Rust源码剖析：Lint - CombinedLintPass
摘要：前一篇文章 介绍了关于 Lint 和 LintPass 的一些概念和实现。基于这些结构，提供了一个简易的 Lint 检查的实现方式。本文主要介绍 CombinedLintPass 这一结构的实现，并基于 CombinedLintPass 进一步优化 Lint 的实现。
背景
在 KusionStack 技术栈中， KCL 配置策略语言是重要的组成部分之一。为了帮助用户更好的编写 KCL 代码，我们也为 KCL 语言开发了一些语言工具，Lint 就是其中一种。Lint 工具帮助用户检查代码中潜在的问题和错误，同时也可以用于自动化的代码检查，保障仓库代码规范和质量。因为 KCL 语言由 Rust 实现，一些功能也学习和参考了 Rustc。本文是在学习 Rustc 过程中的一些思考和沉淀，在这里做一些分享。
前一篇文章(Rustc 源码学习 - Lint 与 LintPass) 介绍了关于 Lint 和 LintPass 的一些概念和实现。基于这些结构，提供了一个简易的 Lint 检查的实现方式。本文主要介绍 CombinedLintPass 这一结构的实现，并基于 CombinedLintPass 进一步优化 Lint 的实现。
CombinedLintpass
Rustc 在 LintPass 的中实现了 Lint 工具检查的具体逻辑。并且使用 Visitor 模式遍历 AST 的同时调用 lintpass 中的 check_*方法。
1impl ast_visit::Visitor for Linter {
2 fn visit_crate(a: ast:crate){
3 for lintpass in lintpasses{
4 lintpass.check_crate(a)
5 }
6 walk_crate();
7 }
8 fn visit_stmt(a: ast:stmt){
9 for lintpass in lintpasses{
10 lintpass.check_stmt(a)
11 }
12 walk_stmt();
13 }
14 ...
15}但是，Rustc 自身和 clippy 提供的 Lint 定义多达550+多个。考虑到性能因素，定义大量的 LintPass，分别注册和调用显然是不合适的。Rustc 提供了一种更优的解决方法：既然可以将多个 Lint 组织为一个 LintPass，同样也可以将多个 LintPass 组合成一个 CombinedLintPass。
Compiler lint passes are combined into one pass
Within the compiler, for performance reasons, we usually do not register dozens of lint passes. Instead, we have a single lint pass of each variety (e.g., BuiltinCombinedModuleLateLintPass) which will internally call all of the individual lint passes; this is because then we get the benefits of static over dynamic dispatch for each of the (often empty) trait methods.
Ideally, we'd not have to do this, since it adds to the complexity of understanding the code. However, with the current type-erased lint store approach, it is beneficial to do so for performance reasons.
BuiltinCombinedEarlyLintPass
CombinedLintPass 同样分为 early 和 late 两类。以 builtin 的 early lint 为例，Rustc 在 rustc_lint::src::lib.rs 中为这些 lintpass 定义了一个 BuiltinCombinedEarlyLintPass 结构。
1early_lint_passes!(declare_combined_early_pass, [BuiltinCombinedEarlyLintPass]);虽然这个定义看起来只有一行，但其中通过若干个宏的展开，汇总了14个 LintPass，并且每个 LintPass 提供了50多个 check_* 方法。接下来一一说明这些宏。
BuiltinCombinedEarlyLintPass 的宏定义
early_lint_passes
1macro_rules! early_lint_passes {
2 ($macro:path, $args:tt) => {
3 $macro!(
4 $args,
5 [
6 UnusedParens: UnusedParens,
7 UnusedBraces: UnusedBraces,
8 UnusedImportBraces: UnusedImportBraces,
9 UnsafeCode: UnsafeCode,
10 AnonymousParameters: AnonymousParameters,
11 EllipsisInclusiveRangePatterns: EllipsisInclusiveRangePatterns::default(),
12 NonCamelCaseTypes: NonCamelCaseTypes,
