2302: Collapse TokenMap and RevTokenMap r=matklad a=matklad



Co-authored-by: Aleksey Kladov <aleksey.kladov@gmail.com>
This commit is contained in:
bors[bot] 2019-11-18 13:09:29 +00:00 committed by GitHub
commit 9cc4d058d2
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
4 changed files with 105 additions and 119 deletions

View file

@ -59,10 +59,8 @@ pub trait AstDatabase: SourceDatabase {
fn intern_macro(&self, macro_call: MacroCallLoc) -> MacroCallId; fn intern_macro(&self, macro_call: MacroCallLoc) -> MacroCallId;
fn macro_arg(&self, id: MacroCallId) -> Option<Arc<(tt::Subtree, mbe::TokenMap)>>; fn macro_arg(&self, id: MacroCallId) -> Option<Arc<(tt::Subtree, mbe::TokenMap)>>;
fn macro_def(&self, id: MacroDefId) -> Option<Arc<(TokenExpander, mbe::TokenMap)>>; fn macro_def(&self, id: MacroDefId) -> Option<Arc<(TokenExpander, mbe::TokenMap)>>;
fn parse_macro( fn parse_macro(&self, macro_file: MacroFile)
&self, -> Option<(Parse<SyntaxNode>, Arc<mbe::TokenMap>)>;
macro_file: MacroFile,
) -> Option<(Parse<SyntaxNode>, Arc<mbe::RevTokenMap>)>;
fn macro_expand(&self, macro_call: MacroCallId) -> Result<Arc<tt::Subtree>, String>; fn macro_expand(&self, macro_call: MacroCallId) -> Result<Arc<tt::Subtree>, String>;
} }
@ -136,7 +134,7 @@ pub(crate) fn parse_or_expand(db: &dyn AstDatabase, file_id: HirFileId) -> Optio
pub(crate) fn parse_macro( pub(crate) fn parse_macro(
db: &dyn AstDatabase, db: &dyn AstDatabase,
macro_file: MacroFile, macro_file: MacroFile,
) -> Option<(Parse<SyntaxNode>, Arc<mbe::RevTokenMap>)> { ) -> Option<(Parse<SyntaxNode>, Arc<mbe::TokenMap>)> {
let _p = profile("parse_macro_query"); let _p = profile("parse_macro_query");
let macro_call_id = macro_file.macro_call_id; let macro_call_id = macro_file.macro_call_id;

View file

@ -159,7 +159,7 @@ pub struct ExpansionInfo {
macro_def: Arc<(db::TokenExpander, mbe::TokenMap)>, macro_def: Arc<(db::TokenExpander, mbe::TokenMap)>,
macro_arg: Arc<(tt::Subtree, mbe::TokenMap)>, macro_arg: Arc<(tt::Subtree, mbe::TokenMap)>,
exp_map: Arc<mbe::RevTokenMap>, exp_map: Arc<mbe::TokenMap>,
} }
impl ExpansionInfo { impl ExpansionInfo {
@ -186,7 +186,7 @@ impl ExpansionInfo {
mbe::Origin::Def => (&self.macro_def.1, &self.def), 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( let token = algo::find_covering_element(
tt.ast.syntax(), tt.ast.syntax(),
range + tt.ast.syntax().text_range().start(), range + tt.ast.syntax().text_range().start(),

View file

@ -31,7 +31,7 @@ pub enum ExpandError {
} }
pub use crate::syntax_bridge::{ 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 /// This struct contains AST for a single `macro_rules` definition. What might

