src/tools/rust-analyzer/crates/mbe/src/syntax_bridge.rs

   1 //! Conversions between [`SyntaxNode`] and [`tt::TokenTree`].
   2
   3 use rustc_hash::FxHashMap;
   4 use stdx::{always, non_empty_vec::NonEmptyVec};
   5 use syntax::{
   6     ast::{self, make::tokens::doc_comment},
   7     AstToken, Parse, PreorderWithTokens, SmolStr, SyntaxElement, SyntaxKind,
   8     SyntaxKind::*,
   9     SyntaxNode, SyntaxToken, SyntaxTreeBuilder, TextRange, TextSize, WalkEvent, T,
  10 };
  11
  12 use crate::{
  13     to_parser_input::to_parser_input,
  14     tt::{
  15         self,
  16         buffer::{Cursor, TokenBuffer},
  17     },
  18     tt_iter::TtIter,
  19     TokenMap,
  20 };
  21
  22 #[cfg(test)]
  23 mod tests;
  24
  25 /// Convert the syntax node to a `TokenTree` (what macro
  26 /// will consume).
  27 pub fn syntax_node_to_token_tree(node: &SyntaxNode) -> (tt::Subtree, TokenMap) {
  28     let (subtree, token_map, _) = syntax_node_to_token_tree_with_modifications(
  29         node,
  30         Default::default(),
  31         0,
  32         Default::default(),
  33         Default::default(),
  34     );
  35     (subtree, token_map)
  36 }
  37
  38 /// Convert the syntax node to a `TokenTree` (what macro will consume)
  39 /// with the censored range excluded.
  40 pub fn syntax_node_to_token_tree_with_modifications(
  41     node: &SyntaxNode,
  42     existing_token_map: TokenMap,
  43     next_id: u32,
  44     replace: FxHashMap<SyntaxElement, Vec<SyntheticToken>>,
  45     append: FxHashMap<SyntaxElement, Vec<SyntheticToken>>,
  46 ) -> (tt::Subtree, TokenMap, u32) {
  47     let global_offset = node.text_range().start();
  48     let mut c = Converter::new(node, global_offset, existing_token_map, next_id, replace, append);
  49     let subtree = convert_tokens(&mut c);
  50     c.id_alloc.map.shrink_to_fit();
  51     always!(c.replace.is_empty(), "replace: {:?}", c.replace);
  52     always!(c.append.is_empty(), "append: {:?}", c.append);
  53     (subtree, c.id_alloc.map, c.id_alloc.next_id)
  54 }
  55
  56 #[derive(Clone, Copy, Debug, PartialEq, Eq, Hash)]
  57 pub struct SyntheticTokenId(pub u32);
  58
  59 #[derive(Debug, Clone)]
  60 pub struct SyntheticToken {
  61     pub kind: SyntaxKind,
  62     pub text: SmolStr,
  63     pub range: TextRange,
  64     pub id: SyntheticTokenId,
  65 }
  66
  67 // The following items are what `rustc` macro can be parsed into :
  68 // link: https://github.com/rust-lang/rust/blob/9ebf47851a357faa4cd97f4b1dc7835f6376e639/src/libsyntax/ext/expand.rs#L141
  69 // * Expr(P<ast::Expr>)                     -> token_tree_to_expr
  70 // * Pat(P<ast::Pat>)                       -> token_tree_to_pat
  71 // * Ty(P<ast::Ty>)                         -> token_tree_to_ty
  72 // * Stmts(SmallVec<[ast::Stmt; 1]>)        -> token_tree_to_stmts
  73 // * Items(SmallVec<[P<ast::Item>; 1]>)     -> token_tree_to_items
  74 //
  75 // * TraitItems(SmallVec<[ast::TraitItem; 1]>)
  76 // * AssocItems(SmallVec<[ast::AssocItem; 1]>)
  77 // * ForeignItems(SmallVec<[ast::ForeignItem; 1]>
  78
  79 pub fn token_tree_to_syntax_node(
  80     tt: &tt::Subtree,
  81     entry_point: parser::TopEntryPoint,
  82 ) -> (Parse<SyntaxNode>, TokenMap) {
  83     let buffer = match tt {
  84         tt::Subtree {
  85             delimiter: tt::Delimiter { kind: tt::DelimiterKind::Invisible, .. },
  86             token_trees,
  87         } => TokenBuffer::from_tokens(token_trees.as_slice()),
  88         _ => TokenBuffer::from_subtree(tt),
  89     };
  90     let parser_input = to_parser_input(&buffer);
  91     let parser_output = entry_point.parse(&parser_input);
  92     let mut tree_sink = TtTreeSink::new(buffer.begin());
  93     for event in parser_output.iter() {
  94         match event {
  95             parser::Step::Token { kind, n_input_tokens: n_raw_tokens } => {
  96                 tree_sink.token(kind, n_raw_tokens)
  97             }
  98             parser::Step::FloatSplit { ends_in_dot: has_pseudo_dot } => {
  99                 tree_sink.float_split(has_pseudo_dot)
 100             }
 101             parser::Step::Enter { kind } => tree_sink.start_node(kind),
 102             parser::Step::Exit => tree_sink.finish_node(),
 103             parser::Step::Error { msg } => tree_sink.error(msg.to_string()),
 104         }
 105     }
 106     tree_sink.finish()
 107 }
 108
 109 /// Convert a string to a `TokenTree`
 110 pub fn parse_to_token_tree(text: &str) -> Option<(tt::Subtree, TokenMap)> {
 111     let lexed = parser::LexedStr::new(text);
 112     if lexed.errors().next().is_some() {
 113         return None;
 114     }
 115
 116     let mut conv = RawConverter {
 117         lexed,
 118         pos: 0,
