rust/crates/ra_mbe/src/syntax_bridge.rs

430 lines
14 KiB
Rust
Raw Normal View History

//! FIXME: write short doc here
use ra_parser::{FragmentKind, ParseError, TreeSink};
2019-02-11 12:12:06 -06:00
use ra_syntax::{
2019-07-20 12:04:34 -05:00
ast, AstNode, AstToken, NodeOrToken, Parse, SmolStr, SyntaxKind, SyntaxKind::*, SyntaxNode,
2019-07-18 15:19:04 -05:00
SyntaxTreeBuilder, TextRange, TextUnit, T,
2019-02-11 12:12:06 -06:00
};
use tt::buffer::{Cursor, TokenBuffer};
2019-04-07 08:42:53 -05:00
2019-07-20 12:04:34 -05:00
use crate::subtree_source::SubtreeTokenSource;
use crate::ExpandError;
2019-02-11 12:31:54 -06:00
/// Maps `tt::TokenId` to the relative range of the original token.
#[derive(Debug, PartialEq, Eq, Default)]
2019-02-11 12:12:06 -06:00
pub struct TokenMap {
/// Maps `tt::TokenId` to the *relative* source range.
2019-02-26 11:59:32 -06:00
tokens: Vec<TextRange>,
2019-01-31 12:29:04 -06:00
}
2019-11-03 13:12:19 -06:00
/// Maps relative range of the expanded syntax node to `tt::TokenId`
2019-11-03 08:46:12 -06:00
#[derive(Debug, PartialEq, Eq, Default)]
2019-11-07 19:57:01 -06:00
pub struct RevTokenMap {
2019-11-08 22:00:46 -06:00
pub ranges: Vec<(TextRange, tt::TokenId)>,
2019-11-03 08:46:12 -06:00
}
2019-02-11 12:31:54 -06:00
/// Convert the syntax tree (what user has written) to a `TokenTree` (what macro
/// will consume).
2019-02-11 12:12:06 -06:00
pub fn ast_to_token_tree(ast: &ast::TokenTree) -> Option<(tt::Subtree, TokenMap)> {
let mut token_map = TokenMap::default();
let node = ast.syntax();
2019-07-20 04:58:27 -05:00
let tt = convert_tt(&mut token_map, node.text_range().start(), node)?;
2019-02-11 12:12:06 -06:00
Some((tt, token_map))
}
/// 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();
2019-07-20 04:58:27 -05:00
let tt = convert_tt(&mut token_map, node.text_range().start(), node)?;
Some((tt, token_map))
}
// The following items are what `rustc` macro can be parsed into :
// link: https://github.com/rust-lang/rust/blob/9ebf47851a357faa4cd97f4b1dc7835f6376e639/src/libsyntax/ext/expand.rs#L141
2019-04-18 14:49:56 -05:00
// * Expr(P<ast::Expr>) -> token_tree_to_expr
// * Pat(P<ast::Pat>) -> token_tree_to_pat
// * Ty(P<ast::Ty>) -> token_tree_to_