-//! FIXME: write short doc here
+//! Conversions between [`SyntaxNode`] and [`tt::TokenTree`].
use parser::{FragmentKind, ParseError, TreeSink};
use rustc_hash::FxHashMap;
};
use tt::buffer::{Cursor, TokenBuffer};
-use crate::subtree_source::SubtreeTokenSource;
-use crate::ExpandError;
-
-#[derive(Debug, PartialEq, Eq, Clone, Copy)]
-pub enum TokenTextRange {
- Token(TextRange),
- Delimiter(TextRange),
-}
-
-impl TokenTextRange {
- pub fn by_kind(self, kind: SyntaxKind) -> Option<TextRange> {
- match self {
- TokenTextRange::Token(it) => Some(it),
- TokenTextRange::Delimiter(it) => match kind {
- T!['{'] | T!['('] | T!['['] => Some(TextRange::at(it.start(), 1.into())),
- T!['}'] | T![')'] | T![']'] => {
- Some(TextRange::at(it.end() - TextSize::of('}'), 1.into()))
- }
- _ => None,
- },
- }
- }
-}
-
-/// Maps `tt::TokenId` to the relative range of the original token.
-#[derive(Debug, PartialEq, Eq, Clone, Default)]
-pub struct TokenMap {
- /// Maps `tt::TokenId` to the *relative* source range.
- entries: Vec<(tt::TokenId, TokenTextRange)>,
-}
+use crate::{subtree_source::SubtreeTokenSource, tt_iter::TtIter};
+use crate::{ExpandError, TokenMap};
/// Convert the syntax tree (what user has written) to a `TokenTree` (what macro
/// will consume).
-pub fn ast_to_token_tree(ast: &impl ast::AstNode) -> Option<(tt::Subtree, TokenMap)> {
+pub fn ast_to_token_tree(ast: &impl ast::AstNode) -> (tt::Subtree, TokenMap) {
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)> {
+pub fn syntax_node_to_token_tree(node: &SyntaxNode) -> (tt::Subtree, TokenMap) {
let global_offset = node.text_range().start();
let mut c = Convertor::new(node, global_offset);
- let subtree = c.go()?;
- c.id_alloc.map.entries.shrink_to_fit();
- Some((subtree, c.id_alloc.map))
+ let subtree = convert_tokens(&mut c);
+ c.id_alloc.map.shrink_to_fit();
+ (subtree, c.id_alloc.map)
}
// The following items are what `rustc` macro can be parsed into :
},
};
- let subtree = conv.go()?;
+ let subtree = convert_tokens(&mut conv);
Some((subtree, conv.id_alloc.map))
}
-impl TokenMap {
- pub fn token_by_range(&self, relative_range: TextRange) -> Option<tt::TokenId> {
- let &(token_id, _) = self.entries.iter().find(|(_, range)| match range {
- TokenTextRange::Token(it) => *it == relative_range,
- TokenTextRange::Delimiter(it) => {
- let open = TextRange::at(it.start(), 1.into());
- let close = TextRange::at(it.end() - TextSize::of('}'), 1.into());
- open == relative_range || close == relative_range
- }
- })?;
- Some(token_id)
+/// Split token tree with seperate expr: $($e:expr)SEP*
+pub fn parse_exprs_with_sep(tt: &tt::Subtree, sep: char) -> Vec<tt::Subtree> {
+ if tt.token_trees.is_empty() {
+ return Vec::new();
}
- pub fn range_by_token(&self, token_id: tt::TokenId) -> Option<TokenTextRange> {
- let &(_, range) = self.entries.iter().find(|(tid, _)| *tid == token_id)?;
- Some(range)
+ let mut iter = TtIter::new(tt);
+ let mut res = Vec::new();
+
+ while iter.peek_n(0).is_some() {
+ let expanded = iter.expect_fragment(FragmentKind::Expr);
+ if expanded.err.is_some() {
+ break;
+ }
+
+ res.push(match expanded.value {
+ None => break,
+ Some(tt @ tt::TokenTree::Leaf(_)) => {
+ tt::Subtree { delimiter: None, token_trees: vec![tt] }
+ }
+ Some(tt::TokenTree::Subtree(tt)) => tt,
+ });
+
+ let mut fork = iter.clone();
+ if fork.expect_char(sep).is_err() {
+ break;
+ }
+ iter = fork;
}
- fn insert(&mut self, token_id: tt::TokenId, relative_range: TextRange) {
