Merge TokenTreesReader into StringReader.

There is a not-very-useful layering in the lexer, where
`TokenTreesReader` contains a `StringReader`. This commit combines them
and names the result `Lexer`, which is a more obvious name for it.

The methods of `Lexer` are now split across `mod.rs` and `tokentrees.rs`
which isn't ideal, but it doesn't seem worth moving a bunch of code to
avoid it.

Author: nnethercote

compiler/rustc_parse/src/lexer/mod.rs

@@ -18,6 +18,7 @@ use rustc_span::symbol::Symbol;
 use rustc_span::{BytePos, Pos, Span};
 use tracing::debug;
 
+use crate::lexer::diagnostics::TokenTreeDiagInfo;
 use crate::lexer::unicode_chars::UNICODE_ARRAY;
 use crate::{errors, make_unclosed_delims_error};
 
@@ -56,7 +57,7 @@ pub(crate) fn lex_token_trees<'psess, 'src>(
     }
 
     let cursor = Cursor::new(src);
-    let string_reader = StringReader {
+    let mut lexer = Lexer {
         psess,
         start_pos,
         pos: start_pos,
@@ -65,9 +66,12 @@ pub(crate) fn lex_token_trees<'psess, 'src>(
         override_span,
         nbsp_is_whitespace: false,
         last_lifetime: None,
+        token: Token::dummy(),
+        diag_info: TokenTreeDiagInfo::default(),
     };
-    let (stream, res, unmatched_delims) =
-        tokentrees::TokenTreesReader::lex_all_token_trees(string_reader);
+    let (_open_spacing, stream, res) = lexer.lex_token_trees(/* is_delimited */ false);
+    let unmatched_delims = lexer.diag_info.unmatched_delims;
+
     match res {
         Ok(()) if unmatched_delims.is_empty() => Ok(stream),
         _ => {
@@ -92,7 +96,7 @@ pub(crate) fn lex_token_trees<'psess, 'src>(
     }
 }
 
-struct StringReader<'psess, 'src> {
+struct Lexer<'psess, 'src> {
     psess: &'psess ParseSess,
     /// Initial position, read-only.
     start_pos: BytePos,
@@ -111,9 +115,14 @@ struct StringReader<'psess, 'src> {
     /// Track the `Span` for the leading `'` of the last lifetime. Used for
     /// diagnostics to detect possible typo where `"` was meant.
     last_lifetime: Option<Span>,
+
+    /// The current token.
+    token: Token,
+
+    diag_info: TokenTreeDiagInfo,
 }
 
-impl<'psess, 'src> StringReader<'psess, 'src> {
+impl<'psess, 'src> Lexer<'psess, 'src> {
     fn dcx(&self) -> DiagCtxtHandle<'psess> {
         self.psess.dcx()
     }
@@ -124,7 +133,7 @@ impl<'psess, 'src> StringReader<'psess, 'src> {
 
     /// Returns the next token, paired with a bool indicating if the token was
     /// preceded by whitespace.
-    fn next_token(&mut self) -> (Token, bool) {
+    fn next_token_from_cursor(&mut self) -> (Token, bool) {
         let mut preceded_by_whitespace = false;
         let mut swallow_next_invalid = 0;
         // Skip trivial (whitespace & comments) tokens

compiler/rustc_parse/src/lexer/tokentrees.rs

@@ -4,41 +4,19 @@ use rustc_ast_pretty::pprust::token_to_string;
 use rustc_errors::{Applicability, PErr};
 use rustc_span::symbol::kw;
 
-use super::diagnostics::{
-    TokenTreeDiagInfo, report_suspicious_mismatch_block, same_indentation_level,
-};
-use super::{StringReader, UnmatchedDelim};
+use super::diagnostics::{report_suspicious_mismatch_block, same_indentation_level};
+use super::{Lexer, UnmatchedDelim};
 use crate::Parser;
 
-pub(super) struct TokenTreesReader<'psess, 'src> {
-    string_reader: StringReader<'psess, 'src>,
-    /// The "next" token, which has been obtained from the `StringReader` but
-    /// not yet handled by the `TokenTreesReader`.
-    token: Token,
-    diag_info: TokenTreeDiagInfo,
-}
-
-impl<'psess, 'src> TokenTreesReader<'psess, 'src> {
-    pub(super) fn lex_all_token_trees(
-        string_reader: StringReader<'psess, 'src>,
-    ) -> (TokenStream, Result<(), Vec<PErr<'psess>>>, Vec<UnmatchedDelim>) {
-        let mut tt_reader = TokenTreesReader {
-            string_reader,
-            token: Token::dummy(),
-            diag_info: TokenTreeDiagInfo::default(),
-        };
-        let (_open_spacing, stream, res) = tt_reader.lex_token_trees(/* is_delimited */ false);
-        (stream, res, tt_reader.diag_info.unmatched_delims)
-    }
-
+impl<'psess, 'src> Lexer<'psess, 'src> {
     // Lex into a token stream. The `Spacing` in the result is that of the
     // opening delimiter.
-    fn lex_token_trees(
+    pub(super) fn lex_token_trees(
         &mut self,
         is_delimited: bool,
     ) -> (Spacing, TokenStream, Result<(), Vec<PErr<'psess>>>) {
         // Move past the opening delimiter.
-        let (_, open_spacing) = self.bump(false);
+        let open_spacing = self.bump(false).1;
 
