Auto merge of #106745 - m-ou-se:format-args-ast, r=oli-obk

Move format_args!() into AST (and expand it during AST lowering)

Implements https://github.com/rust-lang/compiler-team/issues/541

This moves FormatArgs from rustc_builtin_macros to rustc_ast_lowering. For now, the end result is the same. But this allows for future changes to do smarter things with format_args!(). It also allows Clippy to directly access the ast::FormatArgs, making things a lot easier.

This change turns the format args types into lang items. The builtin macro used to refer to them by their path. After this change, the path is no longer relevant, making it easier to make changes in `core`.

This updates clippy to use the new language items, but this doesn't yet make clippy use the ast::FormatArgs structure that's now available. That should be done after this is merged.

compiler/rustc_ast/src/ast.rs

@@ -18,6 +18,7 @@
 //! - [`Attribute`]: Metadata associated with item.
 //! - [`UnOp`], [`BinOp`], and [`BinOpKind`]: Unary and binary operators.
 
+pub use crate::format::*;
 pub use crate::util::parser::ExprPrecedence;
 pub use GenericArgs::*;
 pub use UnsafeSource::*;
@@ -1269,6 +1270,7 @@ impl Expr {
             ExprKind::Try(..) => ExprPrecedence::Try,
             ExprKind::Yield(..) => ExprPrecedence::Yield,
             ExprKind::Yeet(..) => ExprPrecedence::Yeet,
+            ExprKind::FormatArgs(..) => ExprPrecedence::FormatArgs,
             ExprKind::Err => ExprPrecedence::Err,
         }
     }
@@ -1499,6 +1501,9 @@ pub enum ExprKind {
     /// with a `ByteStr` literal.
     IncludedBytes(Lrc<[u8]>),
 
+    /// A `format_args!()` expression.
+    FormatArgs(P<FormatArgs>),
+
     /// Placeholder for an expression that wasn't syntactically well formed in some way.
     Err,
 }

