From 2b6f1ab0e2303dce8e6e424514fd2a9e74566c89 Mon Sep 17 00:00:00 2001 From: Aleksey Kladov Date: Mon, 18 Nov 2019 16:08:41 +0300 Subject: Collapse TokenMap and RevTokenMap --- crates/ra_hir_expand/src/db.rs | 8 +- crates/ra_hir_expand/src/lib.rs | 4 +- crates/ra_mbe/src/lib.rs | 2 +- crates/ra_mbe/src/syntax_bridge.rs | 210 +++++++++++++++++-------------------- 4 files changed, 105 insertions(+), 119 deletions(-) (limited to 'crates') diff --git a/crates/ra_hir_expand/src/db.rs b/crates/ra_hir_expand/src/db.rs index 9de7c1ea8..3c11c8a22 100644 --- a/crates/ra_hir_expand/src/db.rs +++ b/crates/ra_hir_expand/src/db.rs @@ -59,10 +59,8 @@ pub trait AstDatabase: SourceDatabase { fn intern_macro(&self, macro_call: MacroCallLoc) -> MacroCallId; fn macro_arg(&self, id: MacroCallId) -> Option>; fn macro_def(&self, id: MacroDefId) -> Option>; - fn parse_macro( - &self, - macro_file: MacroFile, - ) -> Option<(Parse, Arc)>; + fn parse_macro(&self, macro_file: MacroFile) + -> Option<(Parse, Arc)>; fn macro_expand(&self, macro_call: MacroCallId) -> Result, String>; } @@ -136,7 +134,7 @@ pub(crate) fn parse_or_expand(db: &dyn AstDatabase, file_id: HirFileId) -> Optio pub(crate) fn parse_macro( db: &dyn AstDatabase, macro_file: MacroFile, -) -> Option<(Parse, Arc)> { +) -> Option<(Parse, Arc)> { let _p = profile("parse_macro_query"); let macro_call_id = macro_file.macro_call_id; diff --git a/crates/ra_hir_expand/src/lib.rs b/crates/ra_hir_expand/src/lib.rs index 73ec1688c..cfe7e6d15 100644 --- a/crates/ra_hir_expand/src/lib.rs +++ b/crates/ra_hir_expand/src/lib.rs @@ -159,7 +159,7 @@ pub struct ExpansionInfo { macro_def: Arc<(db::TokenExpander, mbe::TokenMap)>, macro_arg: Arc<(tt::Subtree, mbe::TokenMap)>, - exp_map: Arc, + exp_map: Arc, } impl ExpansionInfo { @@ -186,7 +186,7 @@ impl ExpansionInfo { mbe::Origin::Def => (&self.macro_def.1, &self.def), }; - let range = token_map.relative_range_of(token_id)?; + let range = token_map.range_by_token(token_id)?; let token = algo::find_covering_element( tt.ast.syntax(), range + tt.ast.syntax().text_range().start(), diff --git a/crates/ra_mbe/src/lib.rs b/crates/ra_mbe/src/lib.rs index 58ca95368..bbddebe67 100644 --- a/crates/ra_mbe/src/lib.rs +++ b/crates/ra_mbe/src/lib.rs @@ -31,7 +31,7 @@ pub enum ExpandError { } pub use crate::syntax_bridge::{ - ast_to_token_tree, syntax_node_to_token_tree, token_tree_to_syntax_node, RevTokenMap, TokenMap, + ast_to_token_tree, syntax_node_to_token_tree, token_tree_to_syntax_node, TokenMap, }; /// This struct contains AST for a single `macro_rules` definition. What might diff --git a/crates/ra_mbe/src/syntax_bridge.rs b/crates/ra_mbe/src/syntax_bridge.rs index fe3b70b8d..d1c49c0b3 100644 --- a/crates/ra_mbe/src/syntax_bridge.rs +++ b/crates/ra_mbe/src/syntax_bridge.rs @@ -14,30 +14,22 @@ use crate::ExpandError; #[derive(Debug, PartialEq, Eq, Default)] pub struct TokenMap { /// Maps `tt::TokenId` to the *relative* source range. - tokens: Vec, -} - -/// Maps relative range of the expanded syntax node to `tt::TokenId` -#[derive(Debug, PartialEq, Eq, Default)] -pub struct RevTokenMap { - ranges: Vec<(TextRange, tt::TokenId)>, + entries: Vec<(tt::TokenId, TextRange)>, } /// Convert the syntax tree (what user has written) to a `TokenTree` (what macro /// will consume). pub fn ast_to_token_tree(ast: &ast::TokenTree) -> Option<(tt::Subtree, TokenMap)> { - let mut token_map = TokenMap::default(); - let node = ast.syntax(); - let tt = convert_tt(&mut token_map, node.text_range().start(), node)?; - Some((tt, token_map)) + syntax_node_to_token_tree(ast.syntax()) } /// Convert the syntax node to a `TokenTree` (what macro /// will consume). pub fn syntax_node_to_token_tree(node: &SyntaxNode) -> Option<(tt::Subtree, TokenMap)> { - let mut token_map = TokenMap::default(); - let tt = convert_tt(&mut token_map, node.text_range().start(), node)?; - Some((tt, token_map)) + let