Merge #11046

11046: internal: move all the lexing to the parser crate r=matklad a=matklad

bors r+
🤖

Co-authored-by: Aleksey Kladov <aleksey.kladov@gmail.com>

Cargo.lock

@@ -609,6 +609,7 @@ dependencies = [
  "hir",
  "ide_db",
  "itertools",
+ "parser",
  "profile",
  "rustc-hash",
  "sourcegen",
@@ -654,6 +655,7 @@ dependencies = [
  "itertools",
  "limit",
  "once_cell",
+ "parser",
  "profile",
  "rayon",
  "rustc-hash",
@@ -695,6 +697,7 @@ dependencies = [
  "hir",
  "ide_db",
  "itertools",
+ "parser",
  "rustc-hash",
  "syntax",
  "test_utils",
@@ -1066,7 +1069,9 @@ name = "parser"
 version = "0.0.0"
 dependencies = [
  "drop_bomb",
+ "expect-test",
  "limit",
+ "rustc-ap-rustc_lexer",
 ]
 
 [[package]]

crates/ide_assists/Cargo.toml

@@ -16,6 +16,7 @@ itertools = "0.10.0"
 either = "1.6.1"
 
 stdx = { path = "../stdx", version = "0.0.0" }
+parser = { path = "../parser", version = "0.0.0" }
 syntax = { path = "../syntax", version = "0.0.0" }
 text_edit = { path = "../text_edit", version = "0.0.0" }
 profile = { path = "../profile", version = "0.0.0" }

crates/ide_assists/src/utils/suggest_name.rs

@@ -135,7 +135,7 @@ fn normalize(name: &str) -> Option<String> {
 }
 
 fn is_valid_name(name: &str) -> bool {
-    match syntax::lex_single_syntax_kind(name) {
+    match parser::LexedStr::single_token(name) {
         Some((syntax::SyntaxKind::IDENT, _error)) => true,
         _ => false,
     }

crates/ide_db/Cargo.toml

@@ -22,6 +22,7 @@ arrayvec = "0.7"
 indexmap = "1.7"
 
 stdx = { path = "../stdx", version = "0.0.0" }
+parser = { path = "../parser", version = "0.0.0" }
 syntax = { path = "../syntax", version = "0.0.0" }
 text_edit = { path = "../text_edit", version = "0.0.0" }
 base_db = { path = "../base_db", version = "0.0.0" }

crates/ide_db/src/rename.rs

@@ -28,7 +28,7 @@ use hir::{AsAssocItem, FieldSource, HasSource, InFile, ModuleSource, Semantics};
 use stdx::never;
 use syntax::{
     ast::{self, HasName},
-    lex_single_syntax_kind, AstNode, SyntaxKind, TextRange, T,
+    AstNode, SyntaxKind, TextRange, T,
 };
 use text_edit::{TextEdit, TextEditBuilder};
 
@@ -490,7 +490,7 @@ pub enum IdentifierKind {
 
 impl IdentifierKind {
     pub fn classify(new_name: &str) -> Result<IdentifierKind> {
-        match lex_single_syntax_kind(new_name) {
+        match parser::LexedStr::single_token(new_name) {
             Some(res) => match res {
                 (SyntaxKind::IDENT, _) => Ok(IdentifierKind::Ident),
                 (T![_], _) => Ok(IdentifierKind::Underscore),

crates/ide_ssr/Cargo.toml

@@ -16,6 +16,7 @@ rustc-hash = "1.1.0"
 itertools = "0.10.0"
 
 text_edit = { path = "../text_edit", version = "0.0.0" }
+parser = { path = "../parser", version = "0.0.0" }
 syntax = { path = "../syntax", version = "0.0.0" }
 ide_db = { path = "../ide_db", version = "0.0.0" }
 hir = { path = "../hir", version = "0.0.0" }

crates/ide_ssr/src/parsing.rs

@@ -256,19 +256,13 @@ fn validate_rule(rule: &SsrRule) -> Result<(), SsrError> {
 }
 
 fn tokenize(source: &str) -> Result<Vec<Token>, SsrError> {
-    let mut start = 0;
-    let (raw_tokens, errors) = syntax::tokenize(source);
-    if let Some(first_error) = errors.first() {
+    let lexed = parser::LexedStr::new(source);
+    if let Some((_, first_error)) = lexed.errors().next() {
         bail!("Failed to parse pattern: {}", first_error);
     }
     let mut tokens: Vec<Token> = Vec::new();
-    for raw_token in raw_tokens {
-        let token_len = usize::from(raw_token.len);
-        tokens.push(Token {
-            kind: raw_token.kind,
-            text: SmolStr::new(&source[start..start + token_len]),
-        });
-        start += token_len;
+    for i in 0..lexed.len() {
+        tokens.push(Token { kind: lexed.kind(i), text: lexed.text(i).into() });
     }
     Ok(tokens)
 }

crates/mbe/src/syntax_bridge.rs

@@ -4,10 +4,9 @@ use parser::{ParseError, TreeSink};
 use rustc_hash::{FxHashMap, FxHashSet};
 use syntax::{
     ast::{self, make::tokens::doc_comment},
-    tokenize, AstToken, Parse, PreorderWithTokens, SmolStr, SyntaxElement, SyntaxKind,
+    AstToken, Parse, PreorderWithTokens, SmolStr, SyntaxElement, SyntaxKind,
     SyntaxKind::*,
-    SyntaxNode, SyntaxToken, SyntaxTreeBuilder, TextRange, TextSize, Token as RawToken, WalkEvent,
-    T,
+    SyntaxNode, SyntaxToken, SyntaxTreeBuilder, TextRange, TextSize, WalkEvent, T,
 };
 use tt::buffer::{Cursor, TokenBuffer};
 
