difference between a hash and computing a key for the cache. The new trait can be implemented for types that don't implement `Hash` for performance reasons, and we can define additional constraints on the implementation:  For example, we'll want to enforce portability when we add remote caching support. Using a different trait further allows us not to implement it for types without stable identities (e.g. pointers) or use other implementations than the standard hash function.

Cargo.lock

@@ -1982,6 +1982,7 @@ dependencies = [
  "path-absolutize",
  "regex",
  "result-like",
+ "ruff_cache",
  "ruff_macros",
  "ruff_python",
  "ruff_rustpython",
@@ -2005,6 +2006,17 @@ dependencies = [
  "wasm-bindgen-test",
 ]
 
+[[package]]
+name = "ruff_cache"
+version = "0.0.0"
+dependencies = [
+ "filetime",
+ "globset",
+ "itertools",
+ "regex",
+ "ruff_macros",
+]
+
 [[package]]
 name = "ruff_cli"
 version = "0.0.253"
@@ -2033,6 +2045,7 @@ dependencies = [
  "rayon",
  "regex",
  "ruff",
+ "ruff_cache",
  "rustc-hash",
  "serde",
  "serde_json",

crates/ruff/Cargo.toml

@@ -19,6 +19,7 @@ doctest = false
 ruff_macros = { path = "../ruff_macros" }
 ruff_python = { path = "../ruff_python" }
 ruff_rustpython = { path = "../ruff_rustpython" }
+ruff_cache = { path = "../ruff_cache" }
 
 anyhow = { workspace = true }
 bisection = { version = "0.1.0" }

crates/ruff/src/fs.rs

@@ -1,13 +1,12 @@
-use std::ops::Deref;
 use std::path::{Path, PathBuf};
 
 use anyhow::{anyhow, Result};
+use globset::GlobMatcher;
 use log::debug;
 use path_absolutize::{path_dedot, Absolutize};
 use rustc_hash::FxHashSet;
 
 use crate::registry::Rule;
-use crate::settings::hashable::{HashableGlobMatcher, HashableHashSet};
 
 /// Extract the absolute path and basename (as strings) from a Path.
 pub fn extract_path_names(path: &Path) -> Result<(&str, &str)> {
@@ -25,11 +24,7 @@ pub fn extract_path_names(path: &Path) -> Result<(&str, &str)> {
 /// Create a set with codes matching the pattern/code pairs.
 pub(crate) fn ignores_from_path<'a>(
     path: &Path,
-    pattern_code_pairs: &'a [(
+    pattern_code_pairs: &'a [(GlobMatcher, GlobMatcher, FxHashSet<Rule>)],
-        HashableGlobMatcher,
-        HashableGlobMatcher,
-        HashableHashSet<Rule>,
-    )],
 ) -> FxHashSet<&'a Rule> {
     let (file_path, file_basename) = extract_path_names(path).expect("Unable to parse filename");
     pattern_code_pairs
@@ -39,8 +34,8 @@ pub(crate) fn ignores_from_path<'a>(
                 debug!(
                     "Adding per-file ignores for {:?} due to basename match on {:?}: {:?}",
                     path,
-                    basename.deref().glob().regex(),
+                    basename.glob().regex(),
-                    &**codes
+                    codes
                 );
                 return Some(codes.iter());
             }
@@ -48,8 +43,8 @@ pub(crate) fn ignores_from_path<'a>(
                 debug!(
                     "Adding per-file ignores for {:?} due to absolute match on {:?}: {:?}",
                     path,
-                    absolute.deref().glob().regex(),
+                    absolute.glob().regex(),
-                    &**codes
+                    codes
                 );
                 return Some(codes.iter());
             }

crates/ruff/src/lib.rs

@@ -13,7 +13,6 @@ pub use violation::{AutofixKind, Availability as AutofixAvailability};
 
 mod ast;
 mod autofix;
-pub mod cache;
 mod checkers;
 mod codes;
 mod cst;

crates/ruff/src/rules/flake8_annotations/settings.rs

@@ -1,5 +1,6 @@
 //! Settings for the `flake-annotations` plugin.
 
+use ruff_macros::CacheKey;
 use ruff_macros::ConfigurationOptions;
 use schemars::JsonSchema;
 use serde::{Deserialize, Serialize};
@@ -60,7 +61,7 @@ pub struct Options {
     pub ignore_fully_untyped: Option<bool>,
 }
 
-#[derive(Debug, Default, Hash)]
+#[derive(Debug, Default, CacheKey)]
 #[allow(clippy::struct_excessive_bools)]
 pub struct Settings {
     pub mypy_init_return: bool,

crates/ruff/src/rules/flake8_bandit/settings.rs

@@ -1,6 +1,6 @@
 //! Settings for the `flake8-bandit` plugin.
 
-use ruff_macros::ConfigurationOptions;
+use ruff_macros::{CacheKey, ConfigurationOptions};
 use schemars::JsonSchema;
 use serde::{Deserialize, Serialize};
 
@@ -45,7 +45,7 @@ pub struct Options {
     pub check_typed_exception: Option<bool>,
 }
 
-#[derive(Debug, Hash)]
+#[derive(Debug, CacheKey)]
 pub struct Settings {
     pub hardcoded_tmp_directory: Vec<String>,
     pub check_typed_exception: bool,

crates/ruff/src/rules/flake8_bugbear/settings.rs

@@ -1,6 +1,6 @@
 //! Settings for the `flake8-bugbear` plugin.
 
-use ruff_macros::ConfigurationOptions;
+use ruff_macros::{CacheKey, ConfigurationOptions};
 use schemars::JsonSchema;
 use serde::{Deserialize, Serialize};
 
@@ -26,7 +26,7 @@ pub struct Options {
     pub extend_immutable_calls: Option<Vec<String>>,
 }
 
-#[derive(Debug, Default, Hash)]
+#[derive(Debug, Default, CacheKey)]
 pub struct Settings {
     pub extend_immutable_calls: Vec<String>,
 }

crates/ruff/src/rules/flake8_builtins/settings.rs

@@ -1,6 +1,6 @@
 //! Settings for the `flake8-builtins` plugin.
 
-use ruff_macros::ConfigurationOptions;
+use ruff_macros::{CacheKey, ConfigurationOptions};
 use schemars::JsonSchema;
 use serde::{Deserialize, Serialize};
 
@@ -22,7 +22,7 @@ pub struct Options {
     pub builtins_ignorelist: Option<Vec<String>>,
 }
 
-#[derive(Debug, Default, Hash)]
+#[derive(Debug, Default, CacheKey)]
 pub struct Settings {
     pub builtins_ignorelist: Vec<String>,
 }

crates/ruff/src/rules/flake8_comprehensions/settings.rs

@@ -1,6 +1,6 @@
 //! Settings for the `flake8-comprehensions` plugin.
 
-use ruff_macros::ConfigurationOptions;
+use ruff_macros::{CacheKey, ConfigurationOptions};
 use schemars::JsonSchema;
 use serde::{Deserialize, Serialize};
 
@@ -22,7 +22,7 @@ pub struct Options {
     pub allow_dict_calls_with_keyword_arguments: Option<bool>,
 }
 
-#[derive(Debug, Default, Hash)]
+#[derive(Debug, Default, CacheKey)]
 pub struct Settings {
     pub allow_dict_calls_with_keyword_arguments: bool,
 }

crates/ruff/src/rules/flake8_errmsg/settings.rs

@@ -1,6 +1,6 @@
 //! Settings for the `flake8-errmsg` plugin.
 
-use ruff_macros::ConfigurationOptions;
+use ruff_macros::{CacheKey, ConfigurationOptions};
 use schemars::JsonSchema;
 use serde::{Deserialize, Serialize};
 
@@ -18,7 +18,7 @@ pub struct Options {
     pub max_string_length: Option<usize>,
 }
 
-#[derive(Debug, Default, Hash)]
+#[derive(Debug, Default, CacheKey)]
 pub struct Settings {
     pub max_string_length: usize,
 }

crates/ruff/src/rules/flake8_implicit_str_concat/settings.rs

@@ -1,6 +1,6 @@
 //! Settings for the `flake8-implicit-str-concat` plugin.
 
-use ruff_macros::ConfigurationOptions;
+use ruff_macros::{CacheKey, ConfigurationOptions};
 use schemars::JsonSchema;
 use serde::{Deserialize, Serialize};
 
@@ -32,7 +32,7 @@ pub struct Options {
     pub allow_multiline: Option<bool>,
 }
 
-#[derive(Debug, Hash)]
+#[derive(Debug, CacheKey)]
 pub struct Settings {
     pub allow_multiline: bool,
 }

crates/ruff/src/rules/flake8_import_conventions/settings.rs

@@ -1,14 +1,10 @@
 //! Settings for import conventions.
 