global_offset = node.text_range().start(); + let mut c = Convertor { map: TokenMap::default(), global_offset, next_id: 0 }; + let subtree = c.go(node)?; + Some((subtree, c.map)) } // The following items are what `rustc` macro can be parsed into : @@ -55,7 +47,7 @@ pub fn syntax_node_to_token_tree(node: &SyntaxNode) -> Option<(tt::Subtree, Toke pub fn token_tree_to_syntax_node( tt: &tt::Subtree, fragment_kind: FragmentKind, -) -> Result<(Parse, RevTokenMap), ExpandError> { +) -> Result<(Parse, TokenMap), ExpandError> { let tmp; let tokens = match tt { tt::Subtree { delimiter: tt::Delimiter::None, token_trees } => token_trees.as_slice(), @@ -78,35 +70,17 @@ pub fn token_tree_to_syntax_node( impl TokenMap { pub fn token_by_range(&self, relative_range: TextRange) -> Option { - let (idx, _) = - self.tokens.iter().enumerate().find(|(_, range)| **range == relative_range)?; - Some(tt::TokenId(idx as u32)) - } - - pub fn relative_range_of(&self, token_id: tt::TokenId) -> Option { - let idx = token_id.0 as usize; - self.tokens.get(idx).copied() - } - - fn alloc(&mut self, relative_range: TextRange) -> tt::TokenId { - let id = self.tokens.len(); - self.tokens.push(relative_range); - tt::TokenId(id as u32) - } -} - -impl RevTokenMap { - pub fn token_by_range(&self, relative_range: TextRange) -> Option { - self.ranges.iter().find(|&it| it.0 == relative_range).map(|it| it.1) + let &(token_id, _) = self.entries.iter().find(|(_, range)| *range == relative_range)?; + Some(token_id) } pub fn range_by_token(&self, token_id: tt::TokenId) -> Option { - let &(r, _) = self.ranges.iter().find(|(_, tid)| *tid == token_id)?; - Some(r) + let &(_, range) = self.entries.iter().find(|(tid, _)| *tid == token_id)?; + Some(range) } - fn add(&mut self, relative_range: TextRange, token_id: tt::TokenId) { - self.ranges.push((relative_range, token_id.clone())) + fn insert(&mut self, token_id: tt::TokenId, relative_range: TextRange) { + self.entries.push((token_id, relative_range)); } } @@ -171,84 +145,98 @@ fn convert_doc_comment(token: &ra_syntax::SyntaxToken) -> Option Option { - // This tree is empty - if tt.first_child_or_token().is_none() { - return Some(tt::Subtree { token_trees: vec![], delimiter: tt::Delimiter::None }); - } + next_id: u32, +} - let first_child = tt.first_child_or_token()?; - let last_child = tt.last_child_or_token()?; - let (delimiter, skip_first) = match (first_child.kind(), last_child.kind()) { - (T!['('], T![')']) => (tt::Delimiter::Parenthesis, true), - (T!['{'], T!['}']) => (tt::Delimiter::Brace, true), - (T!['['], T![']']) => (tt::Delimiter::Bracket, true), - _ => (tt::Delimiter::None, false), - }; +impl Convertor { + fn go(&mut self, tt: &SyntaxNode) -> Option { + // This tree is empty + if tt.first_child_or_token().is_none() { + return Some(tt::Subtree { token_trees: vec![], delimiter: tt::Delimiter::None }); + } - let mut token_trees = Vec::new(); - let mut child_iter = tt.children_with_tokens().skip(skip_first as usize).peekable(); + let first_child = tt.first_child_or_token()?; + let last_child = tt.last_child_or_token()?; + let (delimiter, skip_first) = match (first_child.kind(), last_child.kind()) { + (T!['('], T![')']) => (tt::Delimiter::Parenthesis, true), + (T!['{'], T!['}']) => (tt::Delimiter::Brace, true), + (T!['['], T![']']) => (tt::Delimiter::Bracket, true), + _ => (tt::Delimiter::None, false), + }; - while let Some(child) = child_iter.next() { - if skip_first && (child == first_child || child == last_child) { - continue; - } + let mut token_trees = Vec::new(); + let mut child_iter = tt.children_with_tokens().skip(skip_first as usize).peekable(); - match child { - NodeOrToken::Token(token) => { - if let Some(doc_tokens) = convert_doc_comment(&token) { - token_trees.extend(doc_tokens); - } else if token.kind().is_trivia() { - continue; - } else if token.kind().is_punct() { - assert!