View file

@ -14,30 +14,22 @@ use crate::ExpandError;
#[derive(Debug, PartialEq, Eq, Default)] #[derive(Debug, PartialEq, Eq, Default)]
pub struct TokenMap { pub struct TokenMap {
/// Maps `tt::TokenId` to the *relative* source range. /// Maps `tt::TokenId` to the *relative* source range.
tokens: Vec<TextRange>, entries: Vec<(tt::TokenId, TextRange)>,
}
/// Maps relative range of the expanded syntax node to `tt::TokenId`
#[derive(Debug, PartialEq, Eq, Default)]
pub struct RevTokenMap {
ranges: Vec<(TextRange, tt::TokenId)>,
} }
/// Convert the syntax tree (what user has written) to a `TokenTree` (what macro /// Convert the syntax tree (what user has written) to a `TokenTree` (what macro
/// will consume). /// will consume).
pub fn ast_to_token_tree(ast: &ast::TokenTree) -> Option<(tt::Subtree, TokenMap)> { pub fn ast_to_token_tree(ast: &ast::TokenTree) -> Option<(tt::Subtree, TokenMap)> {
let mut token_map = TokenMap::default(); syntax_node_to_token_tree(ast.syntax())
let node = ast.syntax();
let tt = convert_tt(&mut token_map, node.text_range().start(), node)?;
Some((tt, token_map))
} }
/// Convert the syntax node to a `TokenTree` (what macro /// Convert the syntax node to a `TokenTree` (what macro
/// will consume). /// will consume).
pub fn syntax_node_to_token_tree(node: &SyntaxNode) -> Option<(tt::Subtree, TokenMap)> { pub fn syntax_node_to_token_tree(node: &SyntaxNode) -> Option<(tt::Subtree, TokenMap)> {
let mut token_map = TokenMap::default(); let global_offset = node.text_range().start();
let tt = convert_tt(&mut token_map, node.text_range().start(), node)?; let mut c = Convertor { map: TokenMap::default(), global_offset, next_id: 0 };
Some((tt, token_map)) let subtree = c.go(node)?;
Some((subtree, c.map))
} }
// The following items are what `rustc` macro can be parsed into : // 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( pub fn token_tree_to_syntax_node(
tt: &tt::Subtree, tt: &tt::Subtree,
fragment_kind: FragmentKind, fragment_kind: FragmentKind,
) -> Result<(Parse<SyntaxNode>, RevTokenMap), ExpandError> { ) -> Result<(Parse<SyntaxNode>, TokenMap), ExpandError> {
let tmp; let tmp;
let tokens = match tt { let tokens = match tt {
tt::Subtree { delimiter: tt::Delimiter::None, token_trees } => token_trees.as_slice(), 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 { impl TokenMap {
pub fn token_by_range(&self, relative_range: TextRange) -> Option<tt::TokenId> { pub fn token_by_range(&self, relative_range: TextRange) -> Option<tt::TokenId> {
let (idx, _) = let &(token_id, _) = self.entries.iter().find(|(_, range)| *range == relative_range)?;
self.tokens.iter().enumerate().find(|(_, range)| **range == relative_range)?; Some(token_id)
Some(tt::TokenId(idx as u32))
}
pub fn relative_range_of(&self, token_id: tt::TokenId) -> Option<TextRange> {
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<tt::TokenId> {
self.ranges.iter().find(|&it| it.0 == relative_range).map(|it| it.1)
} }
pub fn range_by_token(&self, token_id: tt::TokenId) -> Option<TextRange> { pub fn range_by_token(&self, token_id: tt::TokenId) -> Option<TextRange> {
let &(r, _) = self.ranges.iter().find(|(_, tid)| *tid == token_id)?; let &(_, range) = self.entries.iter().find(|(tid, _)| *tid == token_id)?;
Some(r) Some(range)
} }
fn add(&mut self, relative_range: TextRange, token_id: tt::TokenId) { fn insert(&mut self, token_id: tt::TokenId, relative_range: TextRange) {
self.ranges.push((relative_range, token_id.clone())) self.entries.push((token_id, relative_range));
} }
} }
@ -171,84 +145,98 @@ fn convert_doc_comment(token: &ra_syntax::SyntaxToken) -> Option<Vec<tt::TokenTr
} }
} }
