Validate and transcribe raw strings via unescape module

Author: Xanewok

src/libsyntax/parse/lexer/mod.rs

@@ -1086,10 +1086,12 @@ impl<'a> StringReader<'a> {
                 Ok(TokenKind::lit(token::Str, symbol, suffix))
             }
             'r' => {
-                let (kind, symbol) = self.scan_raw_string();
+                let (start, end, hash_count) = self.scan_raw_string();
+                let symbol = self.name_from_to(start, end);
+                self.validate_raw_str_escape(start, end);
                 let suffix = self.scan_optional_raw_name();
 
-                Ok(TokenKind::lit(kind, symbol, suffix))
+                Ok(TokenKind::lit(token::StrRaw(hash_count), symbol, suffix))
             }
             '-' => {
                 if self.nextch_is('>') {
@@ -1243,7 +1245,7 @@ impl<'a> StringReader<'a> {
         id
     }
 
-    fn scan_raw_string(&mut self) -> (token::LitKind, Symbol) {
+    fn scan_raw_string(&mut self) -> (BytePos, BytePos, u16) {
         let start_bpos = self.pos;
         self.bump();
         let mut hash_count: u16 = 0;
@@ -1273,7 +1275,6 @@ impl<'a> StringReader<'a> {
         self.bump();
         let content_start_bpos = self.pos;
         let mut content_end_bpos;
-        let mut valid = true;
         'outer: loop {
             match self.ch {
                 None => {
@@ -1289,29 +1290,14 @@ impl<'a> StringReader<'a> {
                     }
                     break;
                 }
-                Some(c) => {
-                    if c == '\r' && !self.nextch_is('\n') {
-                        let last_bpos = self.pos;
-                        self.err_span_(start_bpos,
-                                        last_bpos,
-                                        "bare CR not allowed in raw string, use \\r \
-                                        instead");
-                        valid = false;
-                    }
-                }
+                _ => (),
             }
             self.bump();
         }
 
         self.bump();
 
-        let symbol = if valid {
-            self.name_from_to(content_start_bpos, content_end_bpos)
-        } else {
-            Symbol::intern("??")
-        };
-
-        (token::StrRaw(hash_count), symbol)
+        (content_start_bpos, content_end_bpos, hash_count)
     }
 
     fn scan_raw_byte_string(&mut self) -> (token::LitKind, Symbol) {
@@ -1421,6 +1407,23 @@ impl<'a> StringReader<'a> {
         });
     }
 
+    fn validate_raw_str_escape(&self, content_start: BytePos, content_end: BytePos) {
+        self.with_str_from_to(content_start, content_end, |lit: &str| {
+            unescape::unescape_raw_str(lit, &mut |range, c| {
+                if let Err(err) = c {
+                    emit_unescape_error(
+                        &self.sess.span_diagnostic,
+                        lit,
+                        self.mk_sp(content_start - BytePos(1), content_end + BytePos(1)),
+                        unescape::Mode::Str,
+                        range,
+                        err,
+                    )
+                }
+            })
+        });
+    }
+
     fn validate_byte_str_escape(&self, start_with_quote: BytePos) {
         self.with_str_from_to(start_with_quote + BytePos(1), self.pos - BytePos(1), |lit| {
             unescape::unescape_byte_str(lit, &mut |range, c| {

src/libsyntax/parse/literal.rs

@@ -4,7 +4,8 @@ use crate::ast::{self, Lit, LitKind};
 use crate::parse::parser::Parser;
 use crate::parse::PResult;
 use crate::parse::token::{self, Token, TokenKind};
-use crate::parse::unescape::{unescape_str, unescape_char, unescape_byte_str, unescape_byte};
+use crate::parse::unescape::{unescape_str, unescape_byte_str, unescape_raw_str};
+use crate::parse::unescape::{unescape_char, unescape_byte};
 use crate::print::pprust;
 use crate::symbol::{kw, sym, Symbol};
 use crate::tokenstream::{TokenStream, TokenTree};
@@ -141,7 +142,17 @@ impl LitKind {
                 // Ditto.
                 let s = symbol.as_str();
                 let symbol = if s.contains('\r') {
-                    Symbol::intern(&raw_str_lit(&s))
+                    let mut buf = String::with_capacity(s.len());
+                    let mut error = Ok(());
+                    unescape_raw_str(&s, &mut |_, unescaped_char| {
+                        match unescaped_char {
+                            Ok(c) => buf.push(c),
+                            Err(_) => error = Err(LitError::LexerError),
+                        }
+                    });
+                    error?;
+                    buf.shrink_to_fit();
+                    Symbol::intern(&buf)
                 } else {
                     symbol
                 };
@@ -350,29 +361,6 @@ crate fn expect_no_suffix(diag: &Handler, sp: Span, kind: &str, suffix: Option<S
     }
 }
 