13 DeprecatedAttr: DeprecatedAttr::new(),
14 WhileTrue: WhileTrue,
15 NonAsciiIdents: NonAsciiIdents,
16 HiddenUnicodeCodepoints: HiddenUnicodeCodepoints,
17 IncompleteFeatures: IncompleteFeatures,
18 RedundantSemicolons: RedundantSemicolons,
19 UnusedDocComment: UnusedDocComment,
20 ]
21 );
22 };
23}首先是 early_lint_passes 宏，这个宏的主要作用是定义了所有的 early lintpass。这里的 lintpass 是成对出现的，:左边为 lintpass 的 Identifier，:右边为 lintpass 的constructor。所以会出现 EllipsisInclusiveRangePatterns::default() 和 DeprecatedAttr::new()这种形式。early_lint_passes 会将定义的 early lintpass 和 第二个参数一起传递给下一个宏。
通过这个宏，之前的BuiltinCombinedEarlyLintPass的定义被展开为：
1declare_combined_early_pass!([BuiltinCombinedEarlyLintPass], [
2 UnusedParens: UnusedParens,
3 UnusedBraces: UnusedBraces,
4 UnusedImportBraces: UnusedImportBraces,
5 UnsafeCode: UnsafeCode,
6 AnonymousParameters: AnonymousParameters,
7 EllipsisInclusiveRangePatterns: EllipsisInclusiveRangePatterns::default(),
8 NonCamelCaseTypes: NonCamelCaseTypes,
9 DeprecatedAttr: DeprecatedAttr::new(),
10 WhileTrue: WhileTrue,
11 NonAsciiIdents: NonAsciiIdents,
12 HiddenUnicodeCodepoints: HiddenUnicodeCodepoints,
13 IncompleteFeatures: IncompleteFeatures,
14 RedundantSemicolons: RedundantSemicolons,
15 UnusedDocComment: UnusedDocComment,
16 ])declare_combined_early_pass
1macro_rules! declare_combined_early_pass {
2 ([$name:ident], $passes:tt) => (
3 early_lint_methods!(declare_combined_early_lint_pass, [pub $name, $passes]);
4 )
5}declare_combined_early_pass 宏接收 early_lint_passes宏传来的 name(BuiltinCombinedEarlyLintPass) 和 passes，并继续传递给 early_lint_methods 宏。
通过这个宏，BuiltinCombinedEarlyLintPass的定义继续展开为：
1early_lint_methods!(declare_combined_early_lint_pass,
2 [pub BuiltinCombinedEarlyLintPass,
3 [
4 UnusedParens: UnusedParens,
5 UnusedBraces: UnusedBraces,
6 UnusedImportBraces: UnusedImportBraces,
7 UnsafeCode: UnsafeCode,
8 AnonymousParameters: AnonymousParameters,
9 EllipsisInclusiveRangePatterns: EllipsisInclusiveRangePatterns::default(),
10 NonCamelCaseTypes: NonCamelCaseTypes,
11 DeprecatedAttr: DeprecatedAttr::new(),
12 WhileTrue: WhileTrue,
13 NonAsciiIdents: NonAsciiIdents,
14 HiddenUnicodeCodepoints: HiddenUnicodeCodepoints,
15 IncompleteFeatures: IncompleteFeatures,
16 RedundantSemicolons: RedundantSemicolons,
17 UnusedDocComment: UnusedDocComment,
18 ]
19 ]);early_lint_methods
1macro_rules! early_lint_methods {
2 ($macro:path, $args:tt) => (
3 $macro!($args, [
4 fn check_param(a: &ast::Param);
5 fn check_ident(a: &ast::Ident);
6 fn check_crate(a: &ast::Crate);
7 fn check_crate_post(a: &ast::Crate);
8 ...
