Merge #2911

2911: Implement collecting errors while tokenizing r=matklad a=Veetaha

Now we are collecting errors from `rustc_lexer` and returning them in `ParsedToken { token, error }` and `ParsedTokens { tokens, errors }` structures **([UPD]: this is now simplified, see updates bellow)**.

The main changes are introduced in `ra_syntax/parsing/lexer.rs`. It now exposes the following functions and types:

```rust
pub fn tokenize(text: &str) -> ParsedTokens;
pub fn tokenize_append(text: &str, parsed_tokens_to_append_to: &mut ParsedTokens);
pub fn first_token(text: &str) -> Option<ParsedToken>; // allows any number of tokens in text
pub fn single_token(text: &str) -> Option<ParsedToken>; // allows only a single token in text

pub struct ParsedToken  { pub token: Token,       pub error: Option<SyntaxError> }
pub struct ParsedTokens { pub tokens: Vec<Token>, pub errors: Vec<SyntaxError>   }

pub enum TokenizeError { /* Simple enum which reflects rustc_lexer tokenization errors */ }
```
In the first commit I implemented it with iterators, but then decided that since this crate is ad hoc for `rust-analyzer` and we clearly see the places of its usage it would be better to simplify it to vectors.

This is currently WIP, because I want to add tests for error messages generated by the lexer.
I'd like to listen to you thoughts how to define these tests in `ra_syntax/test-data` dir.

Related issues: #223 

**[UPD]**

After the PR review the API was simplified:
```rust
pub fn tokenize(text: &str) -> (Vec<Token>, Vec<SyntaxError>);
// Both lex functions do not check for unescape errors
pub fn lex_single_syntax_kind(text: &str) -> Option<(SyntaxKind, Option<SyntaxError>)>;
pub fn lex_single_valid_syntax_kind(text: &str) -> Option<SyntaxKind>;

// This will be removed in the next PR in favour of simlifying `SyntaxError` to `(String, TextRange)`
pub enum TokenizeError { /* Simple enum which reflects rustc_lexer tokenization errors */ }

// this is private, but may be made public if such demand would exist in future (least privilege principle)
fn lex_first_token(text: &str) -> Option<(Token, Option<SyntaxError>)>;
```

Co-authored-by: Veetaha <[email protected]>

3 files changed, 283 insertions, 100 deletions

diff --git a/crates/ra_syntax/src/parsing/lexer.rs b/crates/ra_syntax/src/parsing/lexer.rs
index 6d839208d..f889e6a1d 100644
--- a/crates/ra_syntax/src/parsing/lexer.rs
+++ b/crates/ra_syntax/src/parsing/lexer.rs
@@ -1,8 +1,10 @@
-//! FIXME: write short doc here
+//! Lexer analyzes raw input string and produces lexemes (tokens).
+//! It is just a bridge to `rustc_lexer`.
 
 use crate::{
+    SyntaxError, SyntaxErrorKind,
     SyntaxKind::{self, *},
-    TextUnit,
+    TextRange, TextUnit,
 };
 
 /// A token of Rust source.
@@ -14,91 +16,261 @@ pub struct Token {
     pub len: TextUnit,
 }
 