@@ -69,15 +68,14 @@ pub fn token_tree_to_syntax_node(
 
 /// Convert a string to a `TokenTree`
 pub fn parse_to_token_tree(text: &str) -> Option<(tt::Subtree, TokenMap)> {
-    let (tokens, errors) = tokenize(text);
-    if !errors.is_empty() {
+    let lexed = parser::LexedStr::new(text);
+    if lexed.errors().next().is_some() {
         return None;
     }
 
     let mut conv = RawConvertor {
-        text,
-        offset: TextSize::default(),
-        inner: tokens.iter(),
+        lexed: lexed,
+        pos: 0,
         id_alloc: TokenIdAlloc {
             map: Default::default(),
             global_offset: TextSize::default(),
@@ -146,7 +144,7 @@ fn convert_tokens<C: TokenConvertor>(conv: &mut C) -> tt::Subtree {
             Some(it) => it,
         };
 
-        let k: SyntaxKind = token.kind();
+        let k: SyntaxKind = token.kind(&conv);
         if k == COMMENT {
             if let Some(tokens) = conv.convert_doc_comment(&token) {
                 // FIXME: There has to be a better way to do this
@@ -199,19 +197,19 @@ fn convert_tokens<C: TokenConvertor>(conv: &mut C) -> tt::Subtree {
             } else {
                 let spacing = match conv.peek() {
                     Some(next)
-                        if next.kind().is_trivia()
-                            || next.kind() == T!['[']
-                            || next.kind() == T!['{']
-                            || next.kind() == T!['('] =>
+                        if next.kind(&conv).is_trivia()
+                            || next.kind(&conv) == T!['[']
+                            || next.kind(&conv) == T!['{']
+                            || next.kind(&conv) == T!['('] =>
                     {
                         tt::Spacing::Alone
                     }
-                    Some(next) if next.kind().is_punct() && next.kind() != UNDERSCORE => {
+                    Some(next) if next.kind(&conv).is_punct() && next.kind(&conv) != UNDERSCORE => {
                         tt::Spacing::Joint
                     }
                     _ => tt::Spacing::Alone,
                 };
-                let char = match token.to_char() {
+                let char = match token.to_char(&conv) {
                     Some(c) => c,
                     None => {
                         panic!("Token from lexer must be single char: token = {:#?}", token);
@@ -222,7 +220,7 @@ fn convert_tokens<C: TokenConvertor>(conv: &mut C) -> tt::Subtree {
         } else {
             macro_rules! make_leaf {
                 ($i:ident) => {
-                    tt::$i { id: conv.id_alloc().alloc(range), text: token.to_text() }.into()
+                    tt::$i { id: conv.id_alloc().alloc(range), text: token.to_text(conv) }.into()
                 };
             }
             let leaf: tt::Leaf = match k {
@@ -243,7 +241,7 @@ fn convert_tokens<C: TokenConvertor>(conv: &mut C) -> tt::Subtree {
 
                     let r = TextRange::at(range.start() + char_unit, range.len() - char_unit);
                     let ident = tt::Leaf::from(tt::Ident {
-                        text: SmolStr::new(&token.to_text()[1..]),
+                        text: SmolStr::new(&token.to_text(conv)[1..]),
                         id: conv.id_alloc().alloc(r),
                     });
                     result.push(ident.into());
@@ -392,22 +390,21 @@ impl TokenIdAlloc {
 
 /// A Raw Token (straightly from lexer) convertor
 struct RawConvertor<'a> {
-    text: &'a str,
-    offset: TextSize,
+    lexed: parser::LexedStr<'a>,
+    pos: usize,
     id_alloc: TokenIdAlloc,
-    inner: std::slice::Iter<'a, RawToken>,
 }
 
-trait SrcToken: std::fmt::Debug {
-    fn kind(&self) -> SyntaxKind;
+trait SrcToken<Ctx>: std::fmt::Debug {
+    fn kind(&self, ctx: &Ctx) -> SyntaxKind;
 
-    fn to_char(&self) -> Option<char>;
+    fn to_char(&self, ctx: &Ctx) -> Option<char>;
 
-    fn to_text(&self) -> SmolStr;
+    fn to_text(&self, ctx: &Ctx) -> SmolStr;
 }
 
-trait TokenConvertor {
-    type Token: SrcToken;
+trait TokenConvertor: Sized {
+    type Token: SrcToken<Self>;
 
     fn convert_doc_comment(&self, token: &Self::Token) -> Option<Vec<tt::TokenTree>>;
 
@@ -418,42 +415,45 @@ trait TokenConvertor {
     fn id_alloc(&mut self) -> &mut TokenIdAlloc;
 }
 
-impl<'a> SrcToken for (&'a RawToken, &'a str) {
-    fn kind(&self) -> SyntaxKind {
-        self.0.kind
+impl<'a> SrcToken<RawConvertor<'a>> for usize {
+    fn kind(&self, ctx: &RawConvertor<'a>) -> SyntaxKind {
+        ctx.lexed.kind(*self)
     }
 