-use std::hash::Hash;
+use ruff_macros::{CacheKey, ConfigurationOptions};
-
-use ruff_macros::ConfigurationOptions;
 use rustc_hash::FxHashMap;
 use schemars::JsonSchema;
 use serde::{Deserialize, Serialize};
 
-use crate::settings::hashable::HashableHashMap;
-
 const CONVENTIONAL_ALIASES: &[(&str, &str)] = &[
     ("altair", "alt"),
     ("matplotlib", "mpl"),
@@ -64,9 +60,9 @@ pub struct Options {
     pub extend_aliases: Option<FxHashMap<String, String>>,
 }
 
-#[derive(Debug, Hash)]
+#[derive(Debug, CacheKey)]
 pub struct Settings {
-    pub aliases: HashableHashMap<String, String>,
+    pub aliases: FxHashMap<String, String>,
 }
 
 fn default_aliases() -> FxHashMap<String, String> {
@@ -90,7 +86,7 @@ fn resolve_aliases(options: Options) -> FxHashMap<String, String> {
 impl Default for Settings {
     fn default() -> Self {
         Self {
-            aliases: default_aliases().into(),
+            aliases: default_aliases(),
         }
     }
 }
@@ -98,7 +94,7 @@ impl Default for Settings {
 impl From<Options> for Settings {
     fn from(options: Options) -> Self {
         Self {
-            aliases: resolve_aliases(options).into(),
+            aliases: resolve_aliases(options),
         }
     }
 }
@@ -106,7 +102,7 @@ impl From<Options> for Settings {
 impl From<Settings> for Options {
     fn from(settings: Settings) -> Self {
         Self {
-            aliases: Some(settings.aliases.into()),
+            aliases: Some(settings.aliases),
             extend_aliases: None,
         }
     }

crates/ruff/src/rules/flake8_pytest_style/settings.rs

@@ -1,6 +1,6 @@
 //! Settings for the `flake8-pytest-style` plugin.
 
-use ruff_macros::ConfigurationOptions;
+use ruff_macros::{CacheKey, ConfigurationOptions};
 use schemars::JsonSchema;
 use serde::{Deserialize, Serialize};
 
@@ -113,7 +113,7 @@ pub struct Options {
     pub mark_parentheses: Option<bool>,
 }
 
-#[derive(Debug, Hash)]
+#[derive(Debug, CacheKey)]
 pub struct Settings {
     pub fixture_parentheses: bool,
     pub parametrize_names_type: types::ParametrizeNameType,

crates/ruff/src/rules/flake8_pytest_style/types.rs

@@ -1,9 +1,10 @@
 use std::fmt::{Display, Formatter};
 
+use ruff_macros::CacheKey;
 use schemars::JsonSchema;
 use serde::{Deserialize, Serialize};
 
-#[derive(Clone, Copy, Debug, Hash, PartialEq, Eq, Serialize, Deserialize, JsonSchema)]
+#[derive(Clone, Copy, Debug, CacheKey, PartialEq, Eq, Serialize, Deserialize, JsonSchema)]
 pub enum ParametrizeNameType {
     #[serde(rename = "csv")]
     Csv,
@@ -29,7 +30,7 @@ impl Display for ParametrizeNameType {
     }
 }
 
-#[derive(Clone, Copy, Debug, Hash, PartialEq, Eq, Serialize, Deserialize, JsonSchema)]
+#[derive(Clone, Copy, Debug, CacheKey, PartialEq, Eq, Serialize, Deserialize, JsonSchema)]
 pub enum ParametrizeValuesType {
     #[serde(rename = "tuple")]
     Tuple,
@@ -52,7 +53,7 @@ impl Display for ParametrizeValuesType {
     }
 }
 
-#[derive(Clone, Copy, Debug, Hash, PartialEq, Eq, Serialize, Deserialize, JsonSchema)]
+#[derive(Clone, Copy, Debug, CacheKey, PartialEq, Eq, Serialize, Deserialize, JsonSchema)]
 pub enum ParametrizeValuesRowType {
     #[serde(rename = "tuple")]
     Tuple,

crates/ruff/src/rules/flake8_quotes/settings.rs

@@ -1,10 +1,10 @@
 //! Settings for the `flake8-quotes` plugin.
 
-use ruff_macros::ConfigurationOptions;
+use ruff_macros::{CacheKey, ConfigurationOptions};
 use schemars::JsonSchema;
 use serde::{Deserialize, Serialize};
 
-#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize, Hash, JsonSchema)]
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize, CacheKey, JsonSchema)]
 #[serde(deny_unknown_fields, rename_all = "kebab-case")]
 pub enum Quote {
     /// Use single quotes.
@@ -71,7 +71,7 @@ pub struct Options {
     pub avoid_escape: Option<bool>,
 }
 
-#[derive(Debug, Hash)]
+#[derive(Debug, CacheKey)]
 pub struct Settings {
     pub inline_quotes: Quote,
     pub multiline_quotes: Quote,

crates/ruff/src/rules/flake8_self/settings.rs

@@ -1,6 +1,6 @@
 //! Settings for the `flake8-self` plugin.
 
-use ruff_macros::ConfigurationOptions;
+use ruff_macros::{CacheKey, ConfigurationOptions};
 use schemars::JsonSchema;
 use serde::{Deserialize, Serialize};
 
@@ -28,7 +28,7 @@ pub struct Options {
     pub ignore_names: Option<Vec<String>>,
 }
 
-#[derive(Debug, Hash)]
+#[derive(Debug, CacheKey)]
 pub struct Settings {
     pub ignore_names: Vec<String>,
 }

crates/ruff/src/rules/flake8_tidy_imports/banned_api.rs

@@ -1,4 +1,4 @@
-use ruff_macros::{define_violation, derive_message_formats};
+use ruff_macros::{define_violation, derive_message_formats, CacheKey};
 use rustc_hash::FxHashMap;
 use rustpython_parser::ast::{Alias, Expr, Located};
 use schemars::JsonSchema;
@@ -7,12 +7,11 @@ use serde::{Deserialize, Serialize};
 use crate::ast::types::{CallPath, Range};
 use crate::checkers::ast::Checker;
 use crate::registry::Diagnostic;
-use crate::settings::hashable::HashableHashMap;
 use crate::violation::Violation;
 
-pub type Settings = HashableHashMap<String, ApiBan>;
+pub type Settings = FxHashMap<String, ApiBan>;
 
-#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize, Hash, JsonSchema)]
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize, CacheKey, JsonSchema)]
 #[serde(deny_unknown_fields, rename_all = "kebab-case")]
 pub struct ApiBan {
     /// The message to display when the API is used.
@@ -147,8 +146,7 @@ mod tests {
                                 msg: "Use typing_extensions.TypedDict instead.".to_string(),
                             },
                         ),
-                    ])
+                    ]),
-                    .into(),
                     ..Default::default()
                 },
                 ..Settings::for_rules(vec![Rule::BannedApi])

crates/ruff/src/rules/flake8_tidy_imports/mod.rs

@@ -1,10 +1,12 @@
 //! Rules from [flake8-tidy-imports](https://pypi.org/project/flake8-tidy-imports/).
+use ruff_macros::CacheKey;
+
 pub mod options;
 
 pub mod banned_api;
 pub mod relative_imports;
 
-#[derive(Debug, Hash, Default)]
+#[derive(Debug, CacheKey, Default)]
 pub struct Settings {
     pub ban_relative_imports: relative_imports::Settings,
     pub banned_api: banned_api::Settings,

crates/ruff/src/rules/flake8_tidy_imports/options.rs

@@ -48,7 +48,7 @@ impl From<Options> for Settings {
     fn from(options: Options) -> Self {
         Self {
             ban_relative_imports: options.ban_relative_imports.unwrap_or(Strictness::Parents),
-            banned_api: options.banned_api.unwrap_or_default().into(),
+            banned_api: options.banned_api.unwrap_or_default(),
         }
     }
 }
@@ -57,7 +57,7 @@ impl From<Settings> for Options {
     fn from(settings: Settings) -> Self {
         Self {
             ban_relative_imports: Some(settings.ban_relative_imports),
-            banned_api: Some(settings.banned_api.into()),
+            banned_api: Some(settings.banned_api),
         }
     }
 }

crates/ruff/src/rules/flake8_tidy_imports/relative_imports.rs

@@ -2,7 +2,7 @@ use rustpython_parser::ast::{Stmt, StmtKind};
 use schemars::JsonSchema;
 use serde::{Deserialize, Serialize};
 
-use ruff_macros::{define_violation, derive_message_formats};
+use ruff_macros::{define_violation, derive_message_formats, CacheKey};
 use ruff_python::identifiers::is_module_name;
 
 use crate::ast::helpers::{create_stmt, from_relative_import, unparse_stmt};
@@ -15,7 +15,7 @@ use crate::violation::{AutofixKind, Availability, Violation};
 
 pub type Settings = Strictness;
 
-#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize, Hash, JsonSchema, Default)]
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize, CacheKey, JsonSchema, Default)]
 #[serde(deny_unknown_fields, rename_all = "kebab-case")]
 pub enum Strictness {
     /// Ban imports that extend into the parent module or beyond.

crates/ruff/src/rules/flake8_type_checking/settings.rs

@@ -1,6 +1,6 @@
 //! Settings for the `flake8-type-checking` plugin.
 