-fn match_literal_kind(kind: rustc_lexer::LiteralKind) -> SyntaxKind {
-    match kind {
-        rustc_lexer::LiteralKind::Int { .. } => INT_NUMBER,
-        rustc_lexer::LiteralKind::Float { .. } => FLOAT_NUMBER,
-        rustc_lexer::LiteralKind::Char { .. } => CHAR,
-        rustc_lexer::LiteralKind::Byte { .. } => BYTE,
-        rustc_lexer::LiteralKind::Str { .. } => STRING,
-        rustc_lexer::LiteralKind::ByteStr { .. } => BYTE_STRING,
-        rustc_lexer::LiteralKind::RawStr { .. } => RAW_STRING,
-        rustc_lexer::LiteralKind::RawByteStr { .. } => RAW_BYTE_STRING,
+/// Break a string up into its component tokens.
+/// Beware that it checks for shebang first and its length contributes to resulting
+/// tokens offsets.
+pub fn tokenize(text: &str) -> (Vec<Token>, Vec<SyntaxError>) {
+    // non-empty string is a precondtion of `rustc_lexer::strip_shebang()`.
+    if text.is_empty() {
+        return Default::default();
+    }
+
+    let mut tokens = Vec::new();
+    let mut errors = Vec::new();
+
+    let mut offset: usize = rustc_lexer::strip_shebang(text)
+        .map(|shebang_len| {
+            tokens.push(Token { kind: SHEBANG, len: TextUnit::from_usize(shebang_len) });
+            shebang_len
+        })
+        .unwrap_or(0);
+
+    let text_without_shebang = &text[offset..];
+
+    for rustc_token in rustc_lexer::tokenize(text_without_shebang) {
+        let token_len = TextUnit::from_usize(rustc_token.len);
+        let token_range = TextRange::offset_len(TextUnit::from_usize(offset), token_len);
+
+        let (syntax_kind, error) =
+            rustc_token_kind_to_syntax_kind(&rustc_token.kind, &text[token_range]);
+
+        tokens.push(Token { kind: syntax_kind, len: token_len });
+
+        if let Some(error) = error {
+            errors.push(SyntaxError::new(SyntaxErrorKind::TokenizeError(error), token_range));
+        }
+
+        offset += rustc_token.len;
     }
+
+    (tokens, errors)
+}
+
+/// Returns `SyntaxKind` and `Option<SyntaxError>` of the first token
+/// encountered at the beginning of the string.
+///
+/// Returns `None` if the string contains zero *or two or more* tokens.
+/// The token is malformed if the returned error is not `None`.
+///
+/// Beware that unescape errors are not checked at tokenization time.
+pub fn lex_single_syntax_kind(text: &str) -> Option<(SyntaxKind, Option<SyntaxError>)> {
+    lex_first_token(text)
+        .filter(|(token, _)| token.len.to_usize() == text.len())
+        .map(|(token, error)| (token.kind, error))
+}
+
+/// The same as `lex_single_syntax_kind()` but returns only `SyntaxKind` and
+/// returns `None` if any tokenization error occured.
+///
+/// Beware that unescape errors are not checked at tokenization time.
+pub fn lex_single_valid_syntax_kind(text: &str) -> Option<SyntaxKind> {
+    lex_first_token(text)
+        .filter(|(token, error)| !error.is_some() && token.len.to_usize() == text.len())
+        .map(|(token, _error)| token.kind)
 }
 