-    fn to_char(&self) -> Option<char> {
-        self.1.chars().next()
+    fn to_char(&self, ctx: &RawConvertor<'a>) -> Option<char> {
+        ctx.lexed.text(*self).chars().next()
     }
 
-    fn to_text(&self) -> SmolStr {
-        self.1.into()
+    fn to_text(&self, ctx: &RawConvertor<'_>) -> SmolStr {
+        ctx.lexed.text(*self).into()
     }
 }
 
 impl<'a> TokenConvertor for RawConvertor<'a> {
-    type Token = (&'a RawToken, &'a str);
+    type Token = usize;
 
-    fn convert_doc_comment(&self, token: &Self::Token) -> Option<Vec<tt::TokenTree>> {
-        convert_doc_comment(&doc_comment(token.1))
+    fn convert_doc_comment(&self, token: &usize) -> Option<Vec<tt::TokenTree>> {
+        let text = self.lexed.text(*token);
+        convert_doc_comment(&doc_comment(text))
     }
 
     fn bump(&mut self) -> Option<(Self::Token, TextRange)> {
-        let token = self.inner.next()?;
-        let range = TextRange::at(self.offset, token.len);
-        self.offset += token.len;
+        if self.pos == self.lexed.len() {
+            return None;
+        }
+        let token = self.pos;
+        self.pos += 1;
+        let range = self.lexed.text_range(token);
+        let range = TextRange::new(range.start.try_into().unwrap(), range.end.try_into().unwrap());
 
-        Some(((token, &self.text[range]), range))
+        Some((token, range))
     }
 
     fn peek(&self) -> Option<Self::Token> {
-        let token = self.inner.as_slice().get(0);
-
-        token.map(|it| {
-            let range = TextRange::at(self.offset, it.len);
-            (it, &self.text[range])
-        })
+        if self.pos == self.lexed.len() {
+            return None;
+        }
+        Some(self.pos)
     }
 
     fn id_alloc(&mut self) -> &mut TokenIdAlloc {
@@ -523,17 +523,17 @@ impl SynToken {
     }
 }
 
-impl SrcToken for SynToken {
-    fn kind(&self) -> SyntaxKind {
+impl<'a> SrcToken<Convertor<'a>> for SynToken {
+    fn kind(&self, _ctx: &Convertor<'a>) -> SyntaxKind {
         self.token().kind()
     }
-    fn to_char(&self) -> Option<char> {
+    fn to_char(&self, _ctx: &Convertor<'a>) -> Option<char> {
         match self {
             SynToken::Ordinary(_) => None,
             SynToken::Punch(it, i) => it.text().chars().nth((*i).into()),
         }
     }
-    fn to_text(&self) -> SmolStr {
+    fn to_text(&self, _ctx: &Convertor<'a>) -> SmolStr {
         self.token().text().into()
     }
 }

crates/mbe/src/to_parser_tokens.rs

@@ -1,7 +1,7 @@
 //! Convert macro-by-example tokens which are specific to macro expansion into a
 //! format that works for our parser.
 
-use syntax::{lex_single_syntax_kind, SyntaxKind, SyntaxKind::*, T};
+use syntax::{SyntaxKind, SyntaxKind::*, T};
 use tt::buffer::TokenBuffer;
 
 pub(crate) fn to_parser_tokens(buffer: &TokenBuffer) -> parser::Tokens {
@@ -35,7 +35,7 @@ pub(crate) fn to_parser_tokens(buffer: &TokenBuffer) -> parser::Tokens {
                         let is_negated = lit.text.starts_with('-');
                         let inner_text = &lit.text[if is_negated { 1 } else { 0 }..];
 
-                        let kind = lex_single_syntax_kind(inner_text)
+                        let kind = parser::LexedStr::single_token(inner_text)
                             .map(|(kind, _error)| kind)
                             .filter(|kind| {
                                 kind.is_literal()

crates/parser/Cargo.toml

@@ -11,5 +11,8 @@ doctest = false
 
 [dependencies]
 drop_bomb = "0.1.4"
-
+rustc_lexer = { version = "725.0.0", package = "rustc-ap-rustc_lexer" }
 limit = { path = "../limit", version = "0.0.0" }
+
+[dev-dependencies]
+expect-test = "1.2"