-use ruff_macros::ConfigurationOptions;
+use ruff_macros::{CacheKey, ConfigurationOptions};
 use schemars::JsonSchema;
 use serde::{Deserialize, Serialize};
 
@@ -36,7 +36,7 @@ pub struct Options {
     pub exempt_modules: Option<Vec<String>>,
 }
 
-#[derive(Debug, Hash)]
+#[derive(Debug, CacheKey)]
 pub struct Settings {
     pub strict: bool,
     pub exempt_modules: Vec<String>,

crates/ruff/src/rules/flake8_unused_arguments/settings.rs

@@ -1,6 +1,6 @@
 //! Settings for the `flake8-unused-arguments` plugin.
 
-use ruff_macros::ConfigurationOptions;
+use ruff_macros::{CacheKey, ConfigurationOptions};
 use schemars::JsonSchema;
 use serde::{Deserialize, Serialize};
 
@@ -22,7 +22,7 @@ pub struct Options {
     pub ignore_variadic_names: Option<bool>,
 }
 
-#[derive(Debug, Default, Hash)]
+#[derive(Debug, Default, CacheKey)]
 pub struct Settings {
     pub ignore_variadic_names: bool,
 }

crates/ruff/src/rules/isort/categorize.rs

@@ -3,6 +3,7 @@ use std::fs;
 use std::path::{Path, PathBuf};
 
 use log::debug;
+use ruff_macros::CacheKey;
 use ruff_python::sys::KNOWN_STANDARD_LIBRARY;
 use schemars::JsonSchema;
 use serde::{Deserialize, Serialize};
@@ -12,7 +13,17 @@ use super::types::{ImportBlock, Importable};
 use crate::settings::types::PythonVersion;
 
 #[derive(
-    Debug, PartialOrd, Ord, PartialEq, Eq, Clone, Serialize, Deserialize, JsonSchema, Hash, EnumIter,
+    Debug,
+    PartialOrd,
+    Ord,
+    PartialEq,
+    Eq,
+    Clone,
+    Serialize,
+    Deserialize,
+    JsonSchema,
+    CacheKey,
+    EnumIter,
 )]
 #[serde(deny_unknown_fields, rename_all = "kebab-case")]
 pub enum ImportType {

crates/ruff/src/rules/isort/settings.rs

@@ -2,13 +2,13 @@
 
 use std::collections::BTreeSet;
 
-use ruff_macros::ConfigurationOptions;
+use ruff_macros::{CacheKey, ConfigurationOptions};
 use schemars::JsonSchema;
 use serde::{Deserialize, Serialize};
 
 use super::categorize::ImportType;
 
-#[derive(Debug, Copy, Clone, PartialEq, Eq, Serialize, Deserialize, Hash, JsonSchema)]
+#[derive(Debug, Copy, Clone, PartialEq, Eq, Serialize, Deserialize, CacheKey, JsonSchema)]
 #[serde(deny_unknown_fields, rename_all = "kebab-case")]
 pub enum RelativeImportsOrder {
     /// Place "closer" imports (fewer `.` characters, most local) before
@@ -265,7 +265,7 @@ pub struct Options {
     pub forced_separate: Option<Vec<String>>,
 }
 
-#[derive(Debug, Hash)]
+#[derive(Debug, CacheKey)]
 #[allow(clippy::struct_excessive_bools)]
 pub struct Settings {
     pub required_imports: BTreeSet<String>,

crates/ruff/src/rules/mccabe/settings.rs

@@ -1,6 +1,6 @@
 //! Settings for the `mccabe` plugin.
 
-use ruff_macros::ConfigurationOptions;
+use ruff_macros::{CacheKey, ConfigurationOptions};
 use schemars::JsonSchema;
 use serde::{Deserialize, Serialize};
 
@@ -25,7 +25,7 @@ pub struct Options {
     pub max_complexity: Option<usize>,
 }
 
-#[derive(Debug, Hash)]
+#[derive(Debug, CacheKey)]
 pub struct Settings {
     pub max_complexity: usize,
 }

crates/ruff/src/rules/pep8_naming/settings.rs

@@ -1,6 +1,6 @@
 //! Settings for the `pep8-naming` plugin.
 
-use ruff_macros::ConfigurationOptions;
+use ruff_macros::{CacheKey, ConfigurationOptions};
 use schemars::JsonSchema;
 use serde::{Deserialize, Serialize};
 
@@ -69,7 +69,7 @@ pub struct Options {
     pub staticmethod_decorators: Option<Vec<String>>,
 }
 
-#[derive(Debug, Hash)]
+#[derive(Debug, CacheKey)]
 pub struct Settings {
     pub ignore_names: Vec<String>,
     pub classmethod_decorators: Vec<String>,

crates/ruff/src/rules/pycodestyle/settings.rs

@@ -1,6 +1,6 @@
 //! Settings for the `pycodestyle` plugin.
 
-use ruff_macros::ConfigurationOptions;
+use ruff_macros::{CacheKey, ConfigurationOptions};
 use schemars::JsonSchema;
 use serde::{Deserialize, Serialize};
 
@@ -32,7 +32,7 @@ pub struct Options {
     pub ignore_overlong_task_comments: Option<bool>,
 }
 
-#[derive(Debug, Default, Hash)]
+#[derive(Debug, Default, CacheKey)]
 pub struct Settings {
     pub max_doc_length: Option<usize>,
     pub ignore_overlong_task_comments: bool,

crates/ruff/src/rules/pydocstyle/settings.rs

@@ -2,13 +2,13 @@
 
 use std::collections::BTreeSet;
 
-use ruff_macros::ConfigurationOptions;
+use ruff_macros::{CacheKey, ConfigurationOptions};
 use schemars::JsonSchema;
 use serde::{Deserialize, Serialize};
 
 use crate::registry::Rule;
 
-#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize, Hash, JsonSchema)]
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize, JsonSchema, CacheKey)]
 #[serde(deny_unknown_fields, rename_all = "kebab-case")]
 pub enum Convention {
     /// Use Google-style docstrings.
@@ -112,7 +112,7 @@ pub struct Options {
     pub property_decorators: Option<Vec<String>>,
 }
 
-#[derive(Debug, Default, Hash)]
+#[derive(Debug, Default, CacheKey)]
 pub struct Settings {
     pub convention: Option<Convention>,
     pub ignore_decorators: BTreeSet<String>,

crates/ruff/src/rules/pyflakes/mod.rs

@@ -129,7 +129,7 @@ mod tests {
         let diagnostics = test_path(
             Path::new("pyflakes/F841_0.py"),
             &settings::Settings {
-                dummy_variable_rgx: Regex::new(r"^z$").unwrap().into(),
+                dummy_variable_rgx: Regex::new(r"^z$").unwrap(),
                 ..settings::Settings::for_rule(Rule::UnusedVariable)
             },
         )?;

crates/ruff/src/rules/pylint/mod.rs

@@ -99,7 +99,7 @@ mod tests {
         let diagnostics = test_path(
             Path::new("pylint/too_many_arguments_params.py"),
             &Settings {
-                dummy_variable_rgx: Regex::new(r"skip_.*").unwrap().into(),
+                dummy_variable_rgx: Regex::new(r"skip_.*").unwrap(),
                 ..Settings::for_rules(vec![Rule::TooManyArguments])
             },
         )?;

crates/ruff/src/rules/pylint/rules/redefined_loop_name.rs

@@ -1,3 +1,4 @@
+use regex::Regex;
 use std::{fmt, iter};
 
 use rustpython_parser::ast::{Expr, ExprContext, ExprKind, Stmt, StmtKind, Withitem};
@@ -12,7 +13,6 @@ use crate::ast::visitor;
 use crate::ast::visitor::Visitor;
 use crate::checkers::ast::Checker;
 use crate::registry::Diagnostic;
-use crate::settings::hashable::HashableRegex;
 use crate::violation::Violation;
 
 #[derive(Debug, PartialEq, Eq, Serialize, Deserialize, Clone, Copy)]
@@ -142,7 +142,7 @@ struct ExprWithInnerBindingKind<'a> {
 }
 
 struct InnerForWithAssignTargetsVisitor<'a> {
-    dummy_variable_rgx: &'a HashableRegex,
+    dummy_variable_rgx: &'a Regex,
     assignment_targets: Vec<ExprWithInnerBindingKind<'a>>,
 }
 
@@ -213,7 +213,7 @@ where
 
 fn assignment_targets_from_expr<'a, U>(
     expr: &'a Expr<U>,
-    dummy_variable_rgx: &'a HashableRegex,
+    dummy_variable_rgx: &'a Regex,
 ) -> Box<dyn Iterator<Item = &'a Expr<U>> + 'a> {
     // The Box is necessary to ensure the match arms have the same return type - we can't use
     // a cast to "impl Iterator", since at the time of writing that is only allowed for
@@ -266,7 +266,7 @@ fn assignment_targets_from_expr<'a, U>(
 
 fn assignment_targets_from_with_items<'a, U>(
     items: &'a [Withitem<U>],
-    dummy_variable_rgx: &'a HashableRegex,
+    dummy_variable_rgx: &'a Regex,
 ) -> impl Iterator<Item = &'a Expr<U>> + 'a {
     items
         .iter()
@@ -280,7 +280,7 @@ fn assignment_targets_from_with_items<'a, U>(
 
 fn assignment_targets_from_assign_targets<'a, U>(
     targets: &'a [Expr<U>],
-    dummy_variable_rgx: &'a HashableRegex,
+    dummy_variable_rgx: &'a Regex,
 ) -> impl Iterator<Item = &'a Expr<U>> + 'a {
     targets
         .iter()

crates/ruff/src/rules/pylint/settings.rs

@@ -1,13 +1,11 @@
 //! Settings for the `pylint` plugin.
 