 119         id_alloc: TokenIdAlloc {
 120             map: Default::default(),
 121             global_offset: TextSize::default(),
 122             next_id: 0,
 123         },
 124     };
 125
 126     let subtree = convert_tokens(&mut conv);
 127     Some((subtree, conv.id_alloc.map))
 128 }
 129
 130 /// Split token tree with separate expr: $($e:expr)SEP*
 131 pub fn parse_exprs_with_sep(tt: &tt::Subtree, sep: char) -> Vec<tt::Subtree> {
 132     if tt.token_trees.is_empty() {
 133         return Vec::new();
 134     }
 135
 136     let mut iter = TtIter::new(tt);
 137     let mut res = Vec::new();
 138
 139     while iter.peek_n(0).is_some() {
 140         let expanded = iter.expect_fragment(parser::PrefixEntryPoint::Expr);
 141
 142         res.push(match expanded.value {
 143             None => break,
 144             Some(tt @ tt::TokenTree::Leaf(_)) => {
 145                 tt::Subtree { delimiter: tt::Delimiter::unspecified(), token_trees: vec![tt] }
 146             }
 147             Some(tt::TokenTree::Subtree(tt)) => tt,
 148         });
 149
 150         let mut fork = iter.clone();
 151         if fork.expect_char(sep).is_err() {
 152             break;
 153         }
 154         iter = fork;
 155     }
 156
 157     if iter.peek_n(0).is_some() {
 158         res.push(tt::Subtree {
 159             delimiter: tt::Delimiter::unspecified(),
 160             token_trees: iter.cloned().collect(),
 161         });
 162     }
 163
 164     res
 165 }
 166
 167 fn convert_tokens<C: TokenConverter>(conv: &mut C) -> tt::Subtree {
 168     struct StackEntry {
 169         subtree: tt::Subtree,
 170         idx: usize,
 171         open_range: TextRange,
 172     }
 173
 174     let entry = StackEntry {
 175         subtree: tt::Subtree { delimiter: tt::Delimiter::unspecified(), token_trees: vec![] },
 176         // never used (delimiter is `None`)
 177         idx: !0,
 178         open_range: TextRange::empty(TextSize::of('.')),
 179     };
 180     let mut stack = NonEmptyVec::new(entry);
 181
 182     loop {
 183         let StackEntry { subtree, .. } = stack.last_mut();
 184         let result = &mut subtree.token_trees;
 185         let (token, range) = match conv.bump() {
 186             Some(it) => it,
 187             None => break,
 188         };
 189         let synth_id = token.synthetic_id(conv);
 190
 191         let kind = token.kind(conv);
 192         if kind == COMMENT {
 193             if let Some(tokens) = conv.convert_doc_comment(&token) {
 194                 // FIXME: There has to be a better way to do this
 195                 // Add the comments token id to the converted doc string
 196                 let id = conv.id_alloc().alloc(range, synth_id);
 197                 result.extend(tokens.into_iter().map(|mut tt| {
 198                     if let tt::TokenTree::Subtree(sub) = &mut tt {
 199                         if let Some(tt::TokenTree::Leaf(tt::Leaf::Literal(lit))) =
 200                             sub.token_trees.get_mut(2)
 201                         {
 202                             lit.span = id
 203                         }
 204                     }
 205                     tt
 206                 }));
 207             }
 208             continue;
 209         }
 210         let tt = if kind.is_punct() && kind != UNDERSCORE {
 211             if synth_id.is_none() {
 212                 assert_eq!(range.len(), TextSize::of('.'));
 213             }
 214
 215             let expected = match subtree.delimiter.kind {
 216                 tt::DelimiterKind::Parenthesis => Some(T![')']),
 217                 tt::DelimiterKind::Brace => Some(T!['}']),
 218                 tt::DelimiterKind::Bracket => Some(T![']']),
 219                 tt::DelimiterKind::Invisible => None,
 220             };
 221
 222             if let Some(expected) = expected {
 223                 if kind == expected {
 224                     if let Some(entry) = stack.pop() {
 225                         conv.id_alloc().close_delim(entry.idx, Some(range));
 226                         stack.last_mut().subtree.token_trees.push(entry.subtree.into());
 227                     }
 228                     continue;
 229                 }
 230             }
 231
 232             let delim = match kind {
 233                 T!['('] => Some(tt::DelimiterKind::Parenthesis),
 234                 T!['{'] => Some(tt::DelimiterKind::Brace),
 235                 T!['['] => Some(tt::DelimiterKind::Bracket),
 236                 _ => None,
 237             };
 238
 239             if let Some(kind) = delim {
 240                 let (id, idx) = conv.id_alloc().open_delim(range, synth_id);
 241                 let subtree = tt::Subtree {
