Auto merge of #102302 - nnethercote:more-lexer-improvements, r=matklad

More lexer improvements

A follow-up to #99884.

r? `@matklad`

Author: bors

compiler/rustc_ast/src/token.rs

@@ -13,7 +13,7 @@ use rustc_span::symbol::{kw, sym};
 use rustc_span::symbol::{Ident, Symbol};
 use rustc_span::{self, edition::Edition, Span, DUMMY_SP};
 use std::borrow::Cow;
-use std::{fmt, mem};
+use std::fmt;
 
 #[derive(Clone, Copy, PartialEq, Encodable, Decodable, Debug, HashStable_Generic)]
 pub enum CommentKind {
@@ -335,11 +335,6 @@ impl Token {
         Token::new(Ident(ident.name, ident.is_raw_guess()), ident.span)
     }
 
-    /// Return this token by value and leave a dummy token in its place.
-    pub fn take(&mut self) -> Self {
-        mem::replace(self, Token::dummy())
-    }
-
     /// For interpolated tokens, returns a span of the fragment to which the interpolated
     /// token refers. For all other tokens this is just a regular span.
     /// It is particularly important to use this for identifiers and lifetimes

compiler/rustc_errors/src/lib.rs

@@ -62,7 +62,8 @@ pub mod translation;
 pub use diagnostic_builder::IntoDiagnostic;
 pub use snippet::Style;
 
-pub type PResult<'a, T> = Result<T, DiagnosticBuilder<'a, ErrorGuaranteed>>;
+pub type PErr<'a> = DiagnosticBuilder<'a, ErrorGuaranteed>;
+pub type PResult<'a, T> = Result<T, PErr<'a>>;
 
 // `PResult` is used a lot. Make sure it doesn't unintentionally get bigger.
 // (See also the comment on `DiagnosticBuilder`'s `diagnostic` field.)

compiler/rustc_lexer/src/cursor.rs

@@ -4,8 +4,8 @@ use std::str::Chars;
 ///
 /// Next characters can be peeked via `first` method,
 /// and position can be shifted forward via `bump` method.
-pub(crate) struct Cursor<'a> {
-    initial_len: usize,
+pub struct Cursor<'a> {
+    len_remaining: usize,
     /// Iterator over chars. Slightly faster than a &str.
     chars: Chars<'a>,
     #[cfg(debug_assertions)]
@@ -15,9 +15,9 @@ pub(crate) struct Cursor<'a> {
 pub(crate) const EOF_CHAR: char = '\0';
 
 impl<'a> Cursor<'a> {
-    pub(crate) fn new(input: &'a str) -> Cursor<'a> {
+    pub fn new(input: &'a str) -> Cursor<'a> {
         Cursor {
-            initial_len: input.len(),
+            len_remaining: input.len(),
             chars: input.chars(),
             #[cfg(debug_assertions)]
             prev: EOF_CHAR,
@@ -61,13 +61,13 @@ impl<'a> Cursor<'a> {
     }
 
     /// Returns amount of already consumed symbols.
-    pub(crate) fn len_consumed(&self) -> u32 {
-        (self.initial_len - self.chars.as_str().len()) as u32
+    pub(crate) fn pos_within_token(&self) -> u32 {
+        (self.len_remaining - self.chars.as_str().len()) as u32
     }
 
     /// Resets the number of bytes consumed to 0.
-    pub(crate) fn reset_len_consumed(&mut self) {
-        self.initial_len = self.chars.as_str().len();
+    pub(crate) fn reset_pos_within_token(&mut self) {
+        self.len_remaining = self.chars.as_str().len();
     }
 
     /// Moves to the next character.

compiler/rustc_lexer/src/lib.rs

@@ -29,9 +29,11 @@ pub mod unescape;
 #[cfg(test)]
 mod tests;
 
+pub use crate::cursor::Cursor;
+
 use self::LiteralKind::*;
 use self::TokenKind::*;
-use crate::cursor::{Cursor, EOF_CHAR};
+use crate::cursor::EOF_CHAR;
 use std::convert::TryFrom;
 
 /// Parsed token.
@@ -139,6 +141,9 @@ pub enum TokenKind {
 
     /// Unknown token, not expected by the lexer, e.g. "№"
     Unknown,
+
+    /// End of input.
+    Eof,
 }
 
 #[derive(Clone, Copy, Debug, PartialEq, Eq)]
@@ -219,13 +224,6 @@ pub fn strip_shebang(input: &str) -> Option<usize> {
     None
 }
 
-/// Parses the first token from the provided input string.
-#[inline]
-pub fn first_token(input: &str) -> Token {
-    debug_assert!(!input.is_empty());
-    Cursor::new(input).advance_token()
-}
-
 /// Validates a raw string literal. Used for getting more information about a
 /// problem with a `RawStr`/`RawByteStr` with a `None` field.
 #[inline]
@@ -243,12 +241,8 @@ pub fn validate_raw_str(input: &str, prefix_len: u32) -> Result<(), RawStrError>
 pub fn tokenize(input: &str) -> impl Iterator<Item = Token> + '_ {
     let mut cursor = Cursor::new(input);
     std::iter::from_fn(move || {
-        if cursor.is_eof() {
-            None
-        } else {
-            cursor.reset_len_consumed();
-            Some(cursor.advance_token())
-        }
+        let token = cursor.advance_token();
+        if token.kind != TokenKind::Eof { Some(token) } else { None }
     })
 }
 