-/// Break a string up into its component tokens
-pub fn tokenize(text: &str) -> Vec<Token> {
+/// Returns `SyntaxKind` and `Option<SyntaxError>` of the first token
+/// encountered at the beginning of the string.
+///
+/// Returns `None` if the string contains zero tokens or if the token was parsed
+/// with an error.
+/// The token is malformed if the returned error is not `None`.
+///
+/// Beware that unescape errors are not checked at tokenization time.
+fn lex_first_token(text: &str) -> Option<(Token, Option<SyntaxError>)> {
+    // non-empty string is a precondtion of `rustc_lexer::first_token()`.
     if text.is_empty() {
-        return vec![];
-    }
-    let mut text = text;
-    let mut acc = Vec::new();
-    if let Some(len) = rustc_lexer::strip_shebang(text) {
-        acc.push(Token { kind: SHEBANG, len: TextUnit::from_usize(len) });
-        text = &text[len..];
+        return None;
     }
-    while !text.is_empty() {
-        let rustc_token = rustc_lexer::first_token(text);
-        let kind = match rustc_token.kind {
-            rustc_lexer::TokenKind::LineComment => COMMENT,
-            rustc_lexer::TokenKind::BlockComment { .. } => COMMENT,
-            rustc_lexer::TokenKind::Whitespace => WHITESPACE,
-            rustc_lexer::TokenKind::Ident => {
-                let token_text = &text[..rustc_token.len];
+
+    let rustc_token = rustc_lexer::first_token(text);
+    let (syntax_kind, error) = rustc_token_kind_to_syntax_kind(&rustc_token.kind, text);
+
+    let token = Token { kind: syntax_kind, len: TextUnit::from_usize(rustc_token.len) };
+    let error = error.map(|error| {
+        SyntaxError::new(
+            SyntaxErrorKind::TokenizeError(error),
+            TextRange::from_to(TextUnit::from(0), TextUnit::of_str(text)),
+        )
+    });
+
+    Some((token, error))
+}
+
+// FIXME: simplify TokenizeError to `SyntaxError(String, TextRange)` as per @matklad advice:
+// https://github.com/rust-analyzer/rust-analyzer/pull/2911/files#r371175067
+
+/// Describes the values of `SyntaxErrorKind::TokenizeError` enum variant.
+/// It describes all the types of errors that may happen during the tokenization
+/// of Rust source.
+#[derive(Debug, Clone, PartialEq, Eq, Hash)]
+pub enum TokenizeError {
+    /// Base prefix was provided, but there were no digits
+    /// after it, e.g. `0x`, `0b`.
+    EmptyInt,
+    /// Float exponent lacks digits e.g. `12.34e+`, `12.3E+`, `12e-`, `1_E-`,
+    EmptyExponent,
+
+    /// Block comment lacks trailing delimiter `*/`
+    UnterminatedBlockComment,
+    /// Character literal lacks trailing delimiter `'`
+    UnterminatedChar,
+    /// Characterish byte literal lacks trailing delimiter `'`
+    UnterminatedByte,
+    /// String literal lacks trailing delimiter `"`
+    UnterminatedString,
+    /// Byte string literal lacks trailing delimiter `"`
+    UnterminatedByteString,
+    /// Raw literal lacks trailing delimiter e.g. `"##`
+    UnterminatedRawString,
+    /// Raw byte string literal lacks trailing delimiter e.g. `"##`
+    UnterminatedRawByteString,
+
+    /// Raw string lacks a quote after the pound characters e.g. `r###`
+    UnstartedRawString,
+    /// Raw byte string lacks a quote after the pound characters e.g. `br###`
+    UnstartedRawByteString,
+
+    /// Lifetime starts with a number e.g. `'4ever`
+    LifetimeStartsWithNumber,
+}
+
+fn rustc_token_kind_to_syntax_kind(
+    rustc_token_kind: &rustc_lexer::TokenKind,
+    token_text: &str,
+) -> (SyntaxKind, Option<TokenizeError>) {
+    // A note on an intended tradeoff:
+    // We drop some useful infromation 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 syntax_kind = {
+        use rustc_lexer::TokenKind as TK;
+        use TokenizeError as TE;
+
+        match rustc_token_kind {
+            TK::LineComment => COMMENT,
+
+            TK::BlockComment { terminated: true } => COMMENT,
+            TK::BlockComment { terminated: false } => {
+                return (COMMENT, Some(TE::UnterminatedBlockComment));
+            }
+
+            TK::Whitespace => WHITESPACE,
+
+            TK::Ident => {
                 if token_text == "_" {
                     UNDERSCORE
                 } else {
-                    SyntaxKind::from_keyword(&text[..rustc_token.len]).unwrap_or(IDENT)
+                    SyntaxKind::from_keyword(token_text).unwrap_or(IDENT)
                 }
             }
-            rustc_lexer::TokenKind::RawIdent => IDENT,
-            rustc_lexer::TokenKind::Literal { kind, .. } => match_literal_kind(kind),
-            rustc_lexer::TokenKind::Lifetime { .. } => LIFETIME,
-            rustc_lexer::TokenKind::Semi => SEMI,
-            rustc_lexer::TokenKind::Comma => COMMA,
-            rustc_lexer::TokenKind::Dot => DOT,
-            rustc_lexer::TokenKind::OpenParen => L_PAREN,
-            rustc_lexer::TokenKind::CloseParen => R_PAREN,