compiler/rustc_ast/src/format.rs

@@ -0,0 +1,250 @@
+use crate::ptr::P;
+use crate::Expr;
+use rustc_data_structures::fx::FxHashMap;
+use rustc_span::symbol::{Ident, Symbol};
+use rustc_span::Span;
+
+// Definitions:
+//
+// format_args!("hello {abc:.xyz$}!!", abc="world");
+// └──────────────────────────────────────────────┘
+//                     FormatArgs
+//
+// format_args!("hello {abc:.xyz$}!!", abc="world");
+//                                     └─────────┘
+//                                      argument
+//
+// format_args!("hello {abc:.xyz$}!!", abc="world");
+//              └───────────────────┘
+//                     template
+//
+// format_args!("hello {abc:.xyz$}!!", abc="world");
+//               └────┘└─────────┘└┘
+//                      pieces
+//
+// format_args!("hello {abc:.xyz$}!!", abc="world");
+//               └────┘           └┘
+//                   literal pieces
+//
+// format_args!("hello {abc:.xyz$}!!", abc="world");
+//                     └─────────┘
+//                     placeholder
+//
+// format_args!("hello {abc:.xyz$}!!", abc="world");
+//                      └─┘  └─┘
+//                      positions (could be names, numbers, empty, or `*`)
+
+/// (Parsed) format args.
+///
+/// Basically the "AST" for a complete `format_args!()`.
+///
+/// E.g., `format_args!("hello {name}");`.
+#[derive(Clone, Encodable, Decodable, Debug)]
+pub struct FormatArgs {
+    pub span: Span,
+    pub template: Vec<FormatArgsPiece>,
+    pub arguments: FormatArguments,
+}
+
+/// A piece of a format template string.
+///
+/// E.g. "hello" or "{name}".
+#[derive(Clone, Encodable, Decodable, Debug)]
+pub enum FormatArgsPiece {
+    Literal(Symbol),
+    Placeholder(FormatPlaceholder),
+}
+
+/// The arguments to format_args!().
+///
+/// E.g. `1, 2, name="ferris", n=3`,
+/// but also implicit captured arguments like `x` in `format_args!("{x}")`.
+#[derive(Clone, Encodable, Decodable, Debug)]
+pub struct FormatArguments {
+    arguments: Vec<FormatArgument>,
+    num_unnamed_args: usize,
+    num_explicit_args: usize,
+    names: FxHashMap<Symbol, usize>,
+}
+
+// FIXME: Rustdoc has trouble proving Send/Sync for this. See #106930.
+#[cfg(parallel_compiler)]
+unsafe impl Sync for FormatArguments {}
+#[cfg(parallel_compiler)]
+unsafe impl Send for FormatArguments {}
+
+impl FormatArguments {
+    pub fn new() -> Self {
+        Self {
+            arguments: Vec::new(),
+            names: FxHashMap::default(),
+            num_unnamed_args: 0,
+            num_explicit_args: 0,
+        }
+    }
+
+    pub fn add(&mut self, arg: FormatArgument) -> usize {
+        let index = self.arguments.len();
+        if let Some(name) = arg.kind.ident() {
+            self.names.insert(name.name, index);
+        } else if self.names.is_empty() {
+            // Only count the unnamed args before the first named arg.
+            // (Any later ones are errors.)
+            self.num_unnamed_args += 1;
+        }
+        if !matches!(arg.kind, FormatArgumentKind::Captured(..)) {
+            // This is an explicit argument.
+            // Make sure that all arguments so far are explcit.
+            assert_eq!(
+                self.num_explicit_args,
+                self.arguments.len(),
+                "captured arguments must be added last"
+            );
+            self.num_explicit_args += 1;
+        }
+        self.arguments.push(arg);
+        index
+    }
+
+    pub fn by_name(&self, name: Symbol) -> Option<(usize, &FormatArgument)> {
+        let i = *self.names.get(&name)?;
+        Some((i, &self.arguments[i]))
+    }
+
+    pub fn by_index(&self, i: usize) -> Option<&FormatArgument> {