- self.entries.push((token_id, TokenTextRange::Token(relative_range)));
+ if iter.peek_n(0).is_some() {
+ res.push(tt::Subtree { delimiter: None, token_trees: iter.into_iter().cloned().collect() });
}
- fn insert_delim(
- &mut self,
- token_id: tt::TokenId,
- open_relative_range: TextRange,
- close_relative_range: TextRange,
- ) -> usize {
- let res = self.entries.len();
- let cover = open_relative_range.cover(close_relative_range);
+ res
+}
- self.entries.push((token_id, TokenTextRange::Delimiter(cover)));
- res
+fn convert_tokens<C: TokenConvertor>(conv: &mut C) -> tt::Subtree {
+ let mut subtree = tt::Subtree { delimiter: None, ..Default::default() };
+ while conv.peek().is_some() {
+ collect_leaf(conv, &mut subtree.token_trees);
}
-
- fn update_close_delim(&mut self, idx: usize, close_relative_range: TextRange) {
- let (_, token_text_range) = &mut self.entries[idx];
- if let TokenTextRange::Delimiter(dim) = token_text_range {
- let cover = dim.cover(close_relative_range);
- *token_text_range = TokenTextRange::Delimiter(cover);
+ if subtree.token_trees.len() == 1 {
+ if let tt::TokenTree::Subtree(first) = &subtree.token_trees[0] {
+ return first.clone();
}
}
+ return subtree;
+
+ fn collect_leaf<C: TokenConvertor>(conv: &mut C, result: &mut Vec<tt::TokenTree>) {
+ let (token, range) = match conv.bump() {
+ None => return,
+ Some(it) => it,
+ };
+
+ let k: SyntaxKind = token.kind();
+ if k == COMMENT {
+ if let Some(tokens) = conv.convert_doc_comment(&token) {
+ result.extend(tokens);
+ }
+ return;
+ }
+
+ result.push(if k.is_punct() && k != UNDERSCORE {
+ assert_eq!(range.len(), TextSize::of('.'));
+ let delim = match k {
+ T!['('] => Some((tt::DelimiterKind::Parenthesis, T![')'])),
+ T!['{'] => Some((tt::DelimiterKind::Brace, T!['}'])),
+ T!['['] => Some((tt::DelimiterKind::Bracket, T![']'])),
+ _ => None,
+ };
+
+ if let Some((kind, closed)) = delim {
+ let mut subtree = tt::Subtree::default();
+ let (id, idx) = conv.id_alloc().open_delim(range);
+ subtree.delimiter = Some(tt::Delimiter { id, kind });
+
+ while conv.peek().map_or(false, |it| it.kind() != closed) {
+ collect_leaf(conv, &mut subtree.token_trees);
+ }
+ let last_range = match conv.bump() {
+ None => {
+ // For error resilience, we insert an char punct for the opening delim here
+ conv.id_alloc().close_delim(idx, None);
+ let leaf: tt::Leaf = tt::Punct {
+ id: conv.id_alloc().alloc(range),
+ char: token.to_char().unwrap(),
+ spacing: tt::Spacing::Alone,
+ }
+ .into();
+ result.push(leaf.into());
+ result.extend(subtree.token_trees);
+ return;
+ }
+ Some(it) => it.1,
+ };
+ conv.id_alloc().close_delim(idx, Some(last_range));
+ subtree.into()
+ } else {
+ let spacing = match conv.peek() {
+ Some(next)
+ if next.kind().is_trivia()
+ || next.kind() == T!['[']
+ || next.kind() == T!['{']
+ || next.kind() == T!['('] =>
+ {
+ tt::Spacing::Alone
+ }
+ Some(next) if next.kind().is_punct() && next.kind() != UNDERSCORE => {
+ tt::Spacing::Joint
+ }
+ _ => tt::Spacing::Alone,
+ };
+ let char = match token.to_char() {
+ Some(c) => c,
+ None => {
+ panic!("Token from lexer must be single char: token = {:#?}", token);
+ }
+ };
+ tt::Leaf::from(tt::Punct { char, spacing, id: conv.id_alloc().alloc(range) }).into()
+ }
+ } else {
+ macro_rules! make_leaf {
+ ($i:ident) => {
+ tt::$i { id: conv.id_alloc().alloc(range), text: token.to_text() }.into()
+ };
+ }
+ let leaf: tt::Leaf = match k {