9 ]);
10 )
11}early_lint_methods 宏在前一篇文章中也介绍过，它定义了 EarlyLintPass 中需要实现的 check_*函数，并且将这些函数以及接收的参数 $args传递给下一个宏。因为 BuiltinCombinedEarlyLintPass 也是 early lint 的一种，所以同样需要实现这些函数。
通过这个宏，BuiltinCombinedEarlyLintPass的定义继续展开为：
1declare_combined_early_lint_pass!(
2 [pub BuiltinCombinedEarlyLintPass,
3 [
4 UnusedParens: UnusedParens,
5 UnusedBraces: UnusedBraces,
6 UnusedImportBraces: UnusedImportBraces,
7 UnsafeCode: UnsafeCode,
8 AnonymousParameters: AnonymousParameters,
9 EllipsisInclusiveRangePatterns: EllipsisInclusiveRangePatterns::default(),
10 NonCamelCaseTypes: NonCamelCaseTypes,
11 DeprecatedAttr: DeprecatedAttr::new(),
12 WhileTrue: WhileTrue,
13 NonAsciiIdents: NonAsciiIdents,
14 HiddenUnicodeCodepoints: HiddenUnicodeCodepoints,
15 IncompleteFeatures: IncompleteFeatures,
16 RedundantSemicolons: RedundantSemicolons,
17 UnusedDocComment: UnusedDocComment,
18 ]
19 ],
20 [
21 fn check_param(a: &ast::Param);
22 fn check_ident(a: &ast::Ident);
23 fn check_crate(a: &ast::Crate);
24 fn check_crate_post(a: &ast::Crate);
25 ...
26 ]
27)declare_combined_early_lint_pass
1macro_rules! declare_combined_early_lint_pass {
2 ([$v:vis $name:ident, [$($passes:ident: $constructor:expr,)*]], $methods:tt) => (
3 #[allow(non_snake_case)]
4 $v struct $name {
5 $($passes: $passes,)*
6 }
7 impl $name {
8 $v fn new() -> Self {
9 Self {
10 $($passes: $constructor,)*
11 }
12 }
13 $v fn get_lints() -> LintArray {
14 let mut lints = Vec::new();
15 $(lints.extend_from_slice(&$passes::get_lints());)*
16 lints
17 }
18 }
19 impl EarlyLintPass for $name {
20 expand_combined_early_lint_pass_methods!([$($passes),*], $methods);
21 }
22 #[allow(rustc::lint_pass_impl_without_macro)]
23 impl LintPass for $name {
24 fn name(&self) -> &'static str {
25 panic!()
26 }
27 }
28 )
29}declare_combined_early_lint_pass宏是生成 BuiltinCombinedEarlyLintPass 的主体。这个宏中做了以下工作：
生成一个名为
BuiltinCombinedEarlyLintPass的 struct，其中的属性为宏early_lint_passes提供的 lintpass 的 identifier。实现
fn new()fn name()和fn get_lints()方法。其中new()调用了early_lint_passes提供的 lintpass 的 constructor。调用宏
expand_combined_early_lint_pass_methods，实现自身的check_*方法。
通过这个宏，BuiltinCombinedEarlyLintPass的定义变为：
1pub struct BuiltinCombinedEarlyLintPass {
2 UnusedParens: UnusedParens,
3 UnusedBraces: UnusedBraces,
4 UnusedImportBraces: UnusedImportBraces,
5 UnsafeCode: UnsafeCode,
6 AnonymousParameters: AnonymousParameters,
7 EllipsisInclusiveRangePatterns: EllipsisInclusiveRangePatterns,
8 NonCamelCaseTypes: NonCamelCaseTypes,
9 DeprecatedAttr: DeprecatedAttr,
10 WhileTrue: WhileTrue,
11 NonAsciiIdents: NonAsciiIdents,
12 HiddenUnicodeCodepoints: HiddenUnicodeCodepoints,
13 IncompleteFeatures: IncompleteFeatures,
14 RedundantSemicolons: RedundantSemicolons,
15 UnusedDocComment: UnusedDocComment,
16}
17impl BuiltinCombinedEarlyLintPass {
18 pub fn new() -> Self {
19 Self {
20 UnusedParens: UnusedParens,
21 UnusedBraces: UnusedBraces,
22 UnusedImportBraces: UnusedImportBraces,
23 UnsafeCode: UnsafeCode,
24 AnonymousParameters: AnonymousParameters,
25 EllipsisInclusiveRangePatterns: EllipsisInclusiveRangePatterns::default(),
26 NonCamelCaseTypes: NonCamelCaseTypes,
27 DeprecatedAttr: DeprecatedAttr::new(),