-use std::hash::Hash;
-
 use anyhow::anyhow;
-use ruff_macros::ConfigurationOptions;
+use ruff_macros::{CacheKey, ConfigurationOptions};
 use rustpython_parser::ast::Constant;
 use schemars::JsonSchema;
 use serde::{Deserialize, Serialize};
-#[derive(Debug, Copy, Clone, PartialEq, Eq, Serialize, Deserialize, Hash, JsonSchema)]
+#[derive(Debug, Copy, Clone, PartialEq, Eq, Serialize, Deserialize, CacheKey, JsonSchema)]
 #[serde(deny_unknown_fields, rename_all = "kebab-case")]
 pub enum ConstantType {
     Bytes,
@@ -72,7 +70,7 @@ pub struct Options {
     pub max_statements: Option<usize>,
 }
 
-#[derive(Debug, Hash)]
+#[derive(Debug, CacheKey)]
 pub struct Settings {
     pub allow_magic_value_types: Vec<ConstantType>,
     pub max_args: usize,

crates/ruff/src/rules/pyupgrade/settings.rs

@@ -1,6 +1,6 @@
 //! Settings for the `pyupgrade` plugin.
 
-use ruff_macros::ConfigurationOptions;
+use ruff_macros::{CacheKey, ConfigurationOptions};
 use schemars::JsonSchema;
 use serde::{Deserialize, Serialize};
 
@@ -31,7 +31,7 @@ pub struct Options {
     pub keep_runtime_typing: Option<bool>,
 }
 
-#[derive(Debug, Default, Hash)]
+#[derive(Debug, Default, CacheKey)]
 pub struct Settings {
     pub keep_runtime_typing: bool,
 }

crates/ruff/src/rules/ruff/mod.rs

@@ -34,7 +34,7 @@ mod tests {
         let diagnostics = test_path(
             Path::new("ruff/confusables.py"),
             &settings::Settings {
-                allowed_confusables: FxHashSet::from_iter(['−', 'ρ', '∗']).into(),
+                allowed_confusables: FxHashSet::from_iter(['−', 'ρ', '∗']),
                 ..settings::Settings::for_rules(vec![
                     Rule::AmbiguousUnicodeCharacterString,
                     Rule::AmbiguousUnicodeCharacterDocstring,

crates/ruff/src/settings/defaults.rs

@@ -2,8 +2,8 @@ use once_cell::sync::Lazy;
 use path_absolutize::path_dedot;
 use regex::Regex;
 use rustc_hash::FxHashSet;
+use std::collections::HashSet;
 
-use super::hashable::{HashableGlobSet, HashableHashSet};
 use super::types::{FilePattern, PythonVersion};
 use super::Settings;
 use crate::codes::{self, RuleCodePrefix};
@@ -15,6 +15,7 @@ use crate::rules::{
     flake8_quotes, flake8_self, flake8_tidy_imports, flake8_type_checking, flake8_unused_arguments,
     isort, mccabe, pep8_naming, pycodestyle, pydocstyle, pylint, pyupgrade,
 };
+use crate::settings::types::FilePatternSet;
 
 pub const PREFIXES: &[RuleSelector] = &[
     prefix_to_selector(RuleCodePrefix::Pycodestyle(codes::Pycodestyle::E)),
@@ -59,12 +60,12 @@ impl Default for Settings {
     fn default() -> Self {
         Self {
             rules: PREFIXES.iter().flat_map(IntoIterator::into_iter).into(),
-            allowed_confusables: FxHashSet::from_iter([]).into(),
+            allowed_confusables: FxHashSet::from_iter([]),
             builtins: vec![],
-            dummy_variable_rgx: DUMMY_VARIABLE_RGX.clone().into(),
+            dummy_variable_rgx: DUMMY_VARIABLE_RGX.clone(),
-            exclude: HashableGlobSet::new(EXCLUDE.clone()).unwrap(),
+            exclude: FilePatternSet::try_from_vec(EXCLUDE.clone()).unwrap(),
-            extend_exclude: HashableGlobSet::empty(),
+            extend_exclude: FilePatternSet::default(),
-            external: HashableHashSet::default(),
+            external: HashSet::default(),
             force_exclude: false,
             ignore_init_module_imports: false,
             line_length: LINE_LENGTH,

crates/ruff/src/settings/flags.rs

@@ -1,6 +1,7 @@
 use crate::fix;
+use ruff_macros::CacheKey;
 
-#[derive(Debug, Copy, Clone, Hash, result_like::BoolLike)]
+#[derive(Debug, Copy, Clone, CacheKey, result_like::BoolLike)]
 pub enum Autofix {
     Enabled,
     Disabled,

crates/ruff/src/settings/hashable.rs

@@ -1,179 +0,0 @@
-use derivative::Derivative;
-use std::hash::{Hash, Hasher};
-use std::ops::{Deref, DerefMut};
-
-use globset::{GlobMatcher, GlobSet};
-use itertools::Itertools;
-use regex::Regex;
-use rustc_hash::{FxHashMap, FxHashSet};
-
-use super::types::FilePattern;
-
-#[derive(Debug)]
-pub struct HashableRegex(Regex);
-
-impl Hash for HashableRegex {
-    fn hash<H: Hasher>(&self, state: &mut H) {
-        self.0.as_str().hash(state);
-    }
-}
-
-impl From<Regex> for HashableRegex {
-    fn from(regex: Regex) -> Self {
-        Self(regex)
-    }
-}
-
-impl Deref for HashableRegex {
-    type Target = Regex;
-
-    fn deref(&self) -> &Self::Target {
-        &self.0
-    }
-}
-
-#[derive(Debug)]
-pub struct HashableGlobMatcher(GlobMatcher);
-
-impl From<GlobMatcher> for HashableGlobMatcher {
-    fn from(matcher: GlobMatcher) -> Self {
-        Self(matcher)
-    }
-}
-
-impl Deref for HashableGlobMatcher {
-    type Target = GlobMatcher;
-
-    fn deref(&self) -> &Self::Target {
-        &self.0
-    }
-}
-
-impl Hash for HashableGlobMatcher {
-    fn hash<H: Hasher>(&self, state: &mut H) {
-        self.0.glob().hash(state);
-    }
-}
-
-#[derive(Derivative)]
-#[derivative(Debug)]
-pub struct HashableGlobSet {
-    patterns: Vec<FilePattern>,
-    #[derivative(Debug = "ignore")]
-    globset: GlobSet,
-}
-
-impl HashableGlobSet {
-    pub fn new(patterns: Vec<FilePattern>) -> anyhow::Result<Self> {
-        let mut builder = globset::GlobSetBuilder::new();
-        for pattern in &patterns {
-            pattern.clone().add_to(&mut builder)?;
-        }
-        let globset = builder.build()?;
-        Ok(HashableGlobSet { patterns, globset })
-    }
-
-    pub fn empty() -> Self {
-        Self {
-            patterns: Vec::new(),
-            globset: GlobSet::empty(),
-        }
-    }
-}
-
-impl Deref for HashableGlobSet {
-    type Target = GlobSet;
-
-    fn deref(&self) -> &Self::Target {
-        &self.globset
-    }
-}
-
-impl Hash for HashableGlobSet {
-    fn hash<H: Hasher>(&self, state: &mut H) {
-        for pattern in self.patterns.iter().sorted() {
-            pattern.hash(state);
-        }
-    }
-}
-
-#[derive(Debug, Clone)]
-pub struct HashableHashSet<T>(FxHashSet<T>);
-
-impl<T: Hash + Ord> Hash for HashableHashSet<T> {
-    fn hash<H: Hasher>(&self, state: &mut H) {
-        for v in self.0.iter().sorted() {
-            v.hash(state);
-        }
-    }
-}
-
-impl<T> Default for HashableHashSet<T> {
-    fn default() -> Self {
-        Self(FxHashSet::default())
-    }
-}
-
-impl<T> From<FxHashSet<T>> for HashableHashSet<T> {
-    fn from(set: FxHashSet<T>) -> Self {
-        Self(set)
-    }
-}
-
-impl<T> From<HashableHashSet<T>> for FxHashSet<T> {
-    fn from(set: HashableHashSet<T>) -> Self {
-        set.0
-    }
-}
-
-impl<T> Deref for HashableHashSet<T> {
-    type Target = FxHashSet<T>;
-
-    fn deref(&self) -> &Self::Target {
-        &self.0
-    }
-}
-
-#[derive(Debug, Clone)]
-pub struct HashableHashMap<K, V>(FxHashMap<K, V>);
-
-impl<K: Hash + Ord, V: Hash> Hash for HashableHashMap<K, V> {
-    fn hash<H: Hasher>(&self, state: &mut H) {
-        for key in self.0.keys().sorted() {
-            key.hash(state);
-            self.0[key].hash(state);
-        }
-    }
-}
-
-impl<K, V> Default for HashableHashMap<K, V> {
-    fn default() -> Self {
-        Self(FxHashMap::default())
-    }
-}
-
-impl<K, V> From<FxHashMap<K, V>> for HashableHashMap<K, V> {
-    fn from(map: FxHashMap<K, V>) -> Self {
-        Self(map)
-    }
-}
-
-impl<K, V> From<HashableHashMap<K, V>> for FxHashMap<K, V> {
-    fn from(map: HashableHashMap<K, V>) -> Self {
-        map.0
-    }
-}
-
-impl<K, V> Deref for HashableHashMap<K, V> {
-    type Target = FxHashMap<K, V>;
-
-    fn deref(&self) -> &Self::Target {
-        &self.0
-    }
-}
-
-impl<K, V> DerefMut for HashableHashMap<K, V> {
-    fn deref_mut(&mut self) -> &mut <Self as Deref>::Target {
-        &mut self.0
-    }
-}

