diff --git a/crates/ra_mbe/src/lib.rs b/crates/ra_mbe/src/lib.rs index 4203929d4f4..a21ea4dbcbd 100644 --- a/crates/ra_mbe/src/lib.rs +++ b/crates/ra_mbe/src/lib.rs @@ -15,10 +15,13 @@ macro_rules! impl_froms { } } -mod tt_cursor; +// mod tt_cursor; mod mbe_parser; mod mbe_expander; mod syntax_bridge; +mod tt_cursor; +mod subtree_source; +mod subtree_parser; use ra_syntax::SmolStr; @@ -379,4 +382,54 @@ SOURCE_FILE@[0; 40) // [let] [s] [=] ["rust1"] [;] assert_eq!(to_literal(&stm_tokens[15 + 3]).text, "\"rust1\""); } + + #[test] + fn test_two_idents() { + let rules = create_rules( + r#" + macro_rules! foo { + ($ i:ident, $ j:ident) => { + fn foo() { let a = $ i; let b = $j; } + } + } +"#, + ); + assert_expansion(&rules, "foo! { foo, bar }", "fn foo () {let a = foo ; let b = bar ;}"); + } + + // The following tests are port from intellij-rust directly + // https://github.com/intellij-rust/intellij-rust/blob/c4e9feee4ad46e7953b1948c112533360b6087bb/src/test/kotlin/org/rust/lang/core/macros/RsMacroExpansionTest.kt + + #[test] + fn test_path() { + let rules = create_rules( + r#" + macro_rules! foo { + ($ i:path) => { + fn foo() { let a = $ i; } + } + } +"#, + ); + assert_expansion(&rules, "foo! { foo }", "fn foo () {let a = foo ;}"); + assert_expansion( + &rules, + "foo! { bar::::baz:: }", + "fn foo () {let a = bar ::< u8 > ::baz ::< u8 > ;}", + ); + } + + #[test] + fn test_two_paths() { + let rules = create_rules( + r#" + macro_rules! foo { + ($ i:path, $ j:path) => { + fn foo() { let a = $ i; let b = $j; } + } + } +"#, + ); + assert_expansion(&rules, "foo! { foo, bar }", "fn foo () {let a = foo ; let b = bar ;}"); + } } diff --git a/crates/ra_mbe/src/mbe_expander.rs b/crates/ra_mbe/src/mbe_expander.rs index 2dafd68f624..ce41d722562 100644 --- a/crates/ra_mbe/src/mbe_expander.rs +++ b/crates/ra_mbe/src/mbe_expander.rs @@ -139,6 +139,11 @@ fn match_lhs(pattern: &crate::Subtree, input: &mut TtCursor) -> Result { + let path = + input.eat_path().ok_or(ExpandError::UnexpectedToken)?.clone(); + res.inner.insert(text.clone(), Binding::Simple(path.into())); + } _ => return Err(ExpandError::UnexpectedToken), } } diff --git a/crates/ra_mbe/src/subtree_parser.rs b/crates/ra_mbe/src/subtree_parser.rs new file mode 100644 index 00000000000..ce39a40bb54 --- /dev/null +++ b/crates/ra_mbe/src/subtree_parser.rs @@ -0,0 +1,61 @@ +use crate::subtree_source::SubtreeTokenSource; + +use ra_parser::{TokenSource, TreeSink}; +use ra_syntax::{SyntaxKind}; + +struct OffsetTokenSink { + token_pos: usize, +} + +impl TreeSink for OffsetTokenSink { + fn token(&mut self, _kind: SyntaxKind, n_tokens: u8) { + self.token_pos += n_tokens as usize; + } + fn start_node(&mut self, _kind: SyntaxKind) {} + fn finish_node(&mut self) {} + fn error(&mut self, _error: ra_parser::ParseError) {} +} + +pub(crate) struct Parser<'a> { + subtree: &'a tt::Subtree, + cur_pos: &'a mut usize, +} + +impl<'a> Parser<'a> { + pub fn new(cur_pos: &'a mut usize, subtree: &'a tt::Subtree) -> Parser<'a> { + Parser { cur_pos, subtree } + } + + pub fn parse_path(self) -> Option { + self.parse(ra_parser::parse_path) + } + + fn parse(self, f: F) -> Option + where + F: FnOnce(&dyn TokenSource, &mut dyn TreeSink), + { + let mut src = SubtreeTokenSource::new(self.subtree); + src.start_from_nth(*self.cur_pos); + let mut sink = OffsetTokenSink { token_pos: 0 }; + + f(&src, &mut sink); + + self.finish(sink.token_pos, &mut src) + } + + fn finish(self, parsed_token: usize, src: &mut SubtreeTokenSource) -> Option { + let res = src.bump_n(parsed_token); + *self.cur_pos += res.len(); + + let res: Vec<_> = res.into_iter().cloned().collect(); + + match res.len() { + 0 => None, + 1 => Some(res[0].clone()), + _ => Some(tt::TokenTree::Subtree(tt::Subtree { + delimiter: tt::Delimiter::None, + token_trees: res, + })), + } + } +} diff --git a/crates/ra_mbe/src/subtree_source.rs b/crates/ra_mbe/src/subtree_source.rs new file mode 100644 index 00000000000..4b37c2bdadd --- /dev/null +++ b/crates/ra_mbe/src/subtree_source.rs @@ -0,0 +1,521 @@ +use ra_parser::{TokenSource}; +use ra_syntax::{classify_literal, SmolStr, SyntaxKind, SyntaxKind::*}; +use std::cell::{RefCell}; + +#[derive(Debug, Clone, Eq, PartialEq)] +struct TtToken { + pub kind: SyntaxKind, + pub is_joint_to_next: bool, + pub text: SmolStr, + pub n_tokens: usize, +} + +#[derive(Debug, Clone, Eq, PartialEq)] +enum WalkCursor { + DelimiterBegin(Option), + Token(usize, Option), + DelimiterEnd(Option), + Eof, +} + +#[derive(Debug)] +struct SubTreeWalker<'a> { + pos: usize, + stack: Vec<(&'a tt::Subtree, Option)>, + cursor: WalkCursor, + last_steps: Vec, + subtree: &'a tt::Subtree, +} + +impl<'a> SubTreeWalker<'a> { + fn new(subtree: &tt::Subtree) -> SubTreeWalker { + let mut res = SubTreeWalker { + pos: 0, + stack: vec![], + cursor: WalkCursor::Eof, + last_steps: vec![], + subtree, + }; + + res.reset(); + res + } + + fn is_eof(&self) -> bool { + self.cursor == WalkCursor::Eof + } + + fn reset(&mut self) { + self.pos = 0; + self.stack = vec![(self.subtree, None)]; + self.cursor = WalkCursor::DelimiterBegin(convert_delim(self.subtree.delimiter, false)); + self.last_steps = vec![]; + + while self.is_empty_delimiter() { + self.forward_unchecked(); + } + } + + // This funciton will fast forward the cursor, + // Such that backward will stop at `start_pos` point + fn start_from_nth(&mut self, start_pos: usize) { + self.reset(); + self.pos = start_pos; + self.cursor = self.walk_token(start_pos, 0, false); + + while self.is_empty_delimiter() { + self.forward_unchecked(); + } + } + + fn current(&self) -> Option<&TtToken> { + match &self.cursor { + WalkCursor::DelimiterBegin(t) => t.as_ref(), + WalkCursor::Token(_, t) => t.as_ref(), + WalkCursor::DelimiterEnd(t) => t.as_ref(), + WalkCursor::Eof => None, + } + } + + fn is_empty_delimiter(&self) -> bool { + match &self.cursor { + WalkCursor::DelimiterBegin(None) => true, + WalkCursor::DelimiterEnd(None) => true, + _ => false, + } + } + + /// Move cursor backward by 1 step with empty checking + fn backward(&mut self) { + if self.last_steps.is_empty() { + return; + } + self.pos -= 1; + loop { + self.backward_unchecked(); + // Skip Empty delimiter + if self.last_steps.is_empty() || !self.is_empty_delimiter() { + break; + } + } + + // Move forward if it is empty delimiter + if self.last_steps.is_empty() { + while self.is_empty_delimiter() { + self.forward_unchecked(); + } + } + } + + /// Move cursor backward by 1 step without empty check + /// + /// Depends on the current state of cursor: + /// + /// * Delimiter Begin => Pop the stack, goto last walking token (`walk_token`) + /// * Token => Goto prev token (`walk_token`) + /// * Delimiter End => Goto the last child token (`walk_token`) + /// * Eof => push the root subtree, and set it as Delimiter End + fn backward_unchecked(&mut self) { + if self.last_steps.is_empty() { + return; + } + + let last_step = self.last_steps.pop().unwrap(); + let do_walk_token = match self.cursor { + WalkCursor::DelimiterBegin(_) => None, + WalkCursor::Token(u, _) => Some(u), + WalkCursor::DelimiterEnd(_) => { + let (top, _) = self.stack.last().unwrap(); + Some(top.token_trees.len()) + } + WalkCursor::Eof => None, + }; + + self.cursor = match do_walk_token { + Some(u) => self.walk_token(u, last_step, true), + None => match self.cursor { + WalkCursor::Eof => { + self.stack.push((self.subtree, None)); + WalkCursor::DelimiterEnd(convert_delim( + self.stack.last().unwrap().0.delimiter, + true, + )) + } + _ => { + let (_, last_top_cursor) = self.stack.pop().unwrap(); + assert!(!self.stack.is_empty()); + + self.walk_token(last_top_cursor.unwrap(), last_step, true) + } + }, + }; + } + + /// Move cursor forward by 1 step with empty checking + fn forward(&mut self) { + if self.is_eof() { + return; + } + + self.pos += 1; + loop { + self.forward_unchecked(); + if !self.is_empty_delimiter() { + break; + } + } + } + + /// Move cursor forward by 1 step without empty checking + /// + /// Depends on the current state of cursor: + /// + /// * Delimiter Begin => Goto the first child token (`walk_token`) + /// * Token => Goto next token (`walk_token`) + /// * Delimiter End => Pop the stack, goto last walking token (`walk_token`) + /// + fn forward_unchecked(&mut self) { + if self.is_eof() { + return; + } + + let step = self.current().map(|x| x.n_tokens).unwrap_or(1); + self.last_steps.push(step); + + let do_walk_token = match self.cursor { + WalkCursor::DelimiterBegin(_) => Some((0, 0)), + WalkCursor::Token(u, _) => Some((u, step)), + WalkCursor::DelimiterEnd(_) => None, + _ => unreachable!(), + }; + + self.cursor = match do_walk_token { + Some((u, step)) => self.walk_token(u, step, false), + None => { + let (_, last_top_idx) = self.stack.pop().unwrap(); + match self.stack.last() { + Some(_) => self.walk_token(last_top_idx.unwrap(), 1, false), + None => WalkCursor::Eof, + } + } + }; + } + + /// Traversal child token + /// Depends on the new position, it returns: + /// + /// * new position < 0 => DelimiterBegin + /// * new position > token_tree.len() => DelimiterEnd + /// * if new position is a subtree, depends on traversal direction: + /// ** backward => DelimiterEnd + /// ** forward => DelimiterBegin + /// * if new psoition is a leaf, return walk_leaf() + fn walk_token(&mut self, pos: usize, offset: usize, backward: bool) -> WalkCursor { + let (top, _) = self.stack.last().unwrap(); + + if backward && pos < offset { + return WalkCursor::DelimiterBegin(convert_delim( + self.stack.last().unwrap().0.delimiter, + false, + )); + } + + if !backward && pos + offset >= top.token_trees.len() { + return WalkCursor::DelimiterEnd(convert_delim( + self.stack.last().unwrap().0.delimiter, + true, + )); + } + + let pos = if backward { pos - offset } else { pos + offset }; + + match &top.token_trees[pos] { + tt::TokenTree::Subtree(subtree) => { + self.stack.push((subtree, Some(pos))); + let delim = convert_delim(self.stack.last().unwrap().0.delimiter, backward); + if backward { + WalkCursor::DelimiterEnd(delim) + } else { + WalkCursor::DelimiterBegin(delim) + } + } + tt::TokenTree::Leaf(leaf) => WalkCursor::Token(pos, Some(self.walk_leaf(leaf, pos))), + } + } + + fn walk_leaf(&mut self, leaf: &tt::Leaf, pos: usize) -> TtToken { + match leaf { + tt::Leaf::Literal(l) => convert_literal(l), + tt::Leaf::Ident(ident) => convert_ident(ident), + tt::Leaf::Punct(punct) => { + let (top, _) = self.stack.last().unwrap(); + convert_punct(punct, top, pos) + } + } + } +} + +pub(crate) trait Querier { + fn token(&self, uidx: usize) -> (SyntaxKind, SmolStr); +} + +// A wrapper class for ref cell +#[derive(Debug)] +pub(crate) struct WalkerOwner<'a> { + walker: RefCell>, + offset: usize, +} + +impl<'a> WalkerOwner<'a> { + fn new(subtree: &'a tt::Subtree) -> Self { + WalkerOwner { walker: RefCell::new(SubTreeWalker::new(subtree)), offset: 0 } + } + + fn get<'b>(&self, pos: usize) -> Option { + self.set_walker_pos(pos); + let walker = self.walker.borrow(); + walker.current().cloned() + } + + fn start_from_nth(&mut self, pos: usize) { + self.offset = pos; + self.walker.borrow_mut().start_from_nth(pos); + } + + fn set_walker_pos(&self, mut pos: usize) { + pos += self.offset; + let mut walker = self.walker.borrow_mut(); + while pos > walker.pos && !walker.is_eof() { + walker.forward(); + } + while pos < walker.pos { + walker.backward(); + } + } + + fn collect_token_trees(&mut self, n: usize) -> Vec<&tt::TokenTree> { + self.start_from_nth(self.offset); + + let mut res = vec![]; + let mut walker = self.walker.borrow_mut(); + + while walker.pos - self.offset < n { + if let WalkCursor::Token(u, tt) = &walker.cursor { + if walker.stack.len() == 1 { + // We only collect the topmost child + res.push(&walker.stack[0].0.token_trees[*u]); + if let Some(tt) = tt { + for i in 0..tt.n_tokens - 1 { + res.push(&walker.stack[0].0.token_trees[u + i]); + } + } + } + } + + walker.forward(); + } + + res + } +} + +impl<'a> Querier for WalkerOwner<'a> { + fn token(&self, uidx: usize) -> (SyntaxKind, SmolStr) { + let tkn = self.get(uidx).unwrap(); + (tkn.kind, tkn.text) + } +} + +pub(crate) struct SubtreeTokenSource<'a> { + walker: WalkerOwner<'a>, +} + +impl<'a> SubtreeTokenSource<'a> { + pub fn new(subtree: &tt::Subtree) -> SubtreeTokenSource { + SubtreeTokenSource { walker: WalkerOwner::new(subtree) } + } + + pub fn start_from_nth(&mut self, n: usize) { + self.walker.start_from_nth(n); + } + + pub fn querier<'b>(&'a self) -> &'b WalkerOwner<'a> + where + 'a: 'b, + { + &self.walker + } + + pub(crate) fn bump_n(&mut self, parsed_tokens: usize) -> Vec<&tt::TokenTree> { + let res = self.walker.collect_token_trees(parsed_tokens); + res + } +} + +impl<'a> TokenSource for SubtreeTokenSource<'a> { + fn token_kind(&self, pos: usize) -> SyntaxKind { + if let Some(tok) = self.walker.get(pos) { + tok.kind + } else { + SyntaxKind::EOF + } + } + fn is_token_joint_to_next(&self, pos: usize) -> bool { + self.walker.get(pos).unwrap().is_joint_to_next + } + fn is_keyword(&self, pos: usize, kw: &str) -> bool { + self.walker.get(pos).unwrap().text == *kw + } +} + +struct TokenPeek<'a, I> +where + I: Iterator, +{ + iter: itertools::MultiPeek, +} + +// helper function +fn to_punct(tt: &tt::TokenTree) -> Option<&tt::Punct> { + if let tt::TokenTree::Leaf(tt::Leaf::Punct(pp)) = tt { + return Some(pp); + } + None +} + +impl<'a, I> TokenPeek<'a, I> +where + I: Iterator, +{ + pub fn new(iter: I) -> Self { + TokenPeek { iter: itertools::multipeek(iter) } + } + + fn current_punct2(&mut self, p: &tt::Punct) -> Option<((char, char), bool)> { + if p.spacing != tt::Spacing::Joint { + return None; + } + + self.iter.reset_peek(); + let p1 = to_punct(self.iter.peek()?)?; + Some(((p.char, p1.char), p1.spacing == tt::Spacing::Joint)) + } + + fn current_punct3(&mut self, p: &tt::Punct) -> Option<((char, char, char), bool)> { + self.current_punct2(p).and_then(|((p0, p1), last_joint)| { + if !last_joint { + None + } else { + let p2 = to_punct(*self.iter.peek()?)