28 WhileTrue: WhileTrue,
29 NonAsciiIdents: NonAsciiIdents,
30 HiddenUnicodeCodepoints: HiddenUnicodeCodepoints,
31 IncompleteFeatures: IncompleteFeatures,
32 RedundantSemicolons: RedundantSemicolons,
33 UnusedDocComment: UnusedDocComment,
34 }
35 }
36 pub fn get_lints() -> LintArray {
37 let mut lints = Vec::new();
38 lints.extend_from_slice(&UnusedParens::get_lints());
39 lints.extend_from_slice(&UnusedBraces::get_lints());
40 lints.extend_from_slice(&UnusedImportBraces::get_lints());
41 lints.extend_from_slice(&UnsafeCode::get_lints());
42 lints.extend_from_slice(&AnonymousParameters::get_lints());
43 lints.extend_from_slice(&EllipsisInclusiveRangePatterns::get_lints());
44 lints.extend_from_slice(&NonCamelCaseTypes::get_lints());
45 lints.extend_from_slice(&DeprecatedAttr::get_lints());
46 lints.extend_from_slice(&WhileTrue::get_lints());
47 lints.extend_from_slice(&NonAsciiIdents::get_lints());
48 lints.extend_from_slice(&HiddenUnicodeCodepoints::get_lints());
49 lints.extend_from_slice(&IncompleteFeatures::get_lints());
50 lints.extend_from_slice(&RedundantSemicolons::get_lints());
51 lints.extend_from_slice(&UnusedDocComment::get_lints());
52
53 lints
54 }
55}
56impl EarlyLintPass for BuiltinCombinedEarlyLintPass {
57 expand_combined_early_lint_pass_methods!([$($passes),*], $methods);
58}
59#[allow(rustc::lint_pass_impl_without_macro)]
60impl LintPass for BuiltinCombinedEarlyLintPass {
61 fn name(&self) -> &'static str {
62 panic!()
63 }
64}expand_combined_early_lint_pass_methods
1macro_rules! expand_combined_early_lint_pass_methods {
2 ($passes:tt, [$($(#[$attr:meta])* fn $name:ident($($param:ident: $arg:ty),*);)*]) => (
3 $(fn $name(&mut self, context: &EarlyContext<'_>, $($param: $arg),*) {
4 expand_combined_early_lint_pass_method!($passes, self, $name, (context, $($param),*));
5 })*
6 )
7}expand_combined_early_lint_pass_methods宏在 BuiltinCombinedEarlyLintPass 中展开所有 early_lint_methods 中定义的方法。
通过这个宏，BuiltinCombinedEarlyLintPass的定义变为（省略其他定义）：
1impl EarlyLintPass for BuiltinCombinedEarlyLintPass {
2 fn check_param(&mut self, context: &EarlyContext<'_>, a: &ast::Param) {
3 expand_combined_early_lint_pass_method!($passes, self, $name, (context, $($param),*));
4 }
5 fn check_ident(&mut self, context: &EarlyContext<'_>, a: &ast::Ident) {
6 expand_combined_early_lint_pass_method!($passes, self, $name, (context, $($param),*));
7 }
8 fn check_crate(&mut self, context: &EarlyContext<'_>, a: &ast::Crate) {
9 expand_combined_early_lint_pass_method!($passes, self, $name, (context, $($param),*));
10 }
11 ...
12
13}expand_combined_early_lint_pass_method
1macro_rules! expand_combined_early_lint_pass_method {
2 ([$($passes:ident),*], $self: ident, $name: ident, $params:tt) => ({
3 $($self.$passes.$name $params;)*
4 })
5}expand_combined_early_lint_pass_method：在展开的check_* 函数中调用每一个 LintPass 的 check_*。
通过这个宏，BuiltinCombinedEarlyLintPass的定义变为（省略其他定义）：
1impl EarlyLintPass for BuiltinCombinedEarlyLintPass {
2 fn check_param(&mut self, context: &EarlyContext<'_>, a: &ast::Param) {
3 self.UnusedParens.check_param(context, a);
4 self.UnusedBraces.check_param(context, a);
5 self.UnusedImportBraces.check_param(context, a);
6 ...