(token.text().len() == 1, "Input ast::token punct must be single char."); - let char = token.text().chars().next().unwrap(); - - let spacing = match child_iter.peek() { - Some(NodeOrToken::Token(token)) => { - if token.kind().is_punct() { - tt::Spacing::Joint - } else { - tt::Spacing::Alone + while let Some(child) = child_iter.next() { + if skip_first && (child == first_child || child == last_child) { + continue; + } + + match child { + NodeOrToken::Token(token) => { + if let Some(doc_tokens) = convert_doc_comment(&token) { + token_trees.extend(doc_tokens); + } else if token.kind().is_trivia() { + continue; + } else if token.kind().is_punct() { + assert!( + token.text().len() == 1, + "Input ast::token punct must be single char." + ); + let char = token.text().chars().next().unwrap(); + + let spacing = match child_iter.peek() { + Some(NodeOrToken::Token(token)) => { + if token.kind().is_punct() { + tt::Spacing::Joint + } else { + tt::Spacing::Alone + } } - } - _ => tt::Spacing::Alone, - }; - - token_trees.push(tt::Leaf::from(tt::Punct { char, spacing }).into()); - } else { - let child: tt::TokenTree = - if token.kind() == T![true] || token.kind() == T![false] { - tt::Leaf::from(tt::Literal { text: token.text().clone() }).into() - } else if token.kind().is_keyword() - || token.kind() == IDENT - || token.kind() == LIFETIME - { - let relative_range = token.text_range() - global_offset; - let id = token_map.alloc(relative_range); - let text = token.text().clone(); - tt::Leaf::from(tt::Ident { text, id }).into() - } else if token.kind().is_literal() { - tt::Leaf::from(tt::Literal { text: token.text().clone() }).into() - } else { - return None; + _ => tt::Spacing::Alone, }; + + token_trees.push(tt::Leaf::from(tt::Punct { char, spacing }).into()); + } else { + let child: tt::TokenTree = + if token.kind() == T![true] || token.kind() == T![false] { + tt::Leaf::from(tt::Literal { text: token.text().clone() }).into() + } else if token.kind().is_keyword() + || token.kind() == IDENT + || token.kind() == LIFETIME + { + let id = self.alloc(token.text_range()); + let text = token.text().clone(); + tt::Leaf::from(tt::Ident { text, id }).into() + } else if token.kind().is_literal() { + tt::Leaf::from(tt::Literal { text: token.text().clone() }).into() + } else { + return None; + }; + token_trees.push(child); + } + } + NodeOrToken::Node(node) => { + let child = self.go(&node)?.into(); token_trees.push(child); } - } - NodeOrToken::Node(node) => { - let child = convert_tt(token_map, global_offset, &node)?.into(); - token_trees.push(child); - } - }; + }; + } + + let res = tt::Subtree { delimiter, token_trees }; + Some(res) } - let res = tt::Subtree { delimiter, token_trees }; - Some(res) + fn alloc(&mut self, absolute_range: TextRange) -> tt::TokenId { + let relative_range = absolute_range - self.global_offset; + let token_id = tt::TokenId(self.next_id); + self.next_id += 1; + self.map.insert(token_id, relative_range); + token_id + } } struct TtTreeSink<'a> { @@ -256,7 +244,7 @@ struct TtTreeSink<'a> { cursor: Cursor<'a>, text_pos: TextUnit, inner: SyntaxTreeBuilder, - range_map: RevTokenMap, + token_map: TokenMap, // Number of roots // Use for detect ill-form tree which is not single root @@ -271,12 +259,12 @@ impl<'a> TtTreeSink<'a> { text_pos: 0.into(), inner: SyntaxTreeBuilder::default(), roots: smallvec::SmallVec::new(), - range_map: RevTokenMap::default(), + token_map: TokenMap::default(), } } - fn finish(self) -> (Parse, RevTokenMap) { - (self.inner.finish(), self.range_map) + fn finish(self) -> (Parse, TokenMap) { + (self.inner.finish(), self.token_map) } } @@ -312,7 +300,7 @@ impl<'a> TreeSink for TtTreeSink<'a> { if kind == IDENT { let range = TextRange::offset_len(self.text_pos, TextUnit::of_str(&ident.text)); - self.range_map.add(range, ident.id); + self.token_map.insert(ident.id, range); } } -- cgit v1.2.3