fn convert_tt( struct Convertor {
token_map: &mut TokenMap, map: TokenMap,
global_offset: TextUnit, global_offset: TextUnit,
tt: &SyntaxNode, next_id: u32,
) -> Option<tt::Subtree> { }
// This tree is empty
if tt.first_child_or_token().is_none() {
return Some(tt::Subtree { token_trees: vec![], delimiter: tt::Delimiter::None });
}
let first_child = tt.first_child_or_token()?; impl Convertor {
let last_child = tt.last_child_or_token()?; fn go(&mut self, tt: &SyntaxNode) -> Option<tt::Subtree> {
let (delimiter, skip_first) = match (first_child.kind(), last_child.kind()) { // This tree is empty
(T!['('], T![')']) => (tt::Delimiter::Parenthesis, true), if tt.first_child_or_token().is_none() {
(T!['{'], T!['}']) => (tt::Delimiter::Brace, true), return Some(tt::Subtree { token_trees: vec![], delimiter: tt::Delimiter::None });
(T!['['], T![']']) => (tt::Delimiter::Bracket, true),
_ => (tt::Delimiter::None, false),
};
let mut token_trees = Vec::new();
let mut child_iter = tt.children_with_tokens().skip(skip_first as usize).peekable();
while let Some(child) = child_iter.next() {
if skip_first && (child == first_child || child == last_child) {
continue;
} }
match child { let first_child = tt.first_child_or_token()?;
NodeOrToken::Token(token) => { let last_child = tt.last_child_or_token()?;
if let Some(doc_tokens) = convert_doc_comment(&token) { let (delimiter, skip_first) = match (first_child.kind(), last_child.kind()) {
token_trees.extend(doc_tokens); (T!['('], T![')']) => (tt::Delimiter::Parenthesis, true),
} else if token.kind().is_trivia() { (T!['{'], T!['}']) => (tt::Delimiter::Brace, true),
continue; (T!['['], T![']']) => (tt::Delimiter::Bracket, true),
} else if token.kind().is_punct() { _ => (tt::Delimiter::None, false),
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() { let mut token_trees = Vec::new();
Some(NodeOrToken::Token(token)) => { let mut child_iter = tt.children_with_tokens().skip(skip_first as usize).peekable();
if token.kind().is_punct() {
tt::Spacing::Joint while let Some(child) = child_iter.next() {
} else { if skip_first && (child == first_child || child == last_child) {
tt::Spacing::Alone 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,
_ => 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;
}; };
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); 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 }; fn alloc(&mut self, absolute_range: TextRange) -> tt::TokenId {
Some(res) 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> { struct TtTreeSink<'a> {
@ -256,7 +244,7 @@ struct TtTreeSink<'a> {
cursor: Cursor<'a>, cursor: Cursor<'a>,
text_pos: TextUnit, text_pos: TextUnit,
inner: SyntaxTreeBuilder, inner: SyntaxTreeBuilder,
range_map: RevTokenMap, token_map: TokenMap,
// Number of roots // Number of roots
// Use for detect ill-form tree which is not single root // Use for detect ill-form tree which is not single root
@ -271,12 +259,12 @@ impl<'a> TtTreeSink<'a> {
text_pos: 0.into(), text_pos: 0.into(),
inner: SyntaxTreeBuilder::default(), inner: SyntaxTreeBuilder::default(),
roots: smallvec::SmallVec::new(), roots: smallvec::SmallVec::new(),
range_map: RevTokenMap::default(), token_map: TokenMap::default(),
} }
} }
fn finish(self) -> (Parse<SyntaxNode>, RevTokenMap) { fn finish(self) -> (Parse<SyntaxNode>, TokenMap) {
(self.inner.finish(), self.range_map) (self.inner.finish(), self.token_map)
} }
} }
@ -312,7 +300,7 @@ impl<'a> TreeSink for TtTreeSink<'a> {
if kind == IDENT { if kind == IDENT {
let range = let range =
TextRange::offset_len(self.text_pos, TextUnit::of_str(&ident.text)); 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);
} }
} }