-            rustc_lexer::TokenKind::OpenBrace => L_CURLY,
-            rustc_lexer::TokenKind::CloseBrace => R_CURLY,
-            rustc_lexer::TokenKind::OpenBracket => L_BRACK,
-            rustc_lexer::TokenKind::CloseBracket => R_BRACK,
-            rustc_lexer::TokenKind::At => AT,
-            rustc_lexer::TokenKind::Pound => POUND,
-            rustc_lexer::TokenKind::Tilde => TILDE,
-            rustc_lexer::TokenKind::Question => QUESTION,
-            rustc_lexer::TokenKind::Colon => COLON,
-            rustc_lexer::TokenKind::Dollar => DOLLAR,
-            rustc_lexer::TokenKind::Eq => EQ,
-            rustc_lexer::TokenKind::Not => EXCL,
-            rustc_lexer::TokenKind::Lt => L_ANGLE,
-            rustc_lexer::TokenKind::Gt => R_ANGLE,
-            rustc_lexer::TokenKind::Minus => MINUS,
-            rustc_lexer::TokenKind::And => AMP,
-            rustc_lexer::TokenKind::Or => PIPE,
-            rustc_lexer::TokenKind::Plus => PLUS,
-            rustc_lexer::TokenKind::Star => STAR,
-            rustc_lexer::TokenKind::Slash => SLASH,
-            rustc_lexer::TokenKind::Caret => CARET,
-            rustc_lexer::TokenKind::Percent => PERCENT,
-            rustc_lexer::TokenKind::Unknown => ERROR,
+
+            TK::RawIdent => IDENT,
+            TK::Literal { kind, .. } => return match_literal_kind(&kind),
+
+            TK::Lifetime { starts_with_number: false } => LIFETIME,
+            TK::Lifetime { starts_with_number: true } => {
+                return (LIFETIME, Some(TE::LifetimeStartsWithNumber))
+            }
+
+            TK::Semi => SEMI,
+            TK::Comma => COMMA,
+            TK::Dot => DOT,
+            TK::OpenParen => L_PAREN,
+            TK::CloseParen => R_PAREN,
+            TK::OpenBrace => L_CURLY,
+            TK::CloseBrace => R_CURLY,
+            TK::OpenBracket => L_BRACK,
+            TK::CloseBracket => R_BRACK,
+            TK::At => AT,
+            TK::Pound => POUND,
+            TK::Tilde => TILDE,
+            TK::Question => QUESTION,
+            TK::Colon => COLON,
+            TK::Dollar => DOLLAR,
+            TK::Eq => EQ,
+            TK::Not => EXCL,
+            TK::Lt => L_ANGLE,
+            TK::Gt => R_ANGLE,
+            TK::Minus => MINUS,
+            TK::And => AMP,
+            TK::Or => PIPE,
+            TK::Plus => PLUS,
+            TK::Star => STAR,
+            TK::Slash => SLASH,
+            TK::Caret => CARET,
+            TK::Percent => PERCENT,
+            TK::Unknown => ERROR,
+        }
+    };
+
+    return (syntax_kind, None);
+
+    fn match_literal_kind(kind: &rustc_lexer::LiteralKind) -> (SyntaxKind, Option<TokenizeError>) {
+        use rustc_lexer::LiteralKind as LK;
+        use TokenizeError as TE;
+
+        #[rustfmt::skip]
+        let syntax_kind = match *kind {
+            LK::Int { empty_int: false, .. } => INT_NUMBER,
+            LK::Int { empty_int: true, .. } => {
+                return (INT_NUMBER, Some(TE::EmptyInt))
+            }
+
+            LK::Float { empty_exponent: false, .. } => FLOAT_NUMBER,
+            LK::Float { empty_exponent: true, .. } => {
+                return (FLOAT_NUMBER, Some(TE::EmptyExponent))
+            }
+
+            LK::Char { terminated: true } => CHAR,
+            LK::Char { terminated: false } => {
+                return (CHAR, Some(TE::UnterminatedChar))
+            }
+
+            LK::Byte { terminated: true } => BYTE,
+            LK::Byte { terminated: false } => {
+                return (BYTE, Some(TE::UnterminatedByte))
+            }
+
+            LK::Str { terminated: true } => STRING,
+            LK::Str { terminated: false } => {
+                return (STRING, Some(TE::UnterminatedString))
+            }
+
+
+            LK::ByteStr { terminated: true } => BYTE_STRING,
+            LK::ByteStr { terminated: false } => {
+                return (BYTE_STRING, Some(TE::UnterminatedByteString))
+            }
+
+            LK::RawStr { started: true, terminated: true, .. } => RAW_STRING,
+            LK::RawStr { started: true, terminated: false, .. } => {
+                return (RAW_STRING, Some(TE::UnterminatedRawString))
+            }
+            LK::RawStr { started: false, .. } => {
+                return (RAW_STRING, Some(TE::UnstartedRawString))
+            }
+
+            LK::RawByteStr { started: true, terminated: true, .. } => RAW_BYTE_STRING,
+            LK::RawByteStr { started: true, terminated: false, .. } => {
+                return (RAW_BYTE_STRING, Some(TE::UnterminatedRawByteString))
+            }
+            LK::RawByteStr { started: false, .. } => {
+                return (RAW_BYTE_STRING, Some(TE::UnstartedRawByteString))
+            }
         };
-        let token = Token { kind, len: TextUnit::from_usize(rustc_token.len) };
-        acc.push(token);
-        text = &text[rustc_token.len..];
-    }
-    acc
-}
 