 242                     delimiter: tt::Delimiter { open: id, close: tt::TokenId::UNSPECIFIED, kind },
 243                     token_trees: vec![],
 244                 };
 245                 stack.push(StackEntry { subtree, idx, open_range: range });
 246                 continue;
 247             }
 248
 249             let spacing = match conv.peek().map(|next| next.kind(conv)) {
 250                 Some(kind) if is_single_token_op(kind) => tt::Spacing::Joint,
 251                 _ => tt::Spacing::Alone,
 252             };
 253             let char = match token.to_char(conv) {
 254                 Some(c) => c,
 255                 None => {
 256                     panic!("Token from lexer must be single char: token = {token:#?}");
 257                 }
 258             };
 259             tt::Leaf::from(tt::Punct {
 260                 char,
 261                 spacing,
 262                 span: conv.id_alloc().alloc(range, synth_id),
 263             })
 264             .into()
 265         } else {
 266             macro_rules! make_leaf {
 267                 ($i:ident) => {
 268                     tt::$i {
 269                         span: conv.id_alloc().alloc(range, synth_id),
 270                         text: token.to_text(conv),
 271                     }
 272                     .into()
 273                 };
 274             }
 275             let leaf: tt::Leaf = match kind {
 276                 T![true] | T![false] => make_leaf!(Ident),
 277                 IDENT => make_leaf!(Ident),
 278                 UNDERSCORE => make_leaf!(Ident),
 279                 k if k.is_keyword() => make_leaf!(Ident),
 280                 k if k.is_literal() => make_leaf!(Literal),
 281                 LIFETIME_IDENT => {
 282                     let char_unit = TextSize::of('\'');
 283                     let r = TextRange::at(range.start(), char_unit);
 284                     let apostrophe = tt::Leaf::from(tt::Punct {
 285                         char: '\'',
 286                         spacing: tt::Spacing::Joint,
 287                         span: conv.id_alloc().alloc(r, synth_id),
 288                     });
 289                     result.push(apostrophe.into());
 290
 291                     let r = TextRange::at(range.start() + char_unit, range.len() - char_unit);
 292                     let ident = tt::Leaf::from(tt::Ident {
 293                         text: SmolStr::new(&token.to_text(conv)[1..]),
 294                         span: conv.id_alloc().alloc(r, synth_id),
 295                     });
 296                     result.push(ident.into());
 297                     continue;
 298                 }
 299                 _ => continue,
 300             };
 301
 302             leaf.into()
 303         };
 304         result.push(tt);
 305     }
 306
 307     // If we get here, we've consumed all input tokens.
 308     // We might have more than one subtree in the stack, if the delimiters are improperly balanced.
 309     // Merge them so we're left with one.
 310     while let Some(entry) = stack.pop() {
 311         let parent = stack.last_mut();
 312
 313         conv.id_alloc().close_delim(entry.idx, None);
 314         let leaf: tt::Leaf = tt::Punct {
 315             span: conv.id_alloc().alloc(entry.open_range, None),
 316             char: match entry.subtree.delimiter.kind {
 317                 tt::DelimiterKind::Parenthesis => '(',
 318                 tt::DelimiterKind::Brace => '{',
 319                 tt::DelimiterKind::Bracket => '[',
 320                 tt::DelimiterKind::Invisible => '$',
 321             },
 322             spacing: tt::Spacing::Alone,
 323         }
 324         .into();
 325         parent.subtree.token_trees.push(leaf.into());
 326         parent.subtree.token_trees.extend(entry.subtree.token_trees);
 327     }
 328
 329     let subtree = stack.into_last().subtree;
 330     if let [tt::TokenTree::Subtree(first)] = &*subtree.token_trees {
 331         first.clone()
 332     } else {
 333         subtree
 334     }
 335 }
 336
 337 fn is_single_token_op(kind: SyntaxKind) -> bool {
 338     matches!(
 339         kind,
 340         EQ | L_ANGLE
 341             | R_ANGLE
 342             | BANG
 343             | AMP
 344             | PIPE
 345             | TILDE
 346             | AT
 347             | DOT
 348             | COMMA
 349             | SEMICOLON
 350             | COLON
 351             | POUND
 352             | DOLLAR
 353             | QUESTION
 354             | PLUS
 355             | MINUS
 356             | STAR
 357             | SLASH
 358             | PERCENT
 359             | CARET
 360             // LIFETIME_IDENT will be split into a sequence of `'` (a single quote) and an
 361             // identifier.