crates/parser/src/lexed_str.rs

@@ -0,0 +1,300 @@
+//! Lexing `&str` into a sequence of Rust tokens.
+//!
+//! Note that strictly speaking the parser in this crate is not required to work
+//! on tokens which originated from text. Macros, eg, can synthesize tokens out
+//! of thin air. So, ideally, lexer should be an orthogonal crate. It is however
+//! convenient to include a text-based lexer here!
+//!
+//! Note that these tokens, unlike the tokens we feed into the parser, do
+//! include info about comments and whitespace.
+
+use std::ops;
+
+use crate::{
+    SyntaxKind::{self, *},
+    T,
+};
+
+pub struct LexedStr<'a> {
+    text: &'a str,
+    kind: Vec<SyntaxKind>,
+    start: Vec<u32>,
+    error: Vec<LexError>,
+}
+
+struct LexError {
+    msg: String,
+    token: u32,
+}
+
+impl<'a> LexedStr<'a> {
+    pub fn new(text: &'a str) -> LexedStr<'a> {
+        let mut res = LexedStr { text, kind: Vec::new(), start: Vec::new(), error: Vec::new() };
+
+        let mut offset = 0;
+        if let Some(shebang_len) = rustc_lexer::strip_shebang(text) {
+            res.push(SHEBANG, offset);
+            offset = shebang_len
+        };
+        for token in rustc_lexer::tokenize(&text[offset..]) {
+            let token_text = &text[offset..][..token.len];
+
+            let (kind, err) = from_rustc(&token.kind, token_text);
+            res.push(kind, offset);
+            offset += token.len;
+
+            if let Some(err) = err {
+                let token = res.len() as u32;
+                let msg = err.to_string();
+                res.error.push(LexError { msg, token });
+            }
+        }
+        res.push(EOF, offset);
+
+        res
+    }
+
+    pub fn single_token(text: &'a str) -> Option<(SyntaxKind, Option<String>)> {
+        if text.is_empty() {
+            return None;
+        }
+
+        let token = rustc_lexer::first_token(text);
+        if token.len != text.len() {
+            return None;
+        }
+
+        let (kind, err) = from_rustc(&token.kind, text);
+        Some((kind, err.map(|it| it.to_owned())))
+    }
+
+    pub fn as_str(&self) -> &str {
+        self.text
+    }
+
+    pub fn len(&self) -> usize {
+        self.kind.len() - 1
+    }
+
+    pub fn is_empty(&self) -> bool {
+        self.len() == 0
+    }
+
+    pub fn kind(&self, i: usize) -> SyntaxKind {
+        assert!(i < self.len());
+        self.kind[i]
+    }
+
+    pub fn text(&self, i: usize) -> &str {
+        self.range_text(i..i + 1)
+    }
+    pub fn range_text(&self, r: ops::Range<usize>) -> &str {
+        assert!(r.start < r.end && r.end <= self.len());
+        let lo = self.start[r.start] as usize;
+        let hi = self.start[r.end] as usize;
+        &self.text[lo..hi]
+    }
+
+    // Naming is hard.
+    pub fn text_range(&self, i: usize) -> ops::Range<usize> {
+        assert!(i < self.len());
+        let lo = self.start[i] as usize;
+        let hi = self.start[i + 1] as usize;
+        lo..hi
+    }
+    pub fn text_start(&self, i: usize) -> usize {
+        assert!(i <= self.len());
+        self.start[i] as usize
+    }
+    pub fn text_len(&self, i: usize) -> usize {
+        assert!(i < self.len());
+        let r = self.text_range(i);
+        r.end - r.start
+    }
+
+    pub fn error(&self, i: usize) -> Option<&str> {
+        assert!(i < self.len());
+        let err = self.error.binary_search_by_key(&(i as u32), |i| i.token).ok()?;
+        Some(self.error[err].msg.as_str())
+    }
+
+    pub fn errors(&self) -> impl Iterator<Item = (usize, &str)> + '_ {
+        self.error.iter().map(|it| (it.token as usize, it.msg.as_str()))
+    }
+
+    pub fn to_tokens(&self) -> crate::Tokens {
+        let mut res = crate::Tokens::default();
+        let mut was_joint = false;
+        for i in 0..self.len() {
+            let kind = self.kind(i);
+            if kind.is_trivia() {
+                was_joint = false
+            } else {
+                if kind == SyntaxKind::IDENT {
+                    let token_text = self.text(i);
+                    let contextual_kw = SyntaxKind::from_contextual_keyword(token_text)
+                        .unwrap_or(SyntaxKind::IDENT);
+                    res.push_ident(contextual_kw);
+                } else {
+                    if was_joint {
+                        res.was_joint();
+                    }
+                    res.push(kind);
+                }
+                was_joint = true;
+            }
+        }
+        res
+    }
+
+    fn push(&mut self, kind: SyntaxKind, offset: usize) {
+        self.kind.push(kind);
+        self.start.push(offset as u32);
+    }
+}
+
+/// Returns `SyntaxKind` and an optional tokenize error message.
+fn from_rustc(
+    kind: &rustc_lexer::TokenKind,
+    token_text: &str,
+) -> (SyntaxKind, Option<&'static str>) {
+    // A note on an intended tradeoff:
+    // We drop some useful information here (see patterns with double dots `..`)
+    // Storing that info in `SyntaxKind` is not possible due to its layout requirements of
+    // being `u16` that come from `rowan::SyntaxKind`.
+    let mut err = "";
+
+    let syntax_kind = {
+        match kind {
+            rustc_lexer::TokenKind::LineComment { doc_style: _ } => COMMENT,
+            rustc_lexer::TokenKind::BlockComment { doc_style: _, terminated } => {
+                if !terminated {
+                    err = "Missing trailing `*/` symbols to terminate the block comment";
+                }
+                COMMENT
+            }
+
+            rustc_lexer::TokenKind::Whitespace => WHITESPACE,
+