crates/ruff/src/settings/mod.rs

@@ -5,13 +5,13 @@
 use std::path::{Path, PathBuf};
 
 use anyhow::{anyhow, Result};
-use globset::Glob;
+use globset::{Glob, GlobMatcher};
+use regex::Regex;
+use ruff_cache::cache_dir;
 use rustc_hash::{FxHashMap, FxHashSet};
 use strum::IntoEnumIterator;
 
-use self::hashable::{HashableGlobMatcher, HashableGlobSet, HashableHashSet, HashableRegex};
 use self::rule_table::RuleTable;
-use crate::cache::cache_dir;
 use crate::registry::{Rule, RuleNamespace, INCOMPATIBLE_CODES};
 use crate::rule_selector::{RuleSelector, Specificity};
 use crate::rules::{
@@ -21,13 +21,13 @@ use crate::rules::{
     isort, mccabe, pep8_naming, pycodestyle, pydocstyle, pylint, pyupgrade,
 };
 use crate::settings::configuration::Configuration;
-use crate::settings::types::{PerFileIgnore, PythonVersion, SerializationFormat};
+use crate::settings::types::{FilePatternSet, PerFileIgnore, PythonVersion, SerializationFormat};
 use crate::warn_user_once;
+use ruff_macros::CacheKey;
 
 pub mod configuration;
 pub mod defaults;
 pub mod flags;
-pub mod hashable;
 pub mod options;
 pub mod options_base;
 pub mod pyproject;
@@ -74,31 +74,27 @@ pub struct CliSettings {
     pub update_check: bool,
 }
 
-#[derive(Debug, Hash)]
+#[derive(Debug, CacheKey)]
 #[allow(clippy::struct_excessive_bools)]
 pub struct Settings {
     pub rules: RuleTable,
-    pub per_file_ignores: Vec<(
+    pub per_file_ignores: Vec<(GlobMatcher, GlobMatcher, FxHashSet<Rule>)>,
-        HashableGlobMatcher,
-        HashableGlobMatcher,
-        HashableHashSet<Rule>,
-    )>,
 
     pub show_source: bool,
     pub target_version: PythonVersion,
 
     // Resolver settings
-    pub exclude: HashableGlobSet,
+    pub exclude: FilePatternSet,
-    pub extend_exclude: HashableGlobSet,
+    pub extend_exclude: FilePatternSet,
     pub force_exclude: bool,
     pub respect_gitignore: bool,
     pub project_root: PathBuf,
 
     // Rule-specific settings
-    pub allowed_confusables: HashableHashSet<char>,
+    pub allowed_confusables: FxHashSet<char>,
     pub builtins: Vec<String>,
-    pub dummy_variable_rgx: HashableRegex,
+    pub dummy_variable_rgx: Regex,
-    pub external: HashableHashSet<String>,
+    pub external: FxHashSet<String>,
     pub ignore_init_module_imports: bool,
     pub line_length: usize,
     pub namespace_packages: Vec<PathBuf>,
@@ -146,18 +142,16 @@ impl Settings {
             allowed_confusables: config
                 .allowed_confusables
                 .map(FxHashSet::from_iter)
-                .unwrap_or_default()
+                .unwrap_or_default(),
-                .into(),
             builtins: config.builtins.unwrap_or_default(),
             dummy_variable_rgx: config
                 .dummy_variable_rgx
-                .unwrap_or_else(|| defaults::DUMMY_VARIABLE_RGX.clone())
+                .unwrap_or_else(|| defaults::DUMMY_VARIABLE_RGX.clone()),
-                .into(),
+            exclude: FilePatternSet::try_from_vec(
-            exclude: HashableGlobSet::new(
                 config.exclude.unwrap_or_else(|| defaults::EXCLUDE.clone()),
             )?,
-            extend_exclude: HashableGlobSet::new(config.extend_exclude)?,
+            extend_exclude: FilePatternSet::try_from_vec(config.extend_exclude)?,
-            external: FxHashSet::from_iter(config.external.unwrap_or_default()).into(),
+            external: FxHashSet::from_iter(config.external.unwrap_or_default()),
 
             force_exclude: config.force_exclude.unwrap_or(false),
 
@@ -414,13 +408,7 @@ impl From<&Configuration> for RuleTable {
 /// Given a list of patterns, create a `GlobSet`.
 pub fn resolve_per_file_ignores(
     per_file_ignores: Vec<PerFileIgnore>,
-) -> Result<
+) -> Result<Vec<(GlobMatcher, GlobMatcher, FxHashSet<Rule>)>> {
-    Vec<(
-        HashableGlobMatcher,
-        HashableGlobMatcher,
-        HashableHashSet<Rule>,
-    )>,
-> {
     per_file_ignores
         .into_iter()
         .map(|per_file_ignore| {
@@ -431,7 +419,7 @@ pub fn resolve_per_file_ignores(
             // Construct basename matcher.
             let basename = Glob::new(&per_file_ignore.basename)?.compile_matcher();
 
-            Ok((absolute.into(), basename.into(), per_file_ignore.rules))
+            Ok((absolute, basename, per_file_ignore.rules))
         })
         .collect()
 }

crates/ruff/src/settings/rule_table.rs

@@ -1,23 +1,23 @@
-use std::collections::hash_map;
+use std::collections::{hash_map, HashMap};
 
+use ruff_macros::CacheKey;
 use rustc_hash::FxHashMap;
 
-use super::hashable::HashableHashMap;
 use crate::registry::Rule;
 
 /// A table to keep track of which rules are enabled
 /// and Whether they should be autofixed.
-#[derive(Debug, Hash)]
+#[derive(Debug, CacheKey)]
 pub struct RuleTable {
     /// Maps rule codes to a boolean indicating if the rule should be autofixed.
-    enabled: HashableHashMap<Rule, bool>,
+    enabled: FxHashMap<Rule, bool>,
 }
 
 impl RuleTable {
     /// Creates a new empty rule table.
     pub fn empty() -> Self {
         Self {
-            enabled: HashableHashMap::default(),
+            enabled: HashMap::default(),
         }
     }
 
@@ -53,8 +53,6 @@ impl<I: IntoIterator<Item = Rule>> From<I> for RuleTable {
         for code in codes {
             enabled.insert(code, true);
         }
-        Self {
+        Self { enabled }
-            enabled: enabled.into(),
-        }
     }
 }

crates/ruff/src/settings/types.rs

@@ -1,22 +1,23 @@
-use std::hash::Hash;
+use std::hash::{Hash, Hasher};
 use std::ops::Deref;
 use std::path::{Path, PathBuf};
 use std::str::FromStr;
 
 use anyhow::{anyhow, bail, Result};
 use clap::ValueEnum;
-use globset::{Glob, GlobSetBuilder};
+use globset::{Glob, GlobSet, GlobSetBuilder};
+use ruff_cache::{CacheKey, CacheKeyHasher};
+use ruff_macros::CacheKey;
 use rustc_hash::FxHashSet;
 use schemars::JsonSchema;
 use serde::{de, Deserialize, Deserializer, Serialize};
 