 362             | LIFETIME_IDENT
 363     )
 364 }
 365
 366 /// Returns the textual content of a doc comment block as a quoted string
 367 /// That is, strips leading `///` (or `/**`, etc)
 368 /// and strips the ending `*/`
 369 /// And then quote the string, which is needed to convert to `tt::Literal`
 370 fn doc_comment_text(comment: &ast::Comment) -> SmolStr {
 371     let prefix_len = comment.prefix().len();
 372     let mut text = &comment.text()[prefix_len..];
 373
 374     // Remove ending "*/"
 375     if comment.kind().shape == ast::CommentShape::Block {
 376         text = &text[0..text.len() - 2];
 377     }
 378
 379     // Quote the string
 380     // Note that `tt::Literal` expect an escaped string
 381     let text = format!("\"{}\"", text.escape_debug());
 382     text.into()
 383 }
 384
 385 fn convert_doc_comment(token: &syntax::SyntaxToken) -> Option<Vec<tt::TokenTree>> {
 386     cov_mark::hit!(test_meta_doc_comments);
 387     let comment = ast::Comment::cast(token.clone())?;
 388     let doc = comment.kind().doc?;
 389
 390     // Make `doc="\" Comments\""
 391     let meta_tkns = vec![mk_ident("doc"), mk_punct('='), mk_doc_literal(&comment)];
 392
 393     // Make `#![]`
 394     let mut token_trees = Vec::with_capacity(3);
 395     token_trees.push(mk_punct('#'));
 396     if let ast::CommentPlacement::Inner = doc {
 397         token_trees.push(mk_punct('!'));
 398     }
 399     token_trees.push(tt::TokenTree::from(tt::Subtree {
 400         delimiter: tt::Delimiter {
 401             open: tt::TokenId::UNSPECIFIED,
 402             close: tt::TokenId::UNSPECIFIED,
 403             kind: tt::DelimiterKind::Bracket,
 404         },
 405         token_trees: meta_tkns,
 406     }));
 407
 408     return Some(token_trees);
 409
 410     // Helper functions
 411     fn mk_ident(s: &str) -> tt::TokenTree {
 412         tt::TokenTree::from(tt::Leaf::from(tt::Ident {
 413             text: s.into(),
 414             span: tt::TokenId::unspecified(),
 415         }))
 416     }
 417
 418     fn mk_punct(c: char) -> tt::TokenTree {
 419         tt::TokenTree::from(tt::Leaf::from(tt::Punct {
 420             char: c,
 421             spacing: tt::Spacing::Alone,
 422             span: tt::TokenId::unspecified(),
 423         }))
 424     }
 425
 426     fn mk_doc_literal(comment: &ast::Comment) -> tt::TokenTree {
 427         let lit = tt::Literal { text: doc_comment_text(comment), span: tt::TokenId::unspecified() };
 428
 429         tt::TokenTree::from(tt::Leaf::from(lit))
 430     }
 431 }
 432
 433 struct TokenIdAlloc {
 434     map: TokenMap,
 435     global_offset: TextSize,
 436     next_id: u32,
 437 }
 438
 439 impl TokenIdAlloc {
 440     fn alloc(
 441         &mut self,
 442         absolute_range: TextRange,
 443         synthetic_id: Option<SyntheticTokenId>,
 444     ) -> tt::TokenId {
 445         let relative_range = absolute_range - self.global_offset;
 446         let token_id = tt::TokenId(self.next_id);
 447         self.next_id += 1;
 448         self.map.insert(token_id, relative_range);
 449         if let Some(id) = synthetic_id {
 450             self.map.insert_synthetic(token_id, id);
 451         }
 452         token_id
 453     }
 454
 455     fn open_delim(
 456         &mut self,
 457         open_abs_range: TextRange,
 458         synthetic_id: Option<SyntheticTokenId>,
 459     ) -> (tt::TokenId, usize) {
 460         let token_id = tt::TokenId(self.next_id);
 461         self.next_id += 1;
 462         let idx = self.map.insert_delim(
 463             token_id,
 464             open_abs_range - self.global_offset,
 465             open_abs_range - self.global_offset,
 466         );
 467         if let Some(id) = synthetic_id {