?; + Some(((p0, p1, p2.char), p2.spacing == tt::Spacing::Joint)) + } + }) + } +} + +fn convert_multi_char_punct<'b, I>( + p: &tt::Punct, + iter: &mut TokenPeek<'b, I>, +) -> Option<(SyntaxKind, bool, &'static str, usize)> +where + I: Iterator, +{ + if let Some((m, is_joint_to_next)) = iter.current_punct3(p) { + if let Some((kind, text)) = match m { + ('<', '<', '=') => Some((SHLEQ, "<<=")), + ('>', '>', '=') => Some((SHREQ, ">>=")), + ('.', '.', '.') => Some((DOTDOTDOT, "...")), + ('.', '.', '=') => Some((DOTDOTEQ, "..=")), + _ => None, + } { + return Some((kind, is_joint_to_next, text, 3)); + } + } + + if let Some((m, is_joint_to_next)) = iter.current_punct2(p) { + if let Some((kind, text)) = match m { + ('<', '<') => Some((SHL, "<<")), + ('>', '>') => Some((SHR, ">>")), + + ('|', '|') => Some((PIPEPIPE, "||")), + ('&', '&') => Some((AMPAMP, "&&")), + ('%', '=') => Some((PERCENTEQ, "%=")), + ('*', '=') => Some((STAREQ, "*=")), + ('/', '=') => Some((SLASHEQ, "/=")), + ('^', '=') => Some((CARETEQ, "^=")), + + ('&', '=') => Some((AMPEQ, "&=")), + ('|', '=') => Some((PIPEEQ, "|=")), + ('-', '=') => Some((MINUSEQ, "-=")), + ('+', '=') => Some((PLUSEQ, "+=")), + ('>', '=') => Some((GTEQ, ">=")), + ('<', '=') => Some((LTEQ, "<=")), + + ('-', '>') => Some((THIN_ARROW, "->")), + ('!', '=') => Some((NEQ, "!=")), + ('=', '>') => Some((FAT_ARROW, "=>")), + ('=', '=') => Some((EQEQ, "==")), + ('.', '.') => Some((DOTDOT, "..")), + (':', ':') => Some((COLONCOLON, "::")), + + _ => None, + } { + return Some((kind, is_joint_to_next, text, 2)); + } + } + + None +} + +fn convert_delim(d: tt::Delimiter, closing: bool) -> Option { + let (kinds, texts) = match d { + tt::Delimiter::Parenthesis => ([L_PAREN, R_PAREN], "()"), + tt::Delimiter::Brace => ([L_CURLY, R_CURLY], "{}"), + tt::Delimiter::Bracket => ([L_BRACK, R_BRACK], "[]"), + tt::Delimiter::None => return None, + }; + + let idx = closing as usize; + let kind = kinds[idx]; + let text = &texts[idx..texts.len() - (1 - idx)]; + Some(TtToken { kind, is_joint_to_next: false, text: SmolStr::new(text), n_tokens: 1 }) +} + +fn convert_literal(l: &tt::Literal) -> TtToken { + TtToken { + kind: classify_literal(&l.text).unwrap().kind, + is_joint_to_next: false, + text: l.text.clone(), + n_tokens: 1, + } +} + +fn convert_ident(ident: &tt::Ident) -> TtToken { + let kind = SyntaxKind::from_keyword(ident.text.as_str()).unwrap_or(IDENT); + TtToken { kind, is_joint_to_next: false, text: ident.text.clone(), n_tokens: 1 } +} + +fn convert_punct(p: &tt::Punct, parent: &tt::Subtree, next: usize) -> TtToken { + let iter = parent.token_trees[next + 1..].iter(); + let mut peek = TokenPeek::new(iter); + + if let Some((kind, is_joint_to_next, text, size)) = convert_multi_char_punct(p, &mut peek) { + TtToken { kind, is_joint_to_next, text: text.into(), n_tokens: size } + } else { + let kind = match p.char { + // lexer may produce combpund tokens for these ones + '.' => DOT, + ':' => COLON, + '=' => EQ, + '!' => EXCL, + '-' => MINUS, + c => SyntaxKind::from_char(c).unwrap(), + }; + let text = { + let mut buf = [0u8; 4]; + let s: &str = p.char.encode_utf8(&mut buf); + SmolStr::new(s) + }; + TtToken { kind, is_joint_to_next: p.spacing == tt::Spacing::Joint, text, n_tokens: 1 } + } +} diff --git a/crates/ra_mbe/src/syntax_bridge.rs