-use super::hashable::HashableHashSet;
 use crate::registry::Rule;
 use crate::rule_selector::RuleSelector;
 use crate::{fs, warn_user_once};
 
 #[derive(
-    Clone, Copy, Debug, PartialOrd, Ord, PartialEq, Eq, Serialize, Deserialize, Hash, JsonSchema,
+    Clone, Copy, Debug, PartialOrd, Ord, PartialEq, Eq, Serialize, Deserialize, JsonSchema, CacheKey,
 )]
 #[serde(rename_all = "lowercase")]
 pub enum PythonVersion {
@@ -61,7 +62,7 @@ impl PythonVersion {
     }
 }
 
-#[derive(Debug, Clone, Hash, PartialEq, PartialOrd, Eq, Ord)]
+#[derive(Debug, Clone, CacheKey, PartialEq, PartialOrd, Eq, Ord)]
 pub enum FilePattern {
     Builtin(&'static str),
     User(String, PathBuf),
@@ -97,11 +98,51 @@ impl FromStr for FilePattern {
     }
 }
 
+#[derive(Debug, Clone, Default)]
+pub struct FilePatternSet {
+    set: GlobSet,
+    cache_key: u64,
+}
+
+impl FilePatternSet {
+    pub fn try_from_vec(patterns: Vec<FilePattern>) -> Result<Self, anyhow::Error> {
+        let mut builder = GlobSetBuilder::new();
+        let mut hasher = CacheKeyHasher::new();
+
+        for pattern in patterns {
+            pattern.cache_key(&mut hasher);
+            pattern.add_to(&mut builder)?;
+        }
+
+        let set = builder.build()?;
+
+        Ok(FilePatternSet {
+            set,
+            cache_key: hasher.finish(),
+        })
+    }
+}
+
+impl Deref for FilePatternSet {
+    type Target = GlobSet;
+
+    fn deref(&self) -> &Self::Target {
+        &self.set
+    }
+}
+
+impl CacheKey for FilePatternSet {
+    fn cache_key(&self, state: &mut CacheKeyHasher) {
+        state.write_usize(self.set.len());
+        state.write_u64(self.cache_key);
+    }
+}
+
 #[derive(Debug, Clone)]
 pub struct PerFileIgnore {
     pub(crate) basename: String,
     pub(crate) absolute: PathBuf,
-    pub(crate) rules: HashableHashSet<Rule>,
+    pub(crate) rules: FxHashSet<Rule>,
 }
 
 impl PerFileIgnore {
@@ -116,7 +157,7 @@ impl PerFileIgnore {
         Self {
             basename: pattern,
             absolute,
-            rules: rules.into(),
+            rules,
         }
     }
 }

crates/ruff_cache/Cargo.toml

@@ -0,0 +1,15 @@
+[package]
+name = "ruff_cache"
+version = "0.0.0"
+publish = false
+edition = { workspace = true }
+rust-version = { workspace = true }
+
+[dependencies]
+itertools = { workspace = true }
+globset = { version = "0.4.9" }
+regex = { workspace = true }
+filetime = { version = "0.2.17" }
+
+[dev-dependencies]
+ruff_macros = { path = "../ruff_macros" }