 468             self.map.insert_synthetic(token_id, id);
 469         }
 470         (token_id, idx)
 471     }
 472
 473     fn close_delim(&mut self, idx: usize, close_abs_range: Option<TextRange>) {
 474         match close_abs_range {
 475             None => {
 476                 self.map.remove_delim(idx);
 477             }
 478             Some(close) => {
 479                 self.map.update_close_delim(idx, close - self.global_offset);
 480             }
 481         }
 482     }
 483 }
 484
 485 /// A raw token (straight from lexer) converter
 486 struct RawConverter<'a> {
 487     lexed: parser::LexedStr<'a>,
 488     pos: usize,
 489     id_alloc: TokenIdAlloc,
 490 }
 491
 492 trait SrcToken<Ctx>: std::fmt::Debug {
 493     fn kind(&self, ctx: &Ctx) -> SyntaxKind;
 494
 495     fn to_char(&self, ctx: &Ctx) -> Option<char>;
 496
 497     fn to_text(&self, ctx: &Ctx) -> SmolStr;
 498
 499     fn synthetic_id(&self, ctx: &Ctx) -> Option<SyntheticTokenId>;
 500 }
 501
 502 trait TokenConverter: Sized {
 503     type Token: SrcToken<Self>;
 504
 505     fn convert_doc_comment(&self, token: &Self::Token) -> Option<Vec<tt::TokenTree>>;
 506
 507     fn bump(&mut self) -> Option<(Self::Token, TextRange)>;
 508
 509     fn peek(&self) -> Option<Self::Token>;
 510
 511     fn id_alloc(&mut self) -> &mut TokenIdAlloc;
 512 }
 513
 514 impl<'a> SrcToken<RawConverter<'a>> for usize {
 515     fn kind(&self, ctx: &RawConverter<'a>) -> SyntaxKind {
 516         ctx.lexed.kind(*self)
 517     }
 518
 519     fn to_char(&self, ctx: &RawConverter<'a>) -> Option<char> {
 520         ctx.lexed.text(*self).chars().next()
 521     }
 522
 523     fn to_text(&self, ctx: &RawConverter<'_>) -> SmolStr {
 524         ctx.lexed.text(*self).into()
 525     }
 526
 527     fn synthetic_id(&self, _ctx: &RawConverter<'a>) -> Option<SyntheticTokenId> {
 528         None
 529     }
 530 }
 531
 532 impl<'a> TokenConverter for RawConverter<'a> {
 533     type Token = usize;
 534
 535     fn convert_doc_comment(&self, &token: &usize) -> Option<Vec<tt::TokenTree>> {
 536         let text = self.lexed.text(token);
 537         convert_doc_comment(&doc_comment(text))
 538     }
 539
 540     fn bump(&mut self) -> Option<(Self::Token, TextRange)> {
 541         if self.pos == self.lexed.len() {
 542             return None;
 543         }
 544         let token = self.pos;
 545         self.pos += 1;
 546         let range = self.lexed.text_range(token);
 547         let range = TextRange::new(range.start.try_into().unwrap(), range.end.try_into().unwrap());
 548
 549         Some((token, range))
 550     }
 551
 552     fn peek(&self) -> Option<Self::Token> {
 553         if self.pos == self.lexed.len() {
 554             return None;
 555         }
 556         Some(self.pos)
 557     }
 558
 559     fn id_alloc(&mut self) -> &mut TokenIdAlloc {
 560         &mut self.id_alloc
 561     }
 562 }
 563
 564 struct Converter {
 565     id_alloc: TokenIdAlloc,
 566     current: Option<SyntaxToken>,
 567     current_synthetic: Vec<SyntheticToken>,
 568     preorder: PreorderWithTokens,
 569     replace: FxHashMap<SyntaxElement, Vec<SyntheticToken>>,
 570     append: FxHashMap<SyntaxElement, Vec<SyntheticToken>>,
 571     range: TextRange,
 572     punct_offset: Option<(SyntaxToken, TextSize)>,
 573 }
 574
 575 impl Converter {
 576     fn new(
 577         node: &SyntaxNode,
 578         global_offset: TextSize,
 579         existing_token_map: TokenMap,
 580         next_id: u32,
 581         mut replace: FxHashMap<SyntaxElement, Vec<SyntheticToken>>,
 582         mut append: FxHashMap<SyntaxElement, Vec<SyntheticToken>>,
 583     ) -> Converter {
 584         let range = node.text_range();
 585         let mut preorder = node.preorder_with_tokens();