-pub fn classify_literal(text: &str) -> Option<Token> {
-    let t = rustc_lexer::first_token(text);
-    if t.len != text.len() {
-        return None;
+        (syntax_kind, None)
     }
-    let kind = match t.kind {
-        rustc_lexer::TokenKind::Literal { kind, .. } => match_literal_kind(kind),
-        _ => return None,
-    };
-    Some(Token { kind, len: TextUnit::from_usize(t.len) })
 }
diff --git a/crates/ra_syntax/src/parsing/reparsing.rs b/crates/ra_syntax/src/parsing/reparsing.rs
index 06bdda11d..a86da0675 100644
--- a/crates/ra_syntax/src/parsing/reparsing.rs
+++ b/crates/ra_syntax/src/parsing/reparsing.rs
@@ -12,7 +12,7 @@ use ra_text_edit::AtomTextEdit;
 use crate::{
     algo,
     parsing::{
-        lexer::{tokenize, Token},
+        lexer::{lex_single_syntax_kind, tokenize, Token},
         text_token_source::TextTokenSource,
         text_tree_sink::TextTreeSink,
     },
@@ -41,37 +41,42 @@ fn reparse_token<'node>(
     root: &'node SyntaxNode,
     edit: &AtomTextEdit,
 ) -> Option<(GreenNode, TextRange)> {
-    let token = algo::find_covering_element(root, edit.delete).as_token()?.clone();
-    match token.kind() {
+    let prev_token = algo::find_covering_element(root, edit.delete).as_token()?.clone();
+    let prev_token_kind = prev_token.kind();
+    match prev_token_kind {
         WHITESPACE | COMMENT | IDENT | STRING | RAW_STRING => {
-            if token.kind() == WHITESPACE || token.kind() == COMMENT {
+            if prev_token_kind == WHITESPACE || prev_token_kind == COMMENT {
                 // removing a new line may extends previous token
-                if token.text().to_string()[edit.delete - token.text_range().start()].contains('\n')
-                {
+                let deleted_range = edit.delete - prev_token.text_range().start();
+                if prev_token.text()[deleted_range].contains('\n') {
                     return None;
                 }
             }
 
