Introduce `ByteSymbol`.

It's like `Symbol` but for byte strings. The interner is now used for
both `Symbol` and `ByteSymbol`. E.g. if you intern `"dog"` and `b"dog"`
you'll get a `Symbol` and a `ByteSymbol` with the same index and the
characters will only be stored once.

The motivation for this is to eliminate the `Arc`s in `ast::LitKind`, to
make `ast::LitKind` impl `Copy`, and to avoid the need to arena-allocate
`ast::LitKind` in HIR. The latter change reduces peak memory by a
non-trivial amount on literal-heavy benchmarks such as `deep-vector` and
`tuple-stress`.

`Encoder`, `Decoder`, `SpanEncoder`, and `SpanDecoder` all get some
changes so that they can handle normal strings and byte strings.

This change does slow down compilation of programs that use
`include_bytes!` on large files, because the contents of those files are
now interned (hashed). This makes `include_bytes!` more similar to
`include_str!`, though `include_bytes!` contents still aren't escaped,
and hashing is still much cheaper than escaping.

3 files changed, 4 insertions, 5 deletions

diff --git a/compiler/rustc_hir/src/arena.rs b/compiler/rustc_hir/src/arena.rs
index b0dff635a9b..180cb6497e7 100644
--- a/compiler/rustc_hir/src/arena.rs
+++ b/compiler/rustc_hir/src/arena.rs
@@ -8,7 +8,6 @@ macro_rules! arena_types {
             [] asm_template: rustc_ast::InlineAsmTemplatePiece,
             [] attribute: rustc_hir::Attribute,
             [] owner_info: rustc_hir::OwnerInfo<'tcx>,
-            [] lit: rustc_hir::Lit,
             [] macro_def: rustc_ast::MacroDef,
         ]);
     )
diff --git a/compiler/rustc_hir/src/hir.rs b/compiler/rustc_hir/src/hir.rs
index 75dff588669..464ff1487da 100644
--- a/compiler/rustc_hir/src/hir.rs
+++ b/compiler/rustc_hir/src/hir.rs
@@ -1807,7 +1807,7 @@ pub struct PatExpr<'hir> {
 #[derive(Debug, Clone, Copy, HashStable_Generic)]
 pub enum PatExprKind<'hir> {
     Lit {
-        lit: &'hir Lit,
+        lit: Lit,
         // FIXME: move this into `Lit` and handle negated literal expressions
         // once instead of matching on unop neg expressions everywhere.
         negated: bool,
@@ -2734,7 +2734,7 @@ pub enum ExprKind<'hir> {
     /// A unary operation (e.g., `!x`, `*x`).
     Unary(UnOp, &'hir Expr<'hir>),
     /// A literal (e.g., `1`, `"foo"`).
-    Lit(&'hir Lit),
+    Lit(Lit),
     /// A cast (e.g., `foo as f64`).
     Cast(&'hir Expr<'hir>, &'hir Ty<'hir>),
     /// A type ascription (e.g., `x: Foo`). See RFC 3307.
diff --git a/compiler/rustc_hir/src/intravisit.rs b/compiler/rustc_hir/src/intravisit.rs
index 57e49625148..a0bc318e2ca 100644
--- a/compiler/rustc_hir/src/intravisit.rs
+++ b/compiler/rustc_hir/src/intravisit.rs
@@ -347,7 +347,7 @@ pub trait Visitor<'v>: Sized {
     fn visit_pat_expr(&mut self, expr: &'v PatExpr<'v>) -> Self::Result {
         walk_pat_expr(self, expr)
     }
-    fn visit_lit(&mut self, _hir_id: HirId, _lit: &'v Lit, _negated: bool) -> Self::Result {
+    fn visit_lit(&mut self, _hir_id: HirId, _lit: Lit, _negated: bool) -> Self::Result {
         Self::Result::output()
     }
     fn visit_anon_const(&mut self, c: &'v AnonConst) -> Self::Result {
@@ -786,7 +786,7 @@ pub fn walk_pat_expr<'v, V: Visitor<'v>>(visitor: &mut V, expr: &'v PatExpr<'v>)
     let PatExpr { hir_id, span, kind } = expr;
     try_visit!(visitor.visit_id(*hir_id));
     match kind {
-        PatExprKind::Lit { lit, negated } => visitor.visit_lit(*hir_id, lit, *negated),
+        PatExprKind::Lit { lit, negated } => visitor.visit_lit(*hir_id, *lit, *negated),
         PatExprKind::ConstBlock(c) => visitor.visit_inline_const(c),
         PatExprKind::Path(qpath) => visitor.visit_qpath(qpath, *hir_id, *span),
     }