+        (i < self.num_explicit_args).then(|| &self.arguments[i])
+    }
+
+    pub fn unnamed_args(&self) -> &[FormatArgument] {
+        &self.arguments[..self.num_unnamed_args]
+    }
+
+    pub fn named_args(&self) -> &[FormatArgument] {
+        &self.arguments[self.num_unnamed_args..self.num_explicit_args]
+    }
+
+    pub fn explicit_args(&self) -> &[FormatArgument] {
+        &self.arguments[..self.num_explicit_args]
+    }
+
+    pub fn all_args(&self) -> &[FormatArgument] {
+        &self.arguments[..]
+    }
+
+    pub fn all_args_mut(&mut self) -> &mut [FormatArgument] {
+        &mut self.arguments[..]
+    }
+}
+
+#[derive(Clone, Encodable, Decodable, Debug)]
+pub struct FormatArgument {
+    pub kind: FormatArgumentKind,
+    pub expr: P<Expr>,
+}
+
+#[derive(Clone, Encodable, Decodable, Debug)]
+pub enum FormatArgumentKind {
+    /// `format_args(…, arg)`
+    Normal,
+    /// `format_args(…, arg = 1)`
+    Named(Ident),
+    /// `format_args("… {arg} …")`
+    Captured(Ident),
+}
+
+impl FormatArgumentKind {
+    pub fn ident(&self) -> Option<Ident> {
+        match self {
+            &Self::Normal => None,
+            &Self::Named(id) => Some(id),
+            &Self::Captured(id) => Some(id),
+        }
+    }
+}
+
+#[derive(Clone, Encodable, Decodable, Debug, PartialEq, Eq)]
+pub struct FormatPlaceholder {
+    /// Index into [`FormatArgs::arguments`].
+    pub argument: FormatArgPosition,
+    /// The span inside the format string for the full `{…}` placeholder.
+    pub span: Option<Span>,
+    /// `{}`, `{:?}`, or `{:x}`, etc.
+    pub format_trait: FormatTrait,
+    /// `{}` or `{:.5}` or `{:-^20}`, etc.
+    pub format_options: FormatOptions,
+}
+
+#[derive(Clone, Encodable, Decodable, Debug, PartialEq, Eq)]
+pub struct FormatArgPosition {
+    /// Which argument this position refers to (Ok),
+    /// or would've referred to if it existed (Err).
+    pub index: Result<usize, usize>,
+    /// What kind of position this is. See [`FormatArgPositionKind`].
+    pub kind: FormatArgPositionKind,
+    /// The span of the name or number.
+    pub span: Option<Span>,
+}
+
+#[derive(Copy, Clone, Encodable, Decodable, Debug, PartialEq, Eq)]
+pub enum FormatArgPositionKind {
+    /// `{}` or `{:.*}`
+    Implicit,
+    /// `{1}` or `{:1$}` or `{:.1$}`
+    Number,
+    /// `{a}` or `{:a$}` or `{:.a$}`
+    Named,
+}
+
+#[derive(Copy, Clone, Encodable, Decodable, Debug, PartialEq, Eq, Hash)]
+pub enum FormatTrait {
+    /// `{}`
+    Display,
+    /// `{:?}`
+    Debug,
+    /// `{:e}`
+    LowerExp,
+    /// `{:E}`
+    UpperExp,
+    /// `{:o}`
+    Octal,
+    /// `{:p}`
+    Pointer,
+    /// `{:b}`
+    Binary,
+    /// `{:x}`
+    LowerHex,
+    /// `{:X}`
+    UpperHex,
+}
+
+#[derive(Clone, Encodable, Decodable, Default, Debug, PartialEq, Eq)]
+pub struct FormatOptions {
+    /// The width. E.g. `{:5}` or `{:width$}`.
+    pub width: Option<FormatCount>,
+    /// The precision. E.g. `{:.5}` or `{:.precision$}`.
+    pub precision: Option<FormatCount>,
+    /// The alignment. E.g. `{:>}` or `{:<}` or `{:^}`.
+    pub alignment: Option<FormatAlignment>,
+    /// The fill character. E.g. the `.` in `{:.>10}`.
+    pub fill: Option<char>,
+    /// The `+`, `-`, `0`, `#`, `x?` and `X?` flags.
+    pub flags: u32,
+}
+
+#[derive(Copy, Clone, Encodable, Decodable, Debug, PartialEq, Eq)]
+pub enum FormatAlignment {
+    /// `{:<}`
+    Left,
+    /// `{:>}`
+    Right,
+    /// `{:^}`
+    Center,
+}
+
+#[derive(Clone, Encodable, Decodable, Debug, PartialEq, Eq)]
+pub enum FormatCount {
+    /// `{:5}` or `{:.5}`
+    Literal(usize),
+    /// `{:.*}`, `{:.5$}`, or `{:a$}`, etc.
+    Argument(FormatArgPosition),
+}