+ T![true] | T![false] => make_leaf!(Ident),
+ IDENT => make_leaf!(Ident),
+ UNDERSCORE => make_leaf!(Ident),
+ k if k.is_keyword() => make_leaf!(Ident),
+ k if k.is_literal() => make_leaf!(Literal),
+ LIFETIME_IDENT => {
+ let char_unit = TextSize::of('\'');
+ let r = TextRange::at(range.start(), char_unit);
+ let apostrophe = tt::Leaf::from(tt::Punct {
+ char: '\'',
+ spacing: tt::Spacing::Joint,
+ id: conv.id_alloc().alloc(r),
+ });
+ result.push(apostrophe.into());
+
+ let r = TextRange::at(range.start() + char_unit, range.len() - char_unit);
+ let ident = tt::Leaf::from(tt::Ident {
+ text: SmolStr::new(&token.to_text()[1..]),
+ id: conv.id_alloc().alloc(r),
+ });
+ result.push(ident.into());
+ return;
+ }
+ _ => return,
+ };
- fn remove_delim(&mut self, idx: usize) {
- // FIXME: This could be accidentally quadratic
- self.entries.remove(idx);
+ leaf.into()
+ });
}
}
// Quote the string
// Note that `tt::Literal` expect an escaped string
- let text = format!("{:?}", text.escape_default().to_string());
+ let text = format!("\"{}\"", text.escape_debug());
text.into()
}
let doc = comment.kind().doc?;
// Make `doc="\" Comments\""
- let mut meta_tkns = Vec::new();
- meta_tkns.push(mk_ident("doc"));
- meta_tkns.push(mk_punct('='));
- meta_tkns.push(mk_doc_literal(&comment));
+ let meta_tkns = vec![mk_ident("doc"), mk_punct('='), mk_doc_literal(&comment)];
// Make `#![]`
- let mut token_trees = Vec::new();
- token_trees.push(mk_punct('#'));
+ let mut token_trees = vec![mk_punct('#')];
if let ast::CommentPlacement::Inner = doc {
token_trees.push(mk_punct('!'));
}
trait TokenConvertor {
type Token: SrcToken;
- fn go(&mut self) -> Option<tt::Subtree> {
- let mut subtree = tt::Subtree::default();
- subtree.delimiter = None;
- while self.peek().is_some() {
- self.collect_leaf(&mut subtree.token_trees);
- }
- if subtree.token_trees.is_empty() {
- return None;
- }
- if subtree.token_trees.len() == 1 {
- if let tt::TokenTree::Subtree(first) = &subtree.token_trees[0] {
- return Some(first.clone());
- }
- }
- Some(subtree)
- }
-
- fn collect_leaf(&mut self, result: &mut Vec<tt::TokenTree>) {
- let (token, range) = match self.bump() {
- None => return,
- Some(it) => it,
- };
-
- let k: SyntaxKind = token.kind();
- if k == COMMENT {
- if let Some(tokens) = self.convert_doc_comment(&token) {
- result.extend(tokens);
- }
- return;
- }
-
- result.push(if k.is_punct() {
- assert_eq!(range.len(), TextSize::of('.'));
- let delim = match k {
- T!['('] => Some((tt::DelimiterKind::Parenthesis, T![')'])),
- T!['{'] => Some((tt::DelimiterKind::Brace, T!['}'])),
- T!['['] => Some((tt::DelimiterKind::Bracket, T![']'])),
- _ => None,
- };
-
- if let Some((kind, closed)) = delim {
- let mut subtree = tt::Subtree::default();
- let (id, idx) = self.id_alloc().open_delim(range);
- subtree.delimiter = Some(tt::Delimiter { kind, id });
-
- while self.peek().map(|it| it.kind() != closed).unwrap_or(false) {
- self.collect_leaf(&mut subtree.token_trees);
- }
- let last_range = match self.bump() {
- None => {
- // For error resilience, we insert an char punct for the opening delim here
- self.id_alloc().close_delim(idx, None);
- let leaf: tt::Leaf = tt::Punct {
- id: self.id_alloc().alloc(range),
- char: token.to_char().unwrap(),
- spacing: tt::Spacing::Alone,
- }
- .into();
- result.push(leaf.into());
- result.extend(subtree.token_trees);
- return;
- }
- Some(it) => it.1,
- };