7 }
8 fn check_ident(&mut self, context: &EarlyContext<'_>, a: &ast::Ident) {
9 self.UnusedParens.check_ident(context, a);
10 self.UnusedBraces.check_ident(context, a);
11 self.UnusedImportBraces.check_ident(context, a);
12 ...
13 }
14 fn check_crate(&mut self, context: &EarlyContext<'_>, a: &ast::Crate) {
15 self.UnusedParens.check_crate(context, a);
16 self.UnusedBraces.check_crate(context, a);
17 self.UnusedImportBraces.check_crate(context, a);
18 ...
19 }
20 ...
21
22}BuiltinCombinedEarlyLintPass 的最终定义
通过以上宏的展开，BuiltinCombinedEarlyLintPass的定义实际为如下形式：
1pub struct BuiltinCombinedEarlyLintPass {
2 UnusedParens: UnusedParens,
3 UnusedBraces: UnusedBraces,
4 ...
5}
6
7impl BuiltinCombinedEarlyLintPass{
8 pub fn new() -> Self {
9 UnusedParens: UnusedParens,
10 UnusedBraces: UnusedBraces,
11 ...
12 }
13
14 pub fn get_lints() -> LintArray {
15 let mut lints = Vec::new();
16 lints.extend_from_slice(&UnusedParens::get_lints());
17 lints.extend_from_slice(&UnusedBraces::get_lints());
18 ...
19 lints
20 }
21}
22
23impl EarlyLintPass for BuiltinCombinedEarlyLintPass {
24 fn check_crates(&mut self, context: &EarlyContext<'_>, a: &ast::Crate){
25 self.UnusedParens.check_crates (context, a);
26 self.UnusedBraces.check_crates (context, a);
27 ...
28 }
29 fn check_ident(&mut self, context: &EarlyContext<'_>, a: Ident){
30 self.UnusedParens.check_ident (context, a);
31 self.UnusedBraces.check_ident (context, a);
32 ...
33 }
34 ..
35}通过这个定义，可以在遍历 AST 时使用 BuiltinCombinedEarlyLintPass 的 check_* 方法实现多个 lintpass 的检查。
Lint 的进一步优化
基于 CombinedLintPass ，可以对上一篇文章中提出的 Linter 的设计做进一步优化。
这里，可以用 CombinedLintPass 的check_* 方法，在 Visitor 遍历 AST 时执行对应的检查。虽然效果与之前一致，但因为宏的关系，所有的 check_* 方法和需要执行的 lintpass 都被收集到了一个结构中，也更容易管理。同样的，因为 CombinedLintPass 实际上调用的是每个 lintpass 各自的 check 方法，虽然调用起来可能下图一样很复杂，但因为 lintpass 中定义的 check 方法大部分是由宏生成的空检查，所以也不会造成性能上的损失。
总结
本文简单介绍了 Rustc 源码中关于 CombinedLintPass 这一结构的定义和实现 ，并以此进一步优化 Linter 的设计。希望能够对理解 Rustc 及 Lint 有所帮助，如有错误，欢迎指正。后续的文章将继续介绍 Rustc 中 Lint 在编译过程中的注册和执行过程，期待继续关注。
参考链接
Rust源码剖析(电子书)：https://github.com/awesome-kusion/rust-code-book
KCL 配置策略语言：https://github.com/KusionStack/KCLVM
KusionStack: https://github.com/KusionStack
Rustc: https://github.com/rust-lang/rust
rustc-dev-guide: https://rustc-dev-guide.rust-lang.org/
Rust Visitor: https://doc.rust-lang.org/nightly/nightly-rustc/rustc_ast/visit/index.html
Rust Clippy: https://github.com/rust-lang/rust-clippy
了解更多...
KusionStack Star 一下✨：
https://github.com/KusionStack/Kusion