+            rustc_lexer::TokenKind::Ident if token_text == "_" => UNDERSCORE,
+            rustc_lexer::TokenKind::Ident => SyntaxKind::from_keyword(token_text).unwrap_or(IDENT),
+
+            rustc_lexer::TokenKind::RawIdent => IDENT,
+            rustc_lexer::TokenKind::Literal { kind, .. } => return from_rustc_literal(kind),
+
+            rustc_lexer::TokenKind::Lifetime { starts_with_number } => {
+                if *starts_with_number {
+                    err = "Lifetime name cannot start with a number";
+                }
+                LIFETIME_IDENT
+            }
+
+            rustc_lexer::TokenKind::Semi => T![;],
+            rustc_lexer::TokenKind::Comma => T![,],
+            rustc_lexer::TokenKind::Dot => T![.],
+            rustc_lexer::TokenKind::OpenParen => T!['('],
+            rustc_lexer::TokenKind::CloseParen => T![')'],
+            rustc_lexer::TokenKind::OpenBrace => T!['{'],
+            rustc_lexer::TokenKind::CloseBrace => T!['}'],
+            rustc_lexer::TokenKind::OpenBracket => T!['['],
+            rustc_lexer::TokenKind::CloseBracket => T![']'],
+            rustc_lexer::TokenKind::At => T![@],
+            rustc_lexer::TokenKind::Pound => T![#],
+            rustc_lexer::TokenKind::Tilde => T![~],
+            rustc_lexer::TokenKind::Question => T![?],
+            rustc_lexer::TokenKind::Colon => T![:],
+            rustc_lexer::TokenKind::Dollar => T![$],
+            rustc_lexer::TokenKind::Eq => T![=],
+            rustc_lexer::TokenKind::Bang => T![!],
+            rustc_lexer::TokenKind::Lt => T![<],
+            rustc_lexer::TokenKind::Gt => T![>],
+            rustc_lexer::TokenKind::Minus => T![-],
+            rustc_lexer::TokenKind::And => T![&],
+            rustc_lexer::TokenKind::Or => T![|],
+            rustc_lexer::TokenKind::Plus => T![+],
+            rustc_lexer::TokenKind::Star => T![*],
+            rustc_lexer::TokenKind::Slash => T![/],
+            rustc_lexer::TokenKind::Caret => T![^],
+            rustc_lexer::TokenKind::Percent => T![%],
+            rustc_lexer::TokenKind::Unknown => ERROR,
+        }
+    };
+
+    let err = if err.is_empty() { None } else { Some(err) };
+    (syntax_kind, err)
+}
+
+fn from_rustc_literal(kind: &rustc_lexer::LiteralKind) -> (SyntaxKind, Option<&'static str>) {
+    let mut err = "";
+
+    let syntax_kind = match *kind {
+        rustc_lexer::LiteralKind::Int { empty_int, base: _ } => {
+            if empty_int {
+                err = "Missing digits after the integer base prefix";
+            }
+            INT_NUMBER
+        }
+        rustc_lexer::LiteralKind::Float { empty_exponent, base: _ } => {
+            if empty_exponent {
+                err = "Missing digits after the exponent symbol";
+            }
+            FLOAT_NUMBER
+        }
+        rustc_lexer::LiteralKind::Char { terminated } => {
+            if !terminated {
+                err = "Missing trailing `'` symbol to terminate the character literal";
+            }
+            CHAR
+        }
+        rustc_lexer::LiteralKind::Byte { terminated } => {
+            if !terminated {
+                err = "Missing trailing `'` symbol to terminate the byte literal";
+            }
+            BYTE
+        }
+        rustc_lexer::LiteralKind::Str { terminated } => {
+            if !terminated {
+                err = "Missing trailing `\"` symbol to terminate the string literal";
+            }
+            STRING
+        }
+        rustc_lexer::LiteralKind::ByteStr { terminated } => {
+            if !terminated {
+                err = "Missing trailing `\"` symbol to terminate the byte string literal";
+            }
+            BYTE_STRING
+        }
+        rustc_lexer::LiteralKind::RawStr { err: raw_str_err, .. } => {
+            if let Some(raw_str_err) = raw_str_err {
+                err = match raw_str_err {
+                    rustc_lexer::RawStrError::InvalidStarter { .. } => "Missing `\"` symbol after `#` symbols to begin the raw string literal",
+                    rustc_lexer::RawStrError::NoTerminator { expected, found, .. } => if expected == found {
+                        "Missing trailing `\"` to terminate the raw string literal"
+                    } else {
+                        "Missing trailing `\"` with `#` symbols to terminate the raw string literal"
+                    },
+                    rustc_lexer::RawStrError::TooManyDelimiters { .. } => "Too many `#` symbols: raw strings may be delimited by up to 65535 `#` symbols",
+                };
+            };
+            STRING
+        }
+        rustc_lexer::LiteralKind::RawByteStr { err: raw_str_err, .. } => {
+            if let Some(raw_str_err) = raw_str_err {
+                err = match raw_str_err {
+                    rustc_lexer::RawStrError::InvalidStarter { .. } => "Missing `\"` symbol after `#` symbols to begin the raw byte string literal",
+                    rustc_lexer::RawStrError::NoTerminator { expected, found, .. } => if expected == found {
+                        "Missing trailing `\"` to terminate the raw byte string literal"
+                    } else {
+                        "Missing trailing `\"` with `#` symbols to terminate the raw byte string literal"
+                    },
+                    rustc_lexer::RawStrError::TooManyDelimiters { .. } => "Too many `#` symbols: raw byte strings may be delimited by up to 65535 `#` symbols",
+                };
+            };
+
+            BYTE_STRING
+        }
+    };
+
+    let err = if err.is_empty() { None } else { Some(err) };
+    (syntax_kind, err)
+}