compiler/rustc_ast/src/lib.rs

@@ -42,6 +42,7 @@ pub mod ast_traits;
 pub mod attr;
 pub mod entry;
 pub mod expand;
+pub mod format;
 pub mod mut_visit;
 pub mod node_id;
 pub mod ptr;
@@ -51,6 +52,7 @@ pub mod visit;
 
 pub use self::ast::*;
 pub use self::ast_traits::{AstDeref, AstNodeWrapper, HasAttrs, HasNodeId, HasSpan, HasTokens};
+pub use self::format::*;
 
 use rustc_data_structures::stable_hasher::{HashStable, StableHasher};
 

compiler/rustc_ast/src/mut_visit.rs

@@ -297,6 +297,10 @@ pub trait MutVisitor: Sized {
     fn visit_inline_asm_sym(&mut self, sym: &mut InlineAsmSym) {
         noop_visit_inline_asm_sym(sym, self)
     }
+
+    fn visit_format_args(&mut self, fmt: &mut FormatArgs) {
+        noop_visit_format_args(fmt, self)
+    }
 }
 
 /// Use a map-style function (`FnOnce(T) -> T`) to overwrite a `&mut T`. Useful
@@ -1284,6 +1288,15 @@ pub fn noop_visit_inline_asm_sym<T: MutVisitor>(
     vis.visit_path(path);
 }
 
+pub fn noop_visit_format_args<T: MutVisitor>(fmt: &mut FormatArgs, vis: &mut T) {
+    for arg in fmt.arguments.all_args_mut() {
+        if let FormatArgumentKind::Named(name) = &mut arg.kind {
+            vis.visit_ident(name);
+        }
+        vis.visit_expr(&mut arg.expr);
+    }
+}
+
 pub fn noop_visit_expr<T: MutVisitor>(
     Expr { kind, id, span, attrs, tokens }: &mut Expr,
     vis: &mut T,
@@ -1425,6 +1438,7 @@ pub fn noop_visit_expr<T: MutVisitor>(
             visit_opt(expr, |expr| vis.visit_expr(expr));
         }
         ExprKind::InlineAsm(asm) => vis.visit_inline_asm(asm),
+        ExprKind::FormatArgs(fmt) => vis.visit_format_args(fmt),
         ExprKind::MacCall(mac) => vis.visit_mac_call(mac),
         ExprKind::Struct(se) => {
             let StructExpr { qself, path, fields, rest } = se.deref_mut();

compiler/rustc_ast/src/util/parser.rs

@@ -271,6 +271,7 @@ pub enum ExprPrecedence {
     Try,
     InlineAsm,
     Mac,
+    FormatArgs,
 
     Array,
     Repeat,
@@ -335,7 +336,8 @@ impl ExprPrecedence {
             | ExprPrecedence::Index
             | ExprPrecedence::Try
             | ExprPrecedence::InlineAsm
-            | ExprPrecedence::Mac => PREC_POSTFIX,
+            | ExprPrecedence::Mac
+            | ExprPrecedence::FormatArgs => PREC_POSTFIX,
 
             // Never need parens
             ExprPrecedence::Array

compiler/rustc_ast/src/visit.rs

@@ -242,6 +242,9 @@ pub trait Visitor<'ast>: Sized {
     fn visit_inline_asm(&mut self, asm: &'ast InlineAsm) {
         walk_inline_asm(self, asm)
     }
+    fn visit_format_args(&mut self, fmt: &'ast FormatArgs) {
+        walk_format_args(self, fmt)
+    }
     fn visit_inline_asm_sym(&mut self, sym: &'ast InlineAsmSym) {
         walk_inline_asm_sym(self, sym)
     }
@@ -756,6 +759,15 @@ pub fn walk_inline_asm_sym<'a, V: Visitor<'a>>(visitor: &mut V, sym: &'a InlineA
     visitor.visit_path(&sym.path, sym.id);
 }
 
+pub fn walk_format_args<'a, V: Visitor<'a>>(visitor: &mut V, fmt: &'a FormatArgs) {
+    for arg in fmt.arguments.all_args() {
+        if let FormatArgumentKind::Named(name) = arg.kind {
+            visitor.visit_ident(name);
+        }
+        visitor.visit_expr(&arg.expr);
+    }
+}
+
 pub fn walk_expr<'a, V: Visitor<'a>>(visitor: &mut V, expression: &'a Expr) {
     walk_list!(visitor, visit_attribute, expression.attrs.iter());
 
@@ -896,6 +908,7 @@ pub fn walk_expr<'a, V: Visitor<'a>>(visitor: &mut V, expression: &'a Expr) {
         ExprKind::MacCall(mac) => visitor.visit_mac_call(mac),
         ExprKind::Paren(subexpression) => visitor.visit_expr(subexpression),
         ExprKind::InlineAsm(asm) => visitor.visit_inline_asm(asm),
+        ExprKind::FormatArgs(f) => visitor.visit_format_args(f),
         ExprKind::Yield(optional_expression) => {
             walk_list!(visitor, visit_expr, optional_expression);
         }