b/crates/ra_mbe/src/syntax_bridge.rs index 139a0fd3379..19c17bd550b 100644 --- a/crates/ra_mbe/src/syntax_bridge.rs +++ b/crates/ra_mbe/src/syntax_bridge.rs @@ -1,9 +1,11 @@ -use ra_parser::{TokenSource, TreeSink, ParseError}; +use ra_parser::{TreeSink, ParseError}; use ra_syntax::{ AstNode, SyntaxNode, TextRange, SyntaxKind, SmolStr, SyntaxTreeBuilder, TreeArc, SyntaxElement, - ast, SyntaxKind::*, TextUnit, classify_literal + ast, SyntaxKind::*, TextUnit }; +use crate::subtree_source::{SubtreeTokenSource, Querier}; + /// Maps `tt::TokenId` to the relative range of the original token. #[derive(Default)] pub struct TokenMap { @@ -22,8 +24,8 @@ pub fn ast_to_token_tree(ast: &ast::TokenTree) -> Option<(tt::Subtree, TokenMap) /// Parses the token tree (result of macro expansion) as a sequence of items pub fn token_tree_to_ast_item_list(tt: &tt::Subtree) -> TreeArc { - let token_source = TtTokenSource::new(tt); - let mut tree_sink = TtTreeSink::new(&token_source.tokens); + let token_source = SubtreeTokenSource::new(tt); + let mut tree_sink = TtTreeSink::new(token_source.querier()); ra_parser::parse(&token_source, &mut tree_sink); let syntax = tree_sink.inner.finish(); ast::SourceFile::cast(&syntax).unwrap().to_owned() @@ -103,229 +105,19 @@ fn convert_tt( Some(res) } -#[derive(Debug)] -struct TtTokenSource { - tokens: Vec, -} - -#[derive(Debug)] -struct TtToken { - kind: SyntaxKind, - is_joint_to_next: bool, - text: SmolStr, -} - -// Some helper functions -fn to_punct(tt: &tt::TokenTree) -> Option<&tt::Punct> { - if let tt::TokenTree::Leaf(tt::Leaf::Punct(pp)) = tt { - return Some(pp); - } - None -} - -struct TokenPeek<'a, I> -where - I: Iterator, -{ - iter: itertools::MultiPeek, -} - -impl<'a, I> TokenPeek<'a, I> -where - I: Iterator, -{ - fn next(&mut self) -> Option<&tt::TokenTree> { - self.iter.next() - } - - fn current_punct2(&mut self, p: &tt::Punct) -> Option<((char, char), bool)> { - if p.spacing != tt::Spacing::Joint { - return None; - } - - self.iter.reset_peek(); - let p1 = to_punct(self.iter.peek()?)?; - Some(((p.char, p1.char), p1.spacing == tt::Spacing::Joint)) - } - - fn current_punct3(&mut self, p: &tt::Punct) -> Option<((char, char, char), bool)> { - self.current_punct2(p).and_then(|((p0, p1), last_joint)| { - if !last_joint { - None - } else { - let p2 = to_punct(*self.iter.peek()?)?; - Some(((p0, p1, p2.char), p2.spacing == tt::Spacing::Joint)) - } - }) - } -} - -impl TtTokenSource { - fn new(tt: &tt::Subtree) -> TtTokenSource { - let mut res = TtTokenSource { tokens: Vec::new() }; - res.convert_subtree(tt); - res - } - fn convert_subtree(&mut self, sub: &tt::Subtree) { - self.push_delim(sub.delimiter, false); - let mut peek = TokenPeek { iter: itertools::multipeek(sub.token_trees.iter()) }; - while let Some(tt) = peek.iter.next() { - self.convert_tt(tt, &mut peek); - } - self.push_delim(sub.delimiter, true) - } - - fn convert_tt<'a, I>(&mut self, tt: &tt::TokenTree, iter: &mut TokenPeek<'a, I>) - where - I: Iterator, - { - match tt { - tt::TokenTree::Leaf(token) => self.convert_token(token, iter), - tt::TokenTree::Subtree(sub) => self.convert_subtree(sub), - } - } - - fn convert_token<'a, I>(&mut self, token: &tt::Leaf, iter: &mut TokenPeek<'a, I>) - where - I: Iterator, - { - let tok = match token { - tt::Leaf::Literal(l) => TtToken { - kind: classify_literal(&l.text).unwrap().kind, - is_joint_to_next: false, - text: l.text.clone(), - }, - tt::Leaf::Punct(p) => { - if let