 586         let (first, synthetic) = Self::next_token(&mut preorder, &mut replace, &mut append);
 587         Converter {
 588             id_alloc: { TokenIdAlloc { map: existing_token_map, global_offset, next_id } },
 589             current: first,
 590             current_synthetic: synthetic,
 591             preorder,
 592             range,
 593             replace,
 594             append,
 595             punct_offset: None,
 596         }
 597     }
 598
 599     fn next_token(
 600         preorder: &mut PreorderWithTokens,
 601         replace: &mut FxHashMap<SyntaxElement, Vec<SyntheticToken>>,
 602         append: &mut FxHashMap<SyntaxElement, Vec<SyntheticToken>>,
 603     ) -> (Option<SyntaxToken>, Vec<SyntheticToken>) {
 604         while let Some(ev) = preorder.next() {
 605             let ele = match ev {
 606                 WalkEvent::Enter(ele) => ele,
 607                 WalkEvent::Leave(ele) => {
 608                     if let Some(mut v) = append.remove(&ele) {
 609                         if !v.is_empty() {
 610                             v.reverse();
 611                             return (None, v);
 612                         }
 613                     }
 614                     continue;
 615                 }
 616             };
 617             if let Some(mut v) = replace.remove(&ele) {
 618                 preorder.skip_subtree();
 619                 if !v.is_empty() {
 620                     v.reverse();
 621                     return (None, v);
 622                 }
 623             }
 624             match ele {
 625                 SyntaxElement::Token(t) => return (Some(t), Vec::new()),
 626                 _ => {}
 627             }
 628         }
 629         (None, Vec::new())
 630     }
 631 }
 632
 633 #[derive(Debug)]
 634 enum SynToken {
 635     Ordinary(SyntaxToken),
 636     // FIXME is this supposed to be `Punct`?
 637     Punch(SyntaxToken, TextSize),
 638     Synthetic(SyntheticToken),
 639 }
 640
 641 impl SynToken {
 642     fn token(&self) -> Option<&SyntaxToken> {
 643         match self {
 644             SynToken::Ordinary(it) | SynToken::Punch(it, _) => Some(it),
 645             SynToken::Synthetic(_) => None,
 646         }
 647     }
 648 }
 649
 650 impl SrcToken<Converter> for SynToken {
 651     fn kind(&self, ctx: &Converter) -> SyntaxKind {
 652         match self {
 653             SynToken::Ordinary(token) => token.kind(),
 654             SynToken::Punch(..) => SyntaxKind::from_char(self.to_char(ctx).unwrap()).unwrap(),
 655             SynToken::Synthetic(token) => token.kind,
 656         }
 657     }
 658     fn to_char(&self, _ctx: &Converter) -> Option<char> {
 659         match self {
 660             SynToken::Ordinary(_) => None,
 661             SynToken::Punch(it, i) => it.text().chars().nth((*i).into()),
 662             SynToken::Synthetic(token) if token.text.len() == 1 => token.text.chars().next(),
 663             SynToken::Synthetic(_) => None,
 664         }
 665     }
 666     fn to_text(&self, _ctx: &Converter) -> SmolStr {
 667         match self {
 668             SynToken::Ordinary(token) => token.text().into(),
 669             SynToken::Punch(token, _) => token.text().into(),
 670             SynToken::Synthetic(token) => token.text.clone(),
 671         }
 672     }
 673
 674     fn synthetic_id(&self, _ctx: &Converter) -> Option<SyntheticTokenId> {
 675         match self {
 676             SynToken::Synthetic(token) => Some(token.id),
 677             _ => None,
 678         }
 679     }
 680 }
 681
 682 impl TokenConverter for Converter {
 683     type Token = SynToken;
 684     fn convert_doc_comment(&self, token: &Self::Token) -> Option<Vec<tt::TokenTree>> {
 685         convert_doc_comment(token.token()?)