@@ -311,8 +305,11 @@ pub fn is_ident(string: &str) -> bool {
 
 impl Cursor<'_> {
     /// Parses a token from the input string.
-    fn advance_token(&mut self) -> Token {
-        let first_char = self.bump().unwrap();
+    pub fn advance_token(&mut self) -> Token {
+        let first_char = match self.bump() {
+            Some(c) => c,
+            None => return Token::new(TokenKind::Eof, 0),
+        };
         let token_kind = match first_char {
             // Slash, comment or block comment.
             '/' => match self.first() {
@@ -329,7 +326,7 @@ impl Cursor<'_> {
                 ('#', c1) if is_id_start(c1) => self.raw_ident(),
                 ('#', _) | ('"', _) => {
                     let res = self.raw_double_quoted_string(1);
-                    let suffix_start = self.len_consumed();
+                    let suffix_start = self.pos_within_token();
                     if res.is_ok() {
                         self.eat_literal_suffix();
                     }
@@ -344,7 +341,7 @@ impl Cursor<'_> {
                 ('\'', _) => {
                     self.bump();
                     let terminated = self.single_quoted_string();
-                    let suffix_start = self.len_consumed();
+                    let suffix_start = self.pos_within_token();
                     if terminated {
                         self.eat_literal_suffix();
                     }
@@ -354,7 +351,7 @@ impl Cursor<'_> {
                 ('"', _) => {
                     self.bump();
                     let terminated = self.double_quoted_string();
-                    let suffix_start = self.len_consumed();
+                    let suffix_start = self.pos_within_token();
                     if terminated {
                         self.eat_literal_suffix();
                     }
@@ -364,7 +361,7 @@ impl Cursor<'_> {
                 ('r', '"') | ('r', '#') => {
                     self.bump();
                     let res = self.raw_double_quoted_string(2);
-                    let suffix_start = self.len_consumed();
+                    let suffix_start = self.pos_within_token();
                     if res.is_ok() {
                         self.eat_literal_suffix();
                     }
@@ -381,7 +378,7 @@ impl Cursor<'_> {
             // Numeric literal.
             c @ '0'..='9' => {
                 let literal_kind = self.number(c);
-                let suffix_start = self.len_consumed();
+                let suffix_start = self.pos_within_token();
                 self.eat_literal_suffix();
                 TokenKind::Literal { kind: literal_kind, suffix_start }
             }
@@ -420,7 +417,7 @@ impl Cursor<'_> {
             // String literal.
             '"' => {
                 let terminated = self.double_quoted_string();
-                let suffix_start = self.len_consumed();
+                let suffix_start = self.pos_within_token();
                 if terminated {
                     self.eat_literal_suffix();
                 }
@@ -433,7 +430,9 @@ impl Cursor<'_> {
             }
             _ => Unknown,
         };
-        Token::new(token_kind, self.len_consumed())
+        let res = Token::new(token_kind, self.pos_within_token());
+        self.reset_pos_within_token();
+        res
     }
 
     fn line_comment(&mut self) -> TokenKind {
@@ -618,7 +617,7 @@ impl Cursor<'_> {
 
         if !can_be_a_lifetime {
             let terminated = self.single_quoted_string();
-            let suffix_start = self.len_consumed();
+            let suffix_start = self.pos_within_token();
             if terminated {
                 self.eat_literal_suffix();
             }
@@ -643,7 +642,7 @@ impl Cursor<'_> {
         if self.first() == '\'' {
             self.bump();
             let kind = Char { terminated: true };
-            Literal { kind, suffix_start: self.len_consumed() }
+            Literal { kind, suffix_start: self.pos_within_token() }
         } else {
             Lifetime { starts_with_number }
         }
@@ -724,7 +723,7 @@ impl Cursor<'_> {
 
     fn raw_string_unvalidated(&mut self, prefix_len: u32) -> Result<u32, RawStrError> {
         debug_assert!(self.prev() == 'r');
-        let start_pos = self.len_consumed();
+        let start_pos = self.pos_within_token();
         let mut possible_terminator_offset = None;
         let mut max_hashes = 0;
 
@@ -778,7 +777,7 @@ impl Cursor<'_> {
                 // Keep track of possible terminators to give a hint about
                 // where there might be a missing terminator
                 possible_terminator_offset =
-                    Some(self.len_consumed() - start_pos - n_end_hashes + prefix_len);
+                    Some(self.pos_within_token() - start_pos - n_end_hashes + prefix_len);
                 max_hashes = n_end_hashes;
             }
         }