Some(tt) = Self::convert_multi_char_punct(p, iter) { - tt - } else { - let kind = match p.char { - // lexer may produce combpund tokens for these ones - '.' => DOT, - ':' => COLON, - '=' => EQ, - '!' => EXCL, - '-' => MINUS, - c => SyntaxKind::from_char(c).unwrap(), - }; - let text = { - let mut buf = [0u8; 4]; - let s: &str = p.char.encode_utf8(&mut buf); - SmolStr::new(s) - }; - TtToken { kind, is_joint_to_next: p.spacing == tt::Spacing::Joint, text } - } - } - tt::Leaf::Ident(ident) => { - let kind = SyntaxKind::from_keyword(ident.text.as_str()).unwrap_or(IDENT); - TtToken { kind, is_joint_to_next: false, text: ident.text.clone() } - } - }; - self.tokens.push(tok) - } - - fn convert_multi_char_punct<'a, I>( - p: &tt::Punct, - iter: &mut TokenPeek<'a, I>, - ) -> Option - where - I: Iterator, - { - if let Some((m, is_joint_to_next)) = iter.current_punct3(p) { - if let Some((kind, text)) = match m { - ('<', '<', '=') => Some((SHLEQ, "<<=")), - ('>', '>', '=') => Some((SHREQ, ">>=")), - ('.', '.', '.') => Some((DOTDOTDOT, "...")), - ('.', '.', '=') => Some((DOTDOTEQ, "..=")), - _ => None, - } { - iter.next(); - iter.next(); - return Some(TtToken { kind, is_joint_to_next, text: text.into() }); - } - } - - if let Some((m, is_joint_to_next)) = iter.current_punct2(p) { - if let Some((kind, text)) = match m { - ('<', '<') => Some((SHL, "<<")), - ('>', '>') => Some((SHR, ">>")), - - ('|', '|') => Some((PIPEPIPE, "||")), - ('&', '&') => Some((AMPAMP, "&&")), - ('%', '=') => Some((PERCENTEQ, "%=")), - ('*', '=') => Some((STAREQ, "*=")), - ('/', '=') => Some((SLASHEQ, "/=")), - ('^', '=') => Some((CARETEQ, "^=")), - - ('&', '=') => Some((AMPEQ, "&=")), - ('|', '=') => Some((PIPEEQ, "|=")), - ('-', '=') => Some((MINUSEQ, "-=")), - ('+', '=') => Some((PLUSEQ, "+=")), - ('>', '=') => Some((GTEQ, ">=")), - ('<', '=') => Some((LTEQ, "<=")), - - ('-', '>') => Some((THIN_ARROW, "->")), - ('!', '=') => Some((NEQ, "!=")), - ('=', '>') => Some((FAT_ARROW, "=>")), - ('=', '=') => Some((EQEQ, "==")), - ('.', '.') => Some((DOTDOT, "..")), - (':', ':') => Some((COLONCOLON, "::")), - - _ => None, - } { - iter.next(); - return Some(TtToken { kind, is_joint_to_next, text: text.into() }); - } - } - - None - } - - fn push_delim(&mut self, d: tt::Delimiter, closing: bool) { - let (kinds, texts) = match d { - tt::Delimiter::Parenthesis => ([L_PAREN, R_PAREN], "()"), - tt::Delimiter::Brace => ([L_CURLY, R_CURLY], "{}"), - tt::Delimiter::Bracket => ([L_BRACK, R_BRACK], "[]"), - tt::Delimiter::None => return, - }; - let idx = closing as usize; - let kind = kinds[idx]; - let text = &texts[idx..texts.len() - (1 - idx)]; - let tok = TtToken { kind, is_joint_to_next: false, text: SmolStr::new(text) }; - self.tokens.push(tok) - } -} - -impl TokenSource for TtTokenSource { - fn token_kind(&self, pos: usize) -> SyntaxKind { - if let Some(tok) = self.tokens.get(pos) { - tok.kind - } else { - SyntaxKind::EOF - } - } - fn is_token_joint_to_next(&self, pos: usize) -> bool { - self.tokens[pos].is_joint_to_next - } - fn is_keyword(&self, pos: usize, kw: &str) -> bool { - self.tokens[pos].text == *kw - } -} - -#[derive(Default)] -struct TtTreeSink<'a> { +struct TtTreeSink<'a, Q: Querier> { buf: String, - tokens: &'a [TtToken], + src_querier: &'a Q, text_pos: TextUnit, token_pos: usize, inner: SyntaxTreeBuilder, } -impl<'a> TtTreeSink<'a> { - fn new(tokens: &'a [TtToken]) -> TtTreeSink { +impl<'a, Q: Querier> TtTreeSink<'a, Q> { + fn new(src_querier: &'a Q) -> Self { TtTreeSink { buf: String::new(), - tokens, + src_querier, text_pos: 0.into(), token_pos: 0, inner: SyntaxTreeBuilder::default(), @@ -333,10 +125,10 @@ impl<'a> TtTreeSink<'a> { } } -impl<'a> TreeSink for TtTreeSink<'a> { +impl<'a, Q: Querier> TreeSink for TtTreeSink<'a, Q> { fn token(&mut self, kind: SyntaxKind, n_tokens: u8) { for _ in 0..n_tokens { - self.buf += self.tokens[self.token_pos].text.as_str(); + self.buf += &self.src_querier.token(self.token_pos).1; self.token_pos += 1; } self.text_pos += TextUnit::of_str(&self.buf); @@ -380,21 +172,23 @@ mod tests { "#, ); let expansion = expand(&rules, "literals!(foo)"); - let tt_src = TtTokenSource::new(&expansion); + let tt_src = SubtreeTokenSource::new(&expansion); + + let query = tt_src.querier(); // [{] // [let] [a] [=] ['c'] [;] - assert_eq!(tt_src.tokens[1 + 3].text, "'c'"); - assert_eq!(tt_src.tokens[1 + 3].kind, CHAR); + assert_eq!(query.token(1 + 3).1, "'c'"); + assert_eq!(query.token(1 + 3).0, CHAR); // [let] [c] [=] [1000] [;] - assert_eq!(tt_src.tokens[1 + 5 + 3].text, "1000"); - assert_eq!(tt_src.tokens[1 + 5 + 3].kind, INT_NUMBER); + assert_eq!(query.token(1 + 5 + 3).1, "1000"); + assert_eq!(query.token(1 + 5 + 3).0, INT_NUMBER); // [let] [f] [=] [12E+99_f64] [;] - assert_eq!(tt_src.tokens[1 + 10 + 3].text, "12E+99_f64"); - assert_eq!(tt_src.tokens[1 + 10 + 3].kind, FLOAT_NUMBER); + assert_eq!(query.token(1 + 10 + 3).1, "12E+99_f64"); + assert_eq!(query.token(1 + 10 + 3).0, FLOAT_NUMBER); // [let] [s] [=] ["rust1"] [;] - assert_eq!(tt_src.tokens[1 + 15 + 3].text, "\"rust1\""); - assert_eq!(tt_src.tokens[1 + 15 + 3].kind, STRING); + assert_eq!(query.token(1 + 15 + 3).1, "\"rust1\""); + assert_eq!(query.token(1 + 15 + 3).0, STRING); } } diff --git a/crates/ra_mbe/src/tt_cursor.rs b/crates/ra_mbe/src/tt_cursor.rs index 3128cb9ae2e..d29faa77ce8 100644 --- a/crates/ra_mbe/src/tt_cursor.rs +++ b/crates/ra_mbe/src/tt_cursor.rs @@ -1,4 +1,5 @@ use crate::ParseError; +use crate::subtree_parser::Parser; #[derive(Clone)] pub(crate) struct TtCursor<'a> { @@ -78,6 +79,11 @@ impl<'a> TtCursor<'a> { }) } + pub(crate) fn eat_path(&mut self) -> Option { + let parser = Parser::new(&mut self.pos, self.subtree); + parser.parse_path() + } + pub(crate) fn expect_char(&mut self, char: char) -> Result<(), ParseError> { if self.at_char(char) { self.bump(); diff --git a/crates/ra_parser/src/grammar.rs b/crates/ra_parser/src/grammar.rs index b2ffeff8c7a..c5f510e6b77 100644 --- a/crates/ra_parser/src/grammar.rs +++ b/crates/ra_parser/src/grammar.rs @@ -49,6 +49,10 @@ pub(crate) fn root(p: &mut Parser) { m.complete(p, SOURCE_FILE); } +pub(crate) fn path(p: &mut Parser) { + paths::type_path(p); +} + pub(crate) fn reparser( node: SyntaxKind, first_child: Option, diff --git a/crates/ra_parser/src/lib.rs b/crates/ra_parser/src/lib.rs index 30ba06aacda..3ceeeebd751 100644 --- a/crates/ra_parser/src/lib.rs +++ b/crates/ra_parser/src/lib.rs @@ -61,6 +61,14 @@ pub fn parse(token_source: &dyn TokenSource, tree_sink: &mut dyn TreeSink) { event::process(tree_sink, events); } +/// Parse given tokens into the given sink as a path +pub fn parse_path(token_source: &dyn TokenSource, tree_sink: &mut dyn TreeSink) { + let mut p = parser::Parser::new(token_source); + grammar::path(&mut p); + let events = p.finish(); + event::process(tree_sink, events); +} + /// A parsing function for a specific braced-block. pub struct Reparser(fn(&mut parser::Parser));