crates/ruff_cache/src/cache_key.rs

@@ -0,0 +1,376 @@
+use itertools::Itertools;
+use regex::Regex;
+use std::borrow::Cow;
+use std::collections::hash_map::DefaultHasher;
+use std::collections::{BTreeMap, BTreeSet, HashMap, HashSet};
+use std::hash::{Hash, Hasher};
+use std::ops::{Deref, DerefMut};
+use std::path::{Path, PathBuf};
+
+#[derive(Clone, Debug, Default)]
+pub struct CacheKeyHasher {
+    inner: DefaultHasher,
+}
+
+impl CacheKeyHasher {
+    pub fn new() -> Self {
+        Self {
+            inner: DefaultHasher::new(),
+        }
+    }
+}
+
+impl Deref for CacheKeyHasher {
+    type Target = DefaultHasher;
+    fn deref(&self) -> &Self::Target {
+        &self.inner
+    }
+}
+
+impl DerefMut for CacheKeyHasher {
+    fn deref_mut(&mut self) -> &mut Self::Target {
+        &mut self.inner
+    }
+}
+
+/// A type that be used as part of a cache key.
+///
+/// A cache looks up artefacts by a cache key. Many cache keys are composed of sub-keys. For example,
+/// caching the lint results of a file depend at least on the file content, the user settings, and linter version.
+/// Types implementing the [`CacheKey`] trait can be used as part of a cache key by which artefacts are queried.
+///
+/// ## Implementing `CacheKey`
+///
+/// You can derive [`CacheKey`] with `#[derive(CacheKey)]` if all fields implement [`CacheKey`]. The resulting
+/// cache key will be the combination of the values from calling `cache_key` on each field.
+///
+/// ```
+/// # use ruff_macros::CacheKey;
+///
+/// #[derive(CacheKey)]
+/// struct Test {
+///     name: String,
+///     version: u32,
+/// }
+/// ```
+///
+/// If you need more control over computing the cache key, you can of course implement the [`CacheKey]` yourself:
+///
+/// ```
+///  use ruff_cache::{CacheKey, CacheKeyHasher};
+///
+/// struct Test {
+///     name: String,
+///     version: u32,
+///     other: String
+/// }
+///
+/// impl CacheKey for Test {
+///     fn cache_key(&self, state: &mut CacheKeyHasher) {
+///         self.name.cache_key(state);
+///         self.version.cache_key(state);
+///     }
+/// }
+/// ```
+///
+/// ## Portability
+///
+/// Ideally, the cache key is portable across platforms but this is not yet a strict requirement.
+///
+/// ## Using [`Hash`]
+///
+/// You can defer to the [`Hash`] implementation for non-composite types.
+/// Be aware, that the [`Hash`] implementation may not be portable.
+///
+/// ## Why a new trait rather than reusing [`Hash`]?
+/// The main reason is that hashes and cache keys have different constraints:
+///
+/// * Cache keys are less performance sensitive: Hashes must be super fast to compute for performant hashed-collections. That's
+///   why some standard types don't implement [`Hash`] where it would be safe to to implement [`CacheKey`], e.g. `HashSet`
+/// * Cache keys must be deterministic where hash keys do not have this constraint. That's why pointers don't implement [`CacheKey`] but they implement [`Hash`].
+/// * Ideally, cache keys are portable
+///
+/// [`Hash`](std::hash::Hash)
+pub trait CacheKey {
+    fn cache_key(&self, state: &mut CacheKeyHasher);
+
+    fn cache_key_slice(data: &[Self], state: &mut CacheKeyHasher)
+    where
+        Self: Sized,
+    {
+        for piece in data {
+            piece.cache_key(state);
+        }
+    }
+}
+
+impl CacheKey for bool {
+    #[inline]
+    fn cache_key(&self, state: &mut CacheKeyHasher) {
+        state.write_u8(u8::from(*self));
+    }
+}
+
+impl CacheKey for char {
+    #[inline]
+    fn cache_key(&self, state: &mut CacheKeyHasher) {
+        state.write_u32(*self as u32);
+    }
+}
+
+impl CacheKey for usize {
+    #[inline]
+    fn cache_key(&self, state: &mut CacheKeyHasher) {
+        state.write_usize(*self);
+    }
+}
+
+impl CacheKey for u128 {
+    #[inline]
+    fn cache_key(&self, state: &mut CacheKeyHasher) {
+        state.write_u128(*self);
+    }
+}
+
+impl CacheKey for u64 {
+    #[inline]
+    fn cache_key(&self, state: &mut CacheKeyHasher) {
+        state.write_u64(*self);
+    }
+}
+
+impl CacheKey for u32 {
+    #[inline]
+    fn cache_key(&self, state: &mut CacheKeyHasher) {
+        state.write_u32(*self);
+    }
+}
+
+impl CacheKey for u16 {
+    #[inline]
+    fn cache_key(&self, state: &mut CacheKeyHasher) {
+        state.write_u16(*self);
+    }
+}
+
+impl CacheKey for u8 {
+    #[inline]
+    fn cache_key(&self, state: &mut CacheKeyHasher) {
+        state.write_u8(*self);
+    }
+}
+
+impl CacheKey for isize {
+    #[inline]
+    fn cache_key(&self, state: &mut CacheKeyHasher) {
+        state.write_isize(*self);
+    }
+}
+
+impl CacheKey for i128 {
+    #[inline]
+    fn cache_key(&self, state: &mut CacheKeyHasher) {
+        state.write_i128(*self);
+    }
+}
+
+impl CacheKey for i64 {
+    #[inline]
+    fn cache_key(&self, state: &mut CacheKeyHasher) {
+        state.write_i64(*self);
+    }
+}
+
+impl CacheKey for i32 {
+    #[inline]
+    fn cache_key(&self, state: &mut CacheKeyHasher) {
+        state.write_i32(*self);
+    }
+}
+
+impl CacheKey for i16 {
+    #[inline]
+    fn cache_key(&self, state: &mut CacheKeyHasher) {
+        state.write_i16(*self);
+    }
+}
+
+impl CacheKey for i8 {
+    #[inline]
+    fn cache_key(&self, state: &mut CacheKeyHasher) {
+        state.write_i8(*self);
+    }
+}
+
+macro_rules! impl_cache_key_tuple {
+    () => (
+        impl CacheKey for () {
+            #[inline]
+            fn cache_key(&self, _state: &mut CacheKeyHasher) {}
+        }
+    );
+
+    ( $($name:ident)+) => (
+        impl<$($name: CacheKey),+> CacheKey for ($($name,)+) where last_type!($($name,)+): ?Sized {
+            #[allow(non_snake_case)]
+            #[inline]
+            fn cache_key(&self, state: &mut CacheKeyHasher) {
+                let ($(ref $name,)+) = *self;
+                $($name.cache_key(state);)+
+            }
+        }
+    );
+}
+
+macro_rules! last_type {
+    ($a:ident,) => { $a };
+    ($a:ident, $($rest_a:ident,)+) => { last_type!($($rest_a,)+) };
+}
+
+impl_cache_key_tuple! {}
+impl_cache_key_tuple! { T }
+impl_cache_key_tuple! { T B }
+impl_cache_key_tuple! { T B C }
+impl_cache_key_tuple! { T B C D }
+impl_cache_key_tuple! { T B C D E }
+impl_cache_key_tuple! { T B C D E F }
+impl_cache_key_tuple! { T B C D E F G }
+impl_cache_key_tuple! { T B C D E F G H }
+impl_cache_key_tuple! { T B C D E F G H I }
+impl_cache_key_tuple! { T B C D E F G H I J }
+impl_cache_key_tuple! { T B C D E F G H I J K }
+impl_cache_key_tuple! { T B C D E F G H I J K L }
+
+impl CacheKey for str {
+    #[inline]
+    fn cache_key(&self, state: &mut CacheKeyHasher) {
+        self.hash(&mut **state);
+    }
+}
+
+impl CacheKey for String {
+    #[inline]
+    fn cache_key(&self, state: &mut CacheKeyHasher) {
+        self.hash(&mut **state);
+    }
+}
+
+impl<T: CacheKey> CacheKey for Option<T> {
+    #[inline]
+    fn cache_key(&self, state: &mut CacheKeyHasher) {
+        match self {
+            None => state.write_usize(0),
+            Some(value) => {
+                state.write_usize(1);
+                value.cache_key(state);
+            }
+        }
+    }
+}
+
+impl<T: CacheKey> CacheKey for [T] {
+    #[inline]
+    fn cache_key(&self, state: &mut CacheKeyHasher) {
+        state.write_usize(self.len());
+        CacheKey::cache_key_slice(self, state);
+    }
+}
+
+impl<T: ?Sized + CacheKey> CacheKey for &T {
+    #[inline]
+    fn cache_key(&self, state: &mut CacheKeyHasher) {
+        (**self).cache_key(state);
+    }
+}
+
+impl<T: ?Sized + CacheKey> CacheKey for &mut T {
+    #[inline]
+    fn cache_key(&self, state: &mut CacheKeyHasher) {
+        (**self).cache_key(state);
+    }
+}
+
+impl<T> CacheKey for Vec<T>
+where
+    T: CacheKey,
+{
+    fn cache_key(&self, state: &mut CacheKeyHasher) {
+        state.write_usize(self.len());
+        CacheKey::cache_key_slice(self, state);
+    }
+}
+
+impl<K, V, S> CacheKey for HashMap<K, V, S>
+where
+    K: CacheKey + Ord,
+    V: CacheKey,
+{
+    fn cache_key(&self, state: &mut CacheKeyHasher) {
+        state.write_usize(self.len());
+        for (key, value) in self
+            .iter()
+            .sorted_by(|(left, _), (right, _)| left.cmp(right))
+        {
+            key.cache_key(state);
+            value.cache_key(state);
+        }
+    }
+}
+
+impl<V: CacheKey + Ord, S> CacheKey for HashSet<V, S> {
+    fn cache_key(&self, state: &mut CacheKeyHasher) {
+        state.write_usize(self.len());
+        for value in self.iter().sorted() {
+            value.cache_key(state);
+        }
+    }
+}
+
+impl<V: CacheKey> CacheKey for BTreeSet<V> {
+    fn cache_key(&self, state: &mut CacheKeyHasher) {
+        state.write_usize(self.len());
+        for item in self {
+            item.cache_key(state);
+        }
+    }
+}
+
+impl<K: CacheKey + Ord, V: CacheKey> CacheKey for BTreeMap<K, V> {
+    fn cache_key(&self, state: &mut CacheKeyHasher) {
+        state.write_usize(self.len());
+
+        for (key, value) in self {
+            key.cache_key(state);
+            value.cache_key(state);
+        }
+    }
+}
+
+impl CacheKey for Path {
+    #[inline]
+    fn cache_key(&self, state: &mut CacheKeyHasher) {
+        self.hash(&mut **state);
+    }
+}
+
+impl CacheKey for PathBuf {
+    #[inline]
+    fn cache_key(&self, state: &mut CacheKeyHasher) {
+        self.hash(&mut **state);
+    }
+}
+
+impl<V: ?Sized> CacheKey for Cow<'_, V>
+where
+    V: CacheKey + ToOwned,
+{
+    fn cache_key(&self, state: &mut CacheKeyHasher) {
+        (**self).cache_key(state);
+    }
+}
+
+impl CacheKey for Regex {
+    fn cache_key(&self, state: &mut CacheKeyHasher) {
+        self.as_str().cache_key(state);
+    }
+}

crates/ruff_cache/src/filetime.rs

@@ -0,0 +1,9 @@
+use crate::{CacheKey, CacheKeyHasher};
+use filetime::FileTime;
+use std::hash::Hash;
+
+impl CacheKey for FileTime {
+    fn cache_key(&self, state: &mut CacheKeyHasher) {
+        self.hash(&mut **state);
+    }
+}

crates/ruff_cache/src/globset.rs

@@ -0,0 +1,14 @@
+use crate::{CacheKey, CacheKeyHasher};
+use globset::{Glob, GlobMatcher};
+
+impl CacheKey for GlobMatcher {
+    fn cache_key(&self, state: &mut CacheKeyHasher) {
+        self.glob().cache_key(state);
+    }
+}
+
+impl CacheKey for Glob {
+    fn cache_key(&self, state: &mut CacheKeyHasher) {
+        self.glob().cache_key(state);
+    }
+}

crates/ruff_cache/src/lib.rs

@@ -1,3 +1,9 @@
+mod cache_key;
+pub mod filetime;
+pub mod globset;
+
+pub use cache_key::{CacheKey, CacheKeyHasher};
+
 use std::path::{Path, PathBuf};
 
 pub const CACHE_DIR_NAME: &str = ".ruff_cache";

crates/ruff_cache/tests/cache_key.rs

@@ -0,0 +1,108 @@
+use ruff_cache::{CacheKey, CacheKeyHasher};
+use ruff_macros::CacheKey;
+use std::collections::hash_map::DefaultHasher;
+use std::hash::{Hash, Hasher};
+
+#[derive(CacheKey, Hash)]
+struct UnitStruct;
+
+#[derive(CacheKey, Hash)]
+struct NamedFieldsStruct {
+    a: String,
+    b: String,
+}
+
+#[derive(CacheKey, Hash)]
+struct UnnamedFieldsStruct(String, String);
+
+#[derive(CacheKey, Hash)]
+enum Enum {
+    Unit,
+    UnnamedFields(String, String),
+    NamedFields { a: String, b: String },
+}
+
+#[test]
+fn unit_struct_cache_key() {
+    let mut key = CacheKeyHasher::new();
+
+    UnitStruct.cache_key(&mut key);
+
+    let mut hash = DefaultHasher::new();
+    UnitStruct.hash(&mut hash);
+
+    assert_eq!(hash.finish(), key.finish());
+}
+
+#[test]
+fn named_field_struct() {
+    let mut key = CacheKeyHasher::new();
+
+    let named_fields = NamedFieldsStruct {
+        a: "Hello".into(),
+        b: "World".into(),
+    };
+
+    named_fields.cache_key(&mut key);
+
+    let mut hash = DefaultHasher::new();
+    named_fields.hash(&mut hash);
+
+    assert_eq!(hash.finish(), key.finish());
+}
+
+#[test]
+fn unnamed_field_struct() {
+    let mut key = CacheKeyHasher::new();
+
+    let unnamed_fields = UnnamedFieldsStruct("Hello".into(), "World".into());
+
+    unnamed_fields.cache_key(&mut key);
+
+    let mut hash = DefaultHasher::new();
+    unnamed_fields.hash(&mut hash);
+
+    assert_eq!(hash.finish(), key.finish());
+}
+
+#[test]
+fn enum_unit_variant() {
+    let mut key = CacheKeyHasher::new();
+
+    let variant = Enum::Unit;
+    variant.cache_key(&mut key);
+
+    let mut hash = DefaultHasher::new();
+    variant.hash(&mut hash);
+
+    assert_eq!(hash.finish(), key.finish());
+}
+
+#[test]
+fn enum_named_fields_variant() {
+    let mut key = CacheKeyHasher::new();
+
+    let variant = Enum::NamedFields {
+        a: "Hello".to_string(),
+        b: "World".to_string(),
+    };
+    variant.cache_key(&mut key);
+
+    let mut hash = DefaultHasher::new();
+    variant.hash(&mut hash);
+
+    assert_eq!(hash.finish(), key.finish());
+}
+
+#[test]
+fn enum_unnamed_fields_variant() {
+    let mut key = CacheKeyHasher::new();
+
+    let variant = Enum::UnnamedFields("Hello".to_string(), "World".to_string());
+    variant.cache_key(&mut key);
+
+    let mut hash = DefaultHasher::new();
+    variant.hash(&mut hash);
+
+    assert_eq!(hash.finish(), key.finish());
+}