crates/parser/src/lib.rs

@@ -18,6 +18,7 @@
 //! [`Parser`]: crate::parser::Parser
 #![allow(rustdoc::private_intra_doc_links)]
 
+mod lexed_str;
 mod token_set;
 mod syntax_kind;
 mod event;
@@ -25,9 +26,12 @@ mod parser;
 mod grammar;
 mod tokens;
 
+#[cfg(test)]
+mod tests;
+
 pub(crate) use token_set::TokenSet;
 
-pub use crate::{syntax_kind::SyntaxKind, tokens::Tokens};
+pub use crate::{lexed_str::LexedStr, syntax_kind::SyntaxKind, tokens::Tokens};
 
 #[derive(Debug, Clone, PartialEq, Eq, Hash)]
 pub struct ParseError(pub Box<String>);

crates/parser/src/tests.rs

@@ -0,0 +1,71 @@
+use std::{
+    fmt::Write,
+    fs,
+    path::{Path, PathBuf},
+};
+
+use expect_test::expect_file;
+
+use crate::LexedStr;
+
+#[test]
+fn valid_lexes_input() {
+    for case in TestCase::list("lexer/ok") {
+        let actual = lex(&case.text);
+        expect_file![case.txt].assert_eq(&actual)
+    }
+}
+
+#[test]
+fn invalid_lexes_input() {
+    for case in TestCase::list("lexer/err") {
+        let actual = lex(&case.text);
+        expect_file![case.txt].assert_eq(&actual)
+    }
+}
+
+fn lex(text: &str) -> String {
+    let lexed = LexedStr::new(text);
+
+    let mut res = String::new();
+    for i in 0..lexed.len() {
+        let kind = lexed.kind(i);
+        let text = lexed.text(i);
+        let error = lexed.error(i);
+
+        let error = error.map(|err| format!(" error: {}", err)).unwrap_or_default();
+        writeln!(res, "{:?} {:?}{}", kind, text, error).unwrap();
+    }
+    res
+}
+
+#[derive(PartialEq, Eq, PartialOrd, Ord)]
+struct TestCase {
+    rs: PathBuf,
+    txt: PathBuf,
+    text: String,
+}
+
+impl TestCase {
+    fn list(path: &'static str) -> Vec<TestCase> {
+        let crate_root_dir = Path::new(env!("CARGO_MANIFEST_DIR"));
+        let test_data_dir = crate_root_dir.join("test_data");
+        let dir = test_data_dir.join(path);
+
+        let mut res = Vec::new();
+        let read_dir = fs::read_dir(&dir)
+            .unwrap_or_else(|err| panic!("can't `read_dir` {}: {}", dir.display(), err));
+        for file in read_dir {
+            let file = file.unwrap();
+            let path = file.path();
+            if path.extension().unwrap_or_default() == "rs" {
+                let rs = path;
+                let txt = rs.with_extension("txt");
+                let text = fs::read_to_string(&rs).unwrap();
+                res.push(TestCase { rs, txt, text });
+            }
+        }
+        res.sort();
+        res
+    }
+}

crates/parser/src/tokens.rs

@@ -1,7 +1,8 @@
 //! Input for the parser -- a sequence of tokens.
 //!
 //! As of now, parser doesn't have access to the *text* of the tokens, and makes
-//! decisions based solely on their classification.
+//! decisions based solely on their classification. Unlike `LexerToken`, the
+//! `Tokens` doesn't include whitespace and comments.
 
 use crate::SyntaxKind;
 

crates/parser/test_data/lexer/err/empty_exponent.txt

@@ -0,0 +1,48 @@
+FLOAT_NUMBER "0e" error: Missing digits after the exponent symbol
+WHITESPACE "\n"
+FLOAT_NUMBER "0E" error: Missing digits after the exponent symbol
+WHITESPACE "\n\n"
+FLOAT_NUMBER "42e+" error: Missing digits after the exponent symbol
+WHITESPACE "\n"
+FLOAT_NUMBER "42e-" error: Missing digits after the exponent symbol
+WHITESPACE "\n"
+FLOAT_NUMBER "42E+" error: Missing digits after the exponent symbol
+WHITESPACE "\n"
+FLOAT_NUMBER "42E-" error: Missing digits after the exponent symbol
+WHITESPACE "\n\n"
+INT_NUMBER "42"
+DOT "."
+IDENT "e"
+PLUS "+"
+WHITESPACE "\n"
+INT_NUMBER "42"
+DOT "."
+IDENT "e"
+MINUS "-"
+WHITESPACE "\n"
+INT_NUMBER "42"
+DOT "."
+IDENT "E"
+PLUS "+"
+WHITESPACE "\n"
+INT_NUMBER "42"
+DOT "."
+IDENT "E"
+MINUS "-"
+WHITESPACE "\n\n"
+FLOAT_NUMBER "42.2e+" error: Missing digits after the exponent symbol
+WHITESPACE "\n"
+FLOAT_NUMBER "42.2e-" error: Missing digits after the exponent symbol
+WHITESPACE "\n"
+FLOAT_NUMBER "42.2E+" error: Missing digits after the exponent symbol
+WHITESPACE "\n"
+FLOAT_NUMBER "42.2E-" error: Missing digits after the exponent symbol
+WHITESPACE "\n\n"
+FLOAT_NUMBER "42.2e+f32" error: Missing digits after the exponent symbol
+WHITESPACE "\n"
+FLOAT_NUMBER "42.2e-f32" error: Missing digits after the exponent symbol
+WHITESPACE "\n"
+FLOAT_NUMBER "42.2E+f32" error: Missing digits after the exponent symbol
+WHITESPACE "\n"
+FLOAT_NUMBER "42.2E-f32" error: Missing digits after the exponent symbol
+WHITESPACE "\n"