 686     }
 687
 688     fn bump(&mut self) -> Option<(Self::Token, TextRange)> {
 689         if let Some((punct, offset)) = self.punct_offset.clone() {
 690             if usize::from(offset) + 1 < punct.text().len() {
 691                 let offset = offset + TextSize::of('.');
 692                 let range = punct.text_range();
 693                 self.punct_offset = Some((punct.clone(), offset));
 694                 let range = TextRange::at(range.start() + offset, TextSize::of('.'));
 695                 return Some((SynToken::Punch(punct, offset), range));
 696             }
 697         }
 698
 699         if let Some(synth_token) = self.current_synthetic.pop() {
 700             if self.current_synthetic.is_empty() {
 701                 let (new_current, new_synth) =
 702                     Self::next_token(&mut self.preorder, &mut self.replace, &mut self.append);
 703                 self.current = new_current;
 704                 self.current_synthetic = new_synth;
 705             }
 706             let range = synth_token.range;
 707             return Some((SynToken::Synthetic(synth_token), range));
 708         }
 709
 710         let curr = self.current.clone()?;
 711         if !self.range.contains_range(curr.text_range()) {
 712             return None;
 713         }
 714         let (new_current, new_synth) =
 715             Self::next_token(&mut self.preorder, &mut self.replace, &mut self.append);
 716         self.current = new_current;
 717         self.current_synthetic = new_synth;
 718         let token = if curr.kind().is_punct() {
 719             self.punct_offset = Some((curr.clone(), 0.into()));
 720             let range = curr.text_range();
 721             let range = TextRange::at(range.start(), TextSize::of('.'));
 722             (SynToken::Punch(curr, 0.into()), range)
 723         } else {
 724             self.punct_offset = None;
 725             let range = curr.text_range();
 726             (SynToken::Ordinary(curr), range)
 727         };
 728
 729         Some(token)
 730     }
 731
 732     fn peek(&self) -> Option<Self::Token> {
 733         if let Some((punct, mut offset)) = self.punct_offset.clone() {
 734             offset += TextSize::of('.');
 735             if usize::from(offset) < punct.text().len() {
 736                 return Some(SynToken::Punch(punct, offset));
 737             }
 738         }
 739
 740         if let Some(synth_token) = self.current_synthetic.last() {
 741             return Some(SynToken::Synthetic(synth_token.clone()));
 742         }
 743
 744         let curr = self.current.clone()?;
 745         if !self.range.contains_range(curr.text_range()) {
 746             return None;
 747         }
 748
 749         let token = if curr.kind().is_punct() {
 750             SynToken::Punch(curr, 0.into())
 751         } else {
 752             SynToken::Ordinary(curr)
 753         };
 754         Some(token)
 755     }
 756
 757     fn id_alloc(&mut self) -> &mut TokenIdAlloc {
 758         &mut self.id_alloc
 759     }
 760 }
 761
 762 struct TtTreeSink<'a> {
 763     buf: String,
 764     cursor: Cursor<'a>,
 765     open_delims: FxHashMap<tt::TokenId, TextSize>,
 766     text_pos: TextSize,
 767     inner: SyntaxTreeBuilder,
 768     token_map: TokenMap,
 769 }
 770
 771 impl<'a> TtTreeSink<'a> {
 772     fn new(cursor: Cursor<'a>) -> Self {
 773         TtTreeSink {
 774             buf: String::new(),
 775             cursor,
 776             open_delims: FxHashMap::default(),
 777             text_pos: 0.into(),
 778             inner: SyntaxTreeBuilder::default(),
 779             token_map: TokenMap::default(),
 780         }
 781     }
 782
 783     fn finish(mut self) -> (Parse<SyntaxNode>, TokenMap) {
 784         self.token_map.shrink_to_fit();
 785         (self.inner.finish(), self.token_map)
 786     }
 787 }
 788
 789 fn delim_to_str(d: tt::DelimiterKind, closing: bool) -> Option<&'static str> {
 790     let texts = match d {
 791         tt::DelimiterKind::Parenthesis => "()",
 792         tt::DelimiterKind::Brace => "{}",
 793         tt::DelimiterKind::Bracket => "[]",
 794         tt::DelimiterKind::Invisible => return None,
 795     };
 796
 797     let idx = closing as usize;
 798     Some(&texts[idx..texts.len() - (1 - idx)])
 799 }
 800
 801 impl<'a> TtTreeSink<'a> {
 802     /// Parses a float literal as if it was a one to two name ref nodes with a dot inbetween.
 803     /// This occurs when a float literal is used as a field access.