-/// Parses a string representing a raw string literal into its final form. The
-/// only operation this does is convert embedded CRLF into a single LF.
-fn raw_str_lit(lit: &str) -> String {
-    debug!("raw_str_lit: {:?}", lit);
-    let mut res = String::with_capacity(lit.len());
-
-    let mut chars = lit.chars().peekable();
-    while let Some(c) = chars.next() {
-        if c == '\r' {
-            if *chars.peek().unwrap() != '\n' {
-                panic!("lexer accepted bare CR");
-            }
-            chars.next();
-            res.push('\n');
-        } else {
-            res.push(c);
-        }
-    }
-
-    res.shrink_to_fit();
-    res
-}
-
 // Checks if `s` looks like i32 or u1234 etc.
 fn looks_like_width_suffix(first_chars: &[char], s: &str) -> bool {
     s.len() > 1 && s.starts_with(first_chars) && s[1..].chars().all(|c| c.is_ascii_digit())

src/libsyntax/parse/unescape.rs

@@ -66,6 +66,28 @@ where
     })
 }
 
+/// Takes a contents of a string literal (without quotes) and produces a
+/// sequence of characters or errors.
+/// NOTE: Raw strings do not perform any explicit character escaping, here we
+/// only translate CRLF to LF and produce errors on bare CR.
+pub(crate) fn unescape_raw_str<F>(literal_text: &str, callback: &mut F)
+where
+    F: FnMut(Range<usize>, Result<char, EscapeError>),
+{
+    let mut byte_offset: usize = 0;
+
+    let mut chars = literal_text.chars().peekable();
+    while let Some(curr) = chars.next() {
+        let result = match (curr, chars.peek()) {
+            ('\r', Some('\n')) => Ok(curr),
+            ('\r', _) => Err(EscapeError::BareCarriageReturn),
+            _ => Ok(curr),
+        };
+        callback(byte_offset..(byte_offset + curr.len_utf8()), result);
+        byte_offset += curr.len_utf8();
+    }
+}
+
 #[derive(Debug, Clone, Copy)]
 pub(crate) enum Mode {
     Char,

src/test/ui/parser/lex-bare-cr-string-literal-doc-comment.rs

@@ -21,7 +21,7 @@ fn main() {
     let _s = "foobar"; //~ ERROR: bare CR not allowed in string
 
     // the following string literal has a bare CR in it
-    let _s = r"barfoo"; //~ ERROR: bare CR not allowed in raw string
+    let _s = r"barfoo"; //~ ERROR: bare CR not allowed in string
 
     // the following string literal has a bare CR in it
     let _s = "foo\bar"; //~ ERROR: unknown character escape: \r

src/test/ui/parser/lex-bare-cr-string-literal-doc-comment.stderr

@@ -28,11 +28,11 @@ error: bare CR not allowed in string, use \r instead
 LL |     let _s = "foobar";
    |                  ^
 
-error: bare CR not allowed in raw string, use \r instead
-  --> $DIR/lex-bare-cr-string-literal-doc-comment.rs:24:14
+error: bare CR not allowed in string, use \r instead
+  --> $DIR/lex-bare-cr-string-literal-doc-comment.rs:24:19
    |
 LL |     let _s = r"barfoo";
-   |              ^^^^^
+   |                   ^
 
 error: unknown character escape: \r
   --> $DIR/lex-bare-cr-string-literal-doc-comment.rs:27:19