crates/parser/test_data/lexer/err/empty_int.txt

@@ -0,0 +1,26 @@
+INT_NUMBER "0b" error: Missing digits after the integer base prefix
+WHITESPACE "\n"
+INT_NUMBER "0o" error: Missing digits after the integer base prefix
+WHITESPACE "\n"
+INT_NUMBER "0x" error: Missing digits after the integer base prefix
+WHITESPACE "\n\n"
+INT_NUMBER "0b_" error: Missing digits after the integer base prefix
+WHITESPACE "\n"
+INT_NUMBER "0o_" error: Missing digits after the integer base prefix
+WHITESPACE "\n"
+INT_NUMBER "0x_" error: Missing digits after the integer base prefix
+WHITESPACE "\n\n"
+INT_NUMBER "0bnoDigit" error: Missing digits after the integer base prefix
+WHITESPACE "\n"
+INT_NUMBER "0onoDigit" error: Missing digits after the integer base prefix
+WHITESPACE "\n"
+INT_NUMBER "0xnoDigit" error: Missing digits after the integer base prefix
+WHITESPACE "\n\n"
+INT_NUMBER "0xG" error: Missing digits after the integer base prefix
+WHITESPACE "\n"
+INT_NUMBER "0xg" error: Missing digits after the integer base prefix
+WHITESPACE "\n\n"
+INT_NUMBER "0x_g" error: Missing digits after the integer base prefix
+WHITESPACE "\n"
+INT_NUMBER "0x_G" error: Missing digits after the integer base prefix
+WHITESPACE "\n"

crates/parser/test_data/lexer/err/lifetime_starts_with_a_number.txt

@@ -0,0 +1,4 @@
+LIFETIME_IDENT "'1" error: Lifetime name cannot start with a number
+WHITESPACE "\n"
+LIFETIME_IDENT "'1lifetime" error: Lifetime name cannot start with a number
+WHITESPACE "\n"

crates/parser/test_data/lexer/err/unclosed_block_comment_at_eof.txt

@@ -0,0 +1 @@
+COMMENT "/*" error: Missing trailing `*/` symbols to terminate the block comment

crates/parser/test_data/lexer/err/unclosed_block_comment_with_content.txt

@@ -0,0 +1 @@
+COMMENT "/* comment\n" error: Missing trailing `*/` symbols to terminate the block comment

crates/parser/test_data/lexer/err/unclosed_byte_at_eof.txt

@@ -0,0 +1 @@
+BYTE "b'" error: Missing trailing `'` symbol to terminate the byte literal

crates/parser/test_data/lexer/err/unclosed_byte_string_at_eof.txt

@@ -0,0 +1 @@
+BYTE_STRING "b\"" error: Missing trailing `"` symbol to terminate the byte string literal

crates/parser/test_data/lexer/err/unclosed_byte_string_with_ascii_escape.txt

@@ -0,0 +1 @@
+BYTE_STRING "b\"\\x7f" error: Missing trailing `"` symbol to terminate the byte string literal

crates/parser/test_data/lexer/err/unclosed_byte_string_with_ferris.txt

@@ -0,0 +1 @@
+BYTE_STRING "b\"🦀" error: Missing trailing `"` symbol to terminate the byte string literal

crates/parser/test_data/lexer/err/unclosed_byte_string_with_slash.txt

@@ -0,0 +1 @@
+BYTE_STRING "b\"\\" error: Missing trailing `"` symbol to terminate the byte string literal

crates/parser/test_data/lexer/err/unclosed_byte_string_with_slash_double_quote.txt

@@ -0,0 +1 @@
+BYTE_STRING "b\"\\\"" error: Missing trailing `"` symbol to terminate the byte string literal

crates/parser/test_data/lexer/err/unclosed_byte_string_with_slash_n.txt

@@ -0,0 +1 @@
+BYTE_STRING "b\"\\n" error: Missing trailing `"` symbol to terminate the byte string literal

crates/parser/test_data/lexer/err/unclosed_byte_string_with_space.txt

@@ -0,0 +1 @@
+BYTE_STRING "b\" " error: Missing trailing `"` symbol to terminate the byte string literal

crates/parser/test_data/lexer/err/unclosed_byte_string_with_unicode_escape.txt

@@ -0,0 +1 @@
+BYTE_STRING "b\"\\u{20AA}" error: Missing trailing `"` symbol to terminate the byte string literal

crates/parser/test_data/lexer/err/unclosed_byte_with_ascii_escape.txt

@@ -0,0 +1 @@
+BYTE "b'\\x7f" error: Missing trailing `'` symbol to terminate the byte literal

crates/parser/test_data/lexer/err/unclosed_byte_with_ferris.txt

@@ -0,0 +1 @@
+BYTE "b'🦀" error: Missing trailing `'` symbol to terminate the byte literal

crates/parser/test_data/lexer/err/unclosed_byte_with_slash.txt

@@ -0,0 +1 @@
+BYTE "b'\\" error: Missing trailing `'` symbol to terminate the byte literal

crates/parser/test_data/lexer/err/unclosed_byte_with_slash_n.txt

@@ -0,0 +1 @@
+BYTE "b'\\n" error: Missing trailing `'` symbol to terminate the byte literal

crates/parser/test_data/lexer/err/unclosed_byte_with_slash_single_quote.txt

@@ -0,0 +1 @@
+BYTE "b'\\'" error: Missing trailing `'` symbol to terminate the byte literal