         let mut buf = Vec::new();
         loop {
@@ -72,7 +50,7 @@ impl<'psess, 'src> TokenTreesReader<'psess, 'src> {
 
     fn eof_err(&mut self) -> PErr<'psess> {
         let msg = "this file contains an unclosed delimiter";
-        let mut err = self.string_reader.dcx().struct_span_err(self.token.span, msg);
+        let mut err = self.dcx().struct_span_err(self.token.span, msg);
 
         let unclosed_delimiter_show_limit = 5;
         let len = usize::min(unclosed_delimiter_show_limit, self.diag_info.open_braces.len());
@@ -102,7 +80,7 @@ impl<'psess, 'src> TokenTreesReader<'psess, 'src> {
             report_suspicious_mismatch_block(
                 &mut err,
                 &self.diag_info,
-                self.string_reader.psess.source_map(),
+                self.psess.source_map(),
                 *delim,
             )
         }
@@ -128,7 +106,7 @@ impl<'psess, 'src> TokenTreesReader<'psess, 'src> {
 
         // Expand to cover the entire delimited token tree.
         let delim_span = DelimSpan::from_pair(pre_span, self.token.span);
-        let sm = self.string_reader.psess.source_map();
+        let sm = self.psess.source_map();
 
         let close_spacing = match self.token.kind {
             // Correct delimiter.
@@ -220,7 +198,7 @@ impl<'psess, 'src> TokenTreesReader<'psess, 'src> {
     // Will glue adjacent single-char tokens together if `glue` is set.
     fn bump(&mut self, glue: bool) -> (Token, Spacing) {
         let (this_spacing, next_tok) = loop {
-            let (next_tok, is_next_tok_preceded_by_whitespace) = self.string_reader.next_token();
+            let (next_tok, is_next_tok_preceded_by_whitespace) = self.next_token_from_cursor();
 
             if is_next_tok_preceded_by_whitespace {
                 break (Spacing::Alone, next_tok);
@@ -248,7 +226,7 @@ impl<'psess, 'src> TokenTreesReader<'psess, 'src> {
     ) -> Vec<PErr<'psess>> {
         // If there are unclosed delims, see if there are diff markers and if so, point them
         // out instead of complaining about the unclosed delims.
-        let mut parser = Parser::new(self.string_reader.psess, tts, None);
+        let mut parser = Parser::new(self.psess, tts, None);
         let mut diff_errs = vec![];
         // Suggest removing a `{` we think appears in an `if`/`while` condition.
         // We want to suggest removing a `{` only if we think we're in an `if`/`while` condition,
@@ -306,14 +284,9 @@ impl<'psess, 'src> TokenTreesReader<'psess, 'src> {
         // An unexpected closing delimiter (i.e., there is no matching opening delimiter).
         let token_str = token_to_string(&self.token);
         let msg = format!("unexpected closing delimiter: `{token_str}`");
-        let mut err = self.string_reader.dcx().struct_span_err(self.token.span, msg);
+        let mut err = self.dcx().struct_span_err(self.token.span, msg);
 
-        report_suspicious_mismatch_block(
-            &mut err,
-            &self.diag_info,
-            self.string_reader.psess.source_map(),
-            delim,
-        );
+        report_suspicious_mismatch_block(&mut err, &self.diag_info, self.psess.source_map(), delim);
         err.span_label(self.token.span, "unexpected closing delimiter");
         err
     }

compiler/rustc_parse/src/lexer/unicode_chars.rs

@@ -4,7 +4,7 @@
 use rustc_span::symbol::kw;
 use rustc_span::{BytePos, Pos, Span};
 
-use super::StringReader;
+use super::Lexer;
 use crate::errors::TokenSubstitution;
 use crate::token::{self, Delimiter};
 
@@ -338,7 +338,7 @@ const ASCII_ARRAY: &[(&str, &str, Option<token::TokenKind>)] = &[
 ];
 
 pub(super) fn check_for_substitution(
-    reader: &StringReader<'_, '_>,
+    lexer: &Lexer<'_, '_>,
     pos: BytePos,
     ch: char,
     count: usize,
@@ -351,11 +351,11 @@ pub(super) fn check_for_substitution(
 
     let Some((_, ascii_name, token)) = ASCII_ARRAY.iter().find(|&&(s, _, _)| s == ascii_str) else {
         let msg = format!("substitution character not found for '{ch}'");
-        reader.dcx().span_bug(span, msg);
+        lexer.dcx().span_bug(span, msg);
     };
 
     // special help suggestion for "directed" double quotes
-    let sugg = if let Some(s) = peek_delimited(&reader.src[reader.src_index(pos)..], '“', '”') {
+    let sugg = if let Some(s) = peek_delimited(&lexer.src[lexer.src_index(pos)..], '“', '”') {
         let span = Span::with_root_ctxt(
             pos,
             pos + Pos::from_usize('“'.len_utf8() + s.len() + '”'.len_utf8()),