- self.id_alloc().close_delim(idx, Some(last_range));
- subtree.into()
- } else {
- let spacing = match self.peek() {
- Some(next)
- if next.kind().is_trivia()
- || next.kind() == T!['[']
- || next.kind() == T!['{']
- || next.kind() == T!['('] =>
- {
- tt::Spacing::Alone
- }
- Some(next) if next.kind().is_punct() => tt::Spacing::Joint,
- _ => tt::Spacing::Alone,
- };
- let char = match token.to_char() {
- Some(c) => c,
- None => {
- panic!("Token from lexer must be single char: token = {:#?}", token);
- }
- };
- tt::Leaf::from(tt::Punct { char, spacing, id: self.id_alloc().alloc(range) }).into()
- }
- } else {
- macro_rules! make_leaf {
- ($i:ident) => {
- tt::$i { id: self.id_alloc().alloc(range), text: token.to_text() }.into()
- };
- }
- let leaf: tt::Leaf = match k {
- T![true] | T![false] => make_leaf!(Ident),
- IDENT => make_leaf!(Ident),
- k if k.is_keyword() => make_leaf!(Ident),
- k if k.is_literal() => make_leaf!(Literal),
- LIFETIME_IDENT => {
- let char_unit = TextSize::of('\'');
- let r = TextRange::at(range.start(), char_unit);
- let apostrophe = tt::Leaf::from(tt::Punct {
- char: '\'',
- spacing: tt::Spacing::Joint,
- id: self.id_alloc().alloc(r),
- });
- result.push(apostrophe.into());
-
- let r = TextRange::at(range.start() + char_unit, range.len() - char_unit);
- let ident = tt::Leaf::from(tt::Ident {
- text: SmolStr::new(&token.to_text()[1..]),
- id: self.id_alloc().alloc(r),
- });
- result.push(ident.into());
- return;
- }
- _ => return,
- };
-
- leaf.into()
- });
- }
-
fn convert_doc_comment(&self, token: &Self::Token) -> Option<Vec<tt::TokenTree>>;
fn bump(&mut self) -> Option<(Self::Token, TextRange)>;
}
}
fn to_text(&self) -> SmolStr {
- self.token().text().clone()
+ self.token().text().into()
}
}
fn peek(&self) -> Option<Self::Token> {
if let Some((punct, mut offset)) = self.punct_offset.clone() {
- offset = offset + TextSize::of('.');
+ offset += TextSize::of('.');
if usize::from(offset) < punct.text().len() {
return Some(SynToken::Punch(punct, offset));
}
}
fn finish(mut self) -> (Parse<SyntaxNode>, TokenMap) {
- self.token_map.entries.shrink_to_fit();
+ self.token_map.shrink_to_fit();
(self.inner.finish(), self.token_map)
}
}
}
}
};
- self.buf += &text;
+ self.buf += text;
self.text_pos += TextSize::of(text);
}
- let text = SmolStr::new(self.buf.as_str());
+ self.inner.token(kind, self.buf.as_str());
self.buf.clear();
- self.inner.token(kind, text);
-
// Add whitespace between adjoint puncts
let next = last.bump();
if let (
// Note: We always assume the semi-colon would be the last token in
// other parts of RA such that we don't add whitespace here.
if curr.spacing == tt::Spacing::Alone && curr.char != ';' {
- self.inner.token(WHITESPACE, " ".into());
+ self.inner.token(WHITESPACE, " ");
self.text_pos += TextSize::of(' ');
}
}
// - T!['}']
// - WHITE_SPACE
let token_tree = ast::TokenTree::cast(token_tree).unwrap();
- let tt = ast_to_token_tree(&token_tree).unwrap().0;
+ let tt = ast_to_token_tree(&token_tree).0;
assert_eq!(tt.delimiter_kind(), Some(tt::DelimiterKind::Brace));
}
fn test_token_tree_multi_char_punct() {
let source_file = ast::SourceFile::parse("struct Foo { a: x::Y }").ok().unwrap();
let struct_def = source_file.syntax().descendants().find_map(ast::Struct::cast).unwrap();
- let tt = ast_to_token_tree(&struct_def).unwrap().0;
+ let tt = ast_to_token_tree(&struct_def).0;
token_tree_to_syntax_node(&tt, FragmentKind::Item).unwrap();
}
}