crates/parser/test_data/lexer/err/unclosed_byte_with_space.txt

@@ -0,0 +1 @@
+BYTE "b' " error: Missing trailing `'` symbol to terminate the byte literal

crates/parser/test_data/lexer/err/unclosed_byte_with_unicode_escape.txt

@@ -0,0 +1 @@
+BYTE "b'\\u{20AA}" error: Missing trailing `'` symbol to terminate the byte literal

crates/parser/test_data/lexer/err/unclosed_char_at_eof.txt

@@ -0,0 +1 @@
+CHAR "'" error: Missing trailing `'` symbol to terminate the character literal

crates/parser/test_data/lexer/err/unclosed_char_with_ascii_escape.txt

@@ -0,0 +1 @@
+CHAR "'\\x7f" error: Missing trailing `'` symbol to terminate the character literal

crates/parser/test_data/lexer/err/unclosed_char_with_ferris.txt

@@ -0,0 +1 @@
+CHAR "'🦀" error: Missing trailing `'` symbol to terminate the character literal

crates/parser/test_data/lexer/err/unclosed_char_with_slash.txt

@@ -0,0 +1 @@
+CHAR "'\\" error: Missing trailing `'` symbol to terminate the character literal

crates/parser/test_data/lexer/err/unclosed_char_with_slash_n.txt

@@ -0,0 +1 @@
+CHAR "'\\n" error: Missing trailing `'` symbol to terminate the character literal

crates/parser/test_data/lexer/err/unclosed_char_with_slash_single_quote.txt

@@ -0,0 +1 @@
+CHAR "'\\'" error: Missing trailing `'` symbol to terminate the character literal

crates/parser/test_data/lexer/err/unclosed_char_with_space.txt

@@ -0,0 +1 @@
+CHAR "' " error: Missing trailing `'` symbol to terminate the character literal

crates/parser/test_data/lexer/err/unclosed_char_with_unicode_escape.txt

@@ -0,0 +1 @@
+CHAR "'\\u{20AA}" error: Missing trailing `'` symbol to terminate the character literal

crates/parser/test_data/lexer/err/unclosed_nested_block_comment_entirely.txt

@@ -0,0 +1 @@
+COMMENT "/* /* /*\n" error: Missing trailing `*/` symbols to terminate the block comment

crates/parser/test_data/lexer/err/unclosed_nested_block_comment_partially.txt

@@ -0,0 +1 @@
+COMMENT "/** /*! /* comment */ */\n" error: Missing trailing `*/` symbols to terminate the block comment

crates/parser/test_data/lexer/err/unclosed_raw_byte_string_at_eof.txt

@@ -0,0 +1 @@
+BYTE_STRING "br##\"" error: Missing trailing `"` with `#` symbols to terminate the raw byte string literal

crates/parser/test_data/lexer/err/unclosed_raw_byte_string_with_ascii_escape.txt

@@ -0,0 +1 @@
+BYTE_STRING "br##\"\\x7f" error: Missing trailing `"` with `#` symbols to terminate the raw byte string literal

crates/parser/test_data/lexer/err/unclosed_raw_byte_string_with_ferris.txt

@@ -0,0 +1 @@
+BYTE_STRING "br##\"🦀" error: Missing trailing `"` with `#` symbols to terminate the raw byte string literal

crates/parser/test_data/lexer/err/unclosed_raw_byte_string_with_slash.txt

@@ -0,0 +1 @@
+BYTE_STRING "br##\"\\" error: Missing trailing `"` with `#` symbols to terminate the raw byte string literal

crates/parser/test_data/lexer/err/unclosed_raw_byte_string_with_slash_n.txt

@@ -0,0 +1 @@
+BYTE_STRING "br##\"\\n" error: Missing trailing `"` with `#` symbols to terminate the raw byte string literal

crates/parser/test_data/lexer/err/unclosed_raw_byte_string_with_space.txt

@@ -0,0 +1 @@
+BYTE_STRING "br##\" " error: Missing trailing `"` with `#` symbols to terminate the raw byte string literal

crates/parser/test_data/lexer/err/unclosed_raw_byte_string_with_unicode_escape.txt

@@ -0,0 +1 @@
+BYTE_STRING "br##\"\\u{20AA}" error: Missing trailing `"` with `#` symbols to terminate the raw byte string literal

crates/parser/test_data/lexer/err/unclosed_raw_string_at_eof.txt

@@ -0,0 +1 @@
+STRING "r##\"" error: Missing trailing `"` with `#` symbols to terminate the raw string literal

crates/parser/test_data/lexer/err/unclosed_raw_string_with_ascii_escape.txt

@@ -0,0 +1 @@
+STRING "r##\"\\x7f" error: Missing trailing `"` with `#` symbols to terminate the raw string literal

crates/parser/test_data/lexer/err/unclosed_raw_string_with_ferris.txt

@@ -0,0 +1 @@
+STRING "r##\"🦀" error: Missing trailing `"` with `#` symbols to terminate the raw string literal

crates/parser/test_data/lexer/err/unclosed_raw_string_with_slash.txt

@@ -0,0 +1 @@
+STRING "r##\"\\" error: Missing trailing `"` with `#` symbols to terminate the raw string literal

crates/parser/test_data/lexer/err/unclosed_raw_string_with_slash_n.txt

@@ -0,0 +1 @@
+STRING "r##\"\\n" error: Missing trailing `"` with `#` symbols to terminate the raw string literal