-            let text = get_text_after_edit(token.clone().into(), &edit);
-            let lex_tokens = tokenize(&text);
-            let lex_token = match lex_tokens[..] {
-                [lex_token] if lex_token.kind == token.kind() => lex_token,
-                _ => return None,
-            };
+            let mut new_text = get_text_after_edit(prev_token.clone().into(), &edit);
+            let (new_token_kind, _error) = lex_single_syntax_kind(&new_text)?;
 
-            if lex_token.kind == IDENT && is_contextual_kw(&text) {
+            if new_token_kind != prev_token_kind
+                || (new_token_kind == IDENT && is_contextual_kw(&new_text))
+            {
                 return None;
             }
 
-            if let Some(next_char) = root.text().char_at(token.text_range().end()) {
-                let tokens_with_next_char = tokenize(&format!("{}{}", text, next_char));
-                if tokens_with_next_char.len() == 1 {
+            // Check that edited token is not a part of the bigger token.
+            // E.g. if for source code `bruh"str"` the user removed `ruh`, then
+            // `b` no longer remains an identifier, but becomes a part of byte string literal
+            if let Some(next_char) = root.text().char_at(prev_token.text_range().end()) {
+                new_text.push(next_char);
+                let token_with_next_char = lex_single_syntax_kind(&new_text);
+                if let Some((_kind, _error)) = token_with_next_char {
                     return None;
                 }
+                new_text.pop();
             }
 
-            let new_token = GreenToken::new(rowan::SyntaxKind(token.kind().into()), text.into());
-            Some((token.replace_with(new_token), token.text_range()))
+            let new_token =
+                GreenToken::new(rowan::SyntaxKind(prev_token_kind.into()), new_text.into());
+            Some((prev_token.replace_with(new_token), prev_token.text_range()))
         }
         _ => None,
     }
@@ -83,20 +88,26 @@ fn reparse_block<'node>(
 ) -> Option<(GreenNode, Vec<SyntaxError>, TextRange)> {
     let (node, reparser) = find_reparsable_node(root, edit.delete)?;
     let text = get_text_after_edit(node.clone().into(), &edit);
-    let tokens = tokenize(&text);
+
+    let (tokens, new_lexer_errors) = tokenize(&text);
     if !is_balanced(&tokens) {
         return None;
     }
+
     let mut token_source = TextTokenSource::new(&text, &tokens);
     let mut tree_sink = TextTreeSink::new(&text, &tokens);
     reparser.parse(&mut token_source, &mut tree_sink);
-    let (green, new_errors) = tree_sink.finish();
-    Some((node.replace_with(green), new_errors, node.text_range()))
+
+    let (green, mut new_parser_errors) = tree_sink.finish();
+    new_parser_errors.extend(new_lexer_errors);
+
+    Some((node.replace_with(green), new_parser_errors, node.text_range()))
 }
 
 fn get_text_after_edit(element: SyntaxElement, edit: &AtomTextEdit) -> String {
     let edit =
         AtomTextEdit::replace(edit.delete - element.text_range().start(), edit.insert.clone());
+
     let text = match element {
         NodeOrToken::Token(token) => token.text().to_string(),
         NodeOrToken::Node(node) => node.text().to_string(),
@@ -113,6 +124,7 @@ fn is_contextual_kw(text: &str) -> bool {
 
 fn find_reparsable_node(node: &SyntaxNode, range: TextRange) -> Option<(SyntaxNode, Reparser)> {
     let node = algo::find_covering_element(node, range);
+
     let mut ancestors = match node {
         NodeOrToken::Token(it) => it.parent().ancestors(),
         NodeOrToken::Node(it) => it.ancestors(),
@@ -182,7 +194,6 @@ mod tests {
         let fully_reparsed = SourceFile::parse(&after);
         let incrementally_reparsed: Parse<SourceFile> = {
             let f = SourceFile::parse(&before);
-            let edit = AtomTextEdit { delete: range, insert: replace_with.to_string() };
             let (green, new_errors, range) =
                 incremental_reparse(f.tree().syntax(), &edit, f.errors.to_vec()).unwrap();
             assert_eq!(range.len(), reparsed_len.into(), "reparsed fragment has wrong length");
diff --git a/crates/ra_syntax/src/parsing/text_tree_sink.rs b/crates/ra_syntax/src/parsing/text_tree_sink.rs
index c36756d6c..dd202601d 100644
--- a/crates/ra_syntax/src/parsing/text_tree_sink.rs
+++ b/crates/ra_syntax/src/parsing/text_tree_sink.rs
@@ -92,8 +92,8 @@ impl<'a> TreeSink for TextTreeSink<'a> {
 }
 
 impl<'a> TextTreeSink<'a> {
-    pub(super) fn new(text: &'a str, tokens: &'a [Token]) -> TextTreeSink<'a> {
-        TextTreeSink {
+    pub(super) fn new(text: &'a str, tokens: &'a [Token]) -> Self {
+        Self {
             text,
             tokens,
             text_pos: 0.into(),