 804     fn float_split(&mut self, has_pseudo_dot: bool) {
 805         let (text, _span) = match self.cursor.token_tree() {
 806             Some(tt::buffer::TokenTreeRef::Leaf(tt::Leaf::Literal(lit), _)) => {
 807                 (lit.text.as_str(), lit.span)
 808             }
 809             _ => unreachable!(),
 810         };
 811         match text.split_once('.') {
 812             Some((left, right)) => {
 813                 assert!(!left.is_empty());
 814                 self.inner.start_node(SyntaxKind::NAME_REF);
 815                 self.inner.token(SyntaxKind::INT_NUMBER, left);
 816                 self.inner.finish_node();
 817
 818                 // here we move the exit up, the original exit has been deleted in process
 819                 self.inner.finish_node();
 820
 821                 self.inner.token(SyntaxKind::DOT, ".");
 822
 823                 if has_pseudo_dot {
 824                     assert!(right.is_empty(), "{left}.{right}");
 825                 } else {
 826                     self.inner.start_node(SyntaxKind::NAME_REF);
 827                     self.inner.token(SyntaxKind::INT_NUMBER, right);
 828                     self.inner.finish_node();
 829
 830                     // the parser creates an unbalanced start node, we are required to close it here
 831                     self.inner.finish_node();
 832                 }
 833             }
 834             None => unreachable!(),
 835         }
 836         self.cursor = self.cursor.bump();
 837     }
 838
 839     fn token(&mut self, kind: SyntaxKind, mut n_tokens: u8) {
 840         if kind == LIFETIME_IDENT {
 841             n_tokens = 2;
 842         }
 843
 844         let mut last = self.cursor;
 845         for _ in 0..n_tokens {
 846             let tmp: u8;
 847             if self.cursor.eof() {
 848                 break;
 849             }
 850             last = self.cursor;
 851             let text: &str = loop {
 852                 break match self.cursor.token_tree() {
 853                     Some(tt::buffer::TokenTreeRef::Leaf(leaf, _)) => {
 854                         // Mark the range if needed
 855                         let (text, id) = match leaf {
 856                             tt::Leaf::Ident(ident) => (ident.text.as_str(), ident.span),
 857                             tt::Leaf::Punct(punct) => {
 858                                 assert!(punct.char.is_ascii());
 859                                 tmp = punct.char as u8;
 860                                 (
 861                                     std::str::from_utf8(std::slice::from_ref(&tmp)).unwrap(),
 862                                     punct.span,
 863                                 )
 864                             }
 865                             tt::Leaf::Literal(lit) => (lit.text.as_str(), lit.span),
 866                         };
 867                         let range = TextRange::at(self.text_pos, TextSize::of(text));
 868                         self.token_map.insert(id, range);
 869                         self.cursor = self.cursor.bump();
 870                         text
 871                     }
 872                     Some(tt::buffer::TokenTreeRef::Subtree(subtree, _)) => {
 873                         self.cursor = self.cursor.subtree().unwrap();
 874                         match delim_to_str(subtree.delimiter.kind, false) {
 875                             Some(it) => {
 876                                 self.open_delims.insert(subtree.delimiter.open, self.text_pos);
 877                                 it
 878                             }
 879                             None => continue,
 880                         }
 881                     }
 882                     None => {
 883                         let parent = self.cursor.end().unwrap();
 884                         self.cursor = self.cursor.bump();
 885                         match delim_to_str(parent.delimiter.kind, true) {
 886                             Some(it) => {
 887                                 if let Some(open_delim) =
 888                                     self.open_delims.get(&parent.delimiter.open)
 889                                 {
 890                                     let open_range = TextRange::at(*open_delim, TextSize::of('('));
 891                                     let close_range =
 892                                         TextRange::at(self.text_pos, TextSize::of('('));
 893                                     self.token_map.insert_delim(
 894                                         parent.delimiter.open,
 895                                         open_range,
 896                                         close_range,
 897                                     );
 898                                 }
 899                                 it
 900                             }
 901                             None => continue,
 902                         }
 903                     }
 904                 };
 905             };
 906             self.buf += text;
 907             self.text_pos += TextSize::of(text);
 908         }
 909
 910         self.inner.token(kind, self.buf.as_str());
 911         self.buf.clear();
 912         // Add whitespace between adjoint puncts
 913         let next = last.bump();
 914         if let (
 915             Some(tt::buffer::TokenTreeRef::Leaf(tt::Leaf::Punct(curr), _)),
 916             Some(tt::buffer::TokenTreeRef::Leaf(tt::Leaf::Punct(next), _)),
 917         ) = (last.token_tree(), next.token_tree())
 918         {
 919             // Note: We always assume the semi-colon would be the last token in
 920             // other parts of RA such that we don't add whitespace here.
 921             //
 922             // When `next` is a `Punct` of `'`, that's a part of a lifetime identifier so we don't
 923             // need to add whitespace either.
 924             if curr.spacing == tt::Spacing::Alone && curr.char != ';' && next.char != '\'' {
 925                 self.inner.token(WHITESPACE, " ");
 926                 self.text_pos += TextSize::of(' ');
 927             }
 928         }
 929     }
 930
 931     fn start_node(&mut self, kind: SyntaxKind) {
 932         self.inner.start_node(kind);
 933     }
 934
 935     fn finish_node(&mut self) {
 936         self.inner.finish_node();
 937     }
 938
 939     fn error(&mut self, error: String) {
 940         self.inner.error(error, self.text_pos)
 941     }
 942 }