crates/ruff_cli/Cargo.toml

@@ -25,6 +25,7 @@ doc = false
 
 [dependencies]
 ruff = { path = "../ruff" }
+ruff_cache = { path = "../ruff_cache" }
 
 annotate-snippets = { version = "0.9.1", features = ["color"] }
 anyhow = { workspace = true }

crates/ruff_cli/src/cache.rs

@@ -1,6 +1,5 @@
-use std::collections::hash_map::DefaultHasher;
 use std::fs;
-use std::hash::{Hash, Hasher};
+use std::hash::Hasher;
 use std::io::Write;
 use std::path::Path;
 
@@ -10,6 +9,7 @@ use log::error;
 use path_absolutize::Absolutize;
 use ruff::message::Message;
 use ruff::settings::{flags, AllSettings, Settings};
+use ruff_cache::{CacheKey, CacheKeyHasher};
 use serde::{Deserialize, Serialize};
 #[cfg(unix)]
 use std::os::unix::fs::PermissionsExt;
@@ -37,18 +37,18 @@ fn cache_key<P: AsRef<Path>>(
     settings: &Settings,
     autofix: flags::Autofix,
 ) -> u64 {
-    let mut hasher = DefaultHasher::new();
+    let mut hasher = CacheKeyHasher::new();
-    CARGO_PKG_VERSION.hash(&mut hasher);
+    CARGO_PKG_VERSION.cache_key(&mut hasher);
-    path.as_ref().absolutize().unwrap().hash(&mut hasher);
+    path.as_ref().absolutize().unwrap().cache_key(&mut hasher);
     package
         .as_ref()
         .map(|path| path.as_ref().absolutize().unwrap())
-        .hash(&mut hasher);
+        .cache_key(&mut hasher);
-    FileTime::from_last_modification_time(metadata).hash(&mut hasher);
+    FileTime::from_last_modification_time(metadata).cache_key(&mut hasher);
     #[cfg(unix)]
-    metadata.permissions().mode().hash(&mut hasher);
+    metadata.permissions().mode().cache_key(&mut hasher);
-    settings.hash(&mut hasher);
+    settings.cache_key(&mut hasher);
-    autofix.hash(&mut hasher);
+    autofix.cache_key(&mut hasher);
     hasher.finish()
 }
 

crates/ruff_cli/src/commands/clean.rs

@@ -6,9 +6,9 @@ use colored::Colorize;
 use path_absolutize::path_dedot;
 use walkdir::WalkDir;
 
-use ruff::cache::CACHE_DIR_NAME;
 use ruff::fs;
 use ruff::logging::LogLevel;
+use ruff_cache::CACHE_DIR_NAME;
 
 /// Clear any caches in the current directory or any subdirectories.
 pub fn clean(level: LogLevel) -> Result<()> {

crates/ruff_macros/Cargo.toml

@@ -15,3 +15,4 @@ quote = { version = "1.0.21" }
 syn = { version = "1.0.103", features = ["derive", "parsing", "extra-traits"] }
 textwrap = { version = "0.16.0" }
 itertools = { workspace = true }
+

crates/ruff_macros/src/cache_key.rs

@@ -0,0 +1,103 @@
+use proc_macro2::TokenStream;
+use quote::{format_ident, quote};
+use syn::spanned::Spanned;
+use syn::{Data, DeriveInput, Error, Fields};
+
+pub fn derive_cache_key(item: &DeriveInput) -> syn::Result<TokenStream> {
+    let fields = match &item.data {
+        Data::Enum(item_enum) => {
+            let arms = item_enum.variants.iter().enumerate().map(|(i, variant)| {
+                let variant_name = &variant.ident;
+
+                match &variant.fields {
+                    Fields::Named(fields) => {
+                        let field_names: Vec<_> = fields
+                            .named
+                            .iter()
+                            .map(|field| field.ident.clone().unwrap())
+                            .collect();
+
+                        let fields_code = field_names
+                            .iter()
+                            .map(|field| quote!(#field.cache_key(key);));
+
+                        quote! {
+                            Self::#variant_name{#(#field_names),*} => {
+                                key.write_usize(#i);
+                                #(#fields_code)*
+                            }
+                        }
+                    }
+                    Fields::Unnamed(fields) => {
+                        let field_names: Vec<_> = fields
+                            .unnamed
+                            .iter()
+                            .enumerate()
+                            .map(|(i, _)| format_ident!("field_{i}"))
+                            .collect();
+
+                        let fields_code = field_names
+                            .iter()
+                            .map(|field| quote!(#field.cache_key(key);));
+
+                        quote! {
+                            Self::#variant_name(#(#field_names),*) => {
+                                key.write_usize(#i);
+                                #(#fields_code)*
+                            }
+                        }
+                    }
+                    Fields::Unit => {
+                        quote! {
+                            Self::#variant_name => {
+                                key.write_usize(#i);
+                            }
+                        }
+                    }
+                }
+            });
+
+            quote! {
+                match self {
+                    #(#arms)*
+                }
+            }
+        }
+
+        Data::Struct(item_struct) => {
+            let fields = item_struct.fields.iter().enumerate().map(|(i, field)| {
+                let field_attr = match &field.ident {
+                    Some(ident) => quote!(self.#ident),
+                    None => {
+                        let index = syn::Index::from(i);
+                        quote!(self.#index)
+                    }
+                };
+
+                quote!(#field_attr.cache_key(key);)
+            });
+
+            quote! {#(#fields)*}
+        }
+
+        Data::Union(_) => {
+            return Err(Error::new(
+                item.span(),
+                "CacheKey does not support unions. Only structs and enums are supported",
+            ))
+        }
+    };
+
+    let name = &item.ident;
+    let (impl_generics, ty_generics, where_clause) = &item.generics.split_for_impl();
+
+    Ok(quote!(
+        impl #impl_generics ruff_cache::CacheKey for #name #ty_generics #where_clause {
+            fn cache_key(&self, key: &mut ruff_cache::CacheKeyHasher) {
+                use std::hash::Hasher;
+                use ruff_cache::CacheKey;
+                #fields
+            }
+        }
+    ))
+}

crates/ruff_macros/src/lib.rs

@@ -1,8 +1,10 @@
 //! This crate implements internal macros for the `ruff` library.
 
+use crate::cache_key::derive_cache_key;
 use proc_macro::TokenStream;
 use syn::{parse_macro_input, DeriveInput, ItemFn};
 
+mod cache_key;
 mod config;
 mod define_violation;
 mod derive_message_formats;
@@ -20,6 +22,16 @@ pub fn derive_config(input: proc_macro::TokenStream) -> proc_macro::TokenStream
         .into()
 }
 
+#[proc_macro_derive(CacheKey)]
+pub fn cache_key(input: TokenStream) -> TokenStream {
+    let item = parse_macro_input!(input as DeriveInput);
+
+    let result = derive_cache_key(&item);
+    let stream = result.unwrap_or_else(|err| err.to_compile_error());
+
+    TokenStream::from(stream)
+}
+
 #[proc_macro]
 pub fn register_rules(item: proc_macro::TokenStream) -> proc_macro::TokenStream {
     let mapping = parse_macro_input!(item as register_rules::Input);

crates/ruff_macros/src/register_rules.rs

@@ -56,6 +56,7 @@ pub fn register_rules(input: &Input) -> proc_macro2::TokenStream {
             Hash,
             PartialOrd,
             Ord,
+            ::ruff_macros::CacheKey,
             AsRefStr,
             ::strum_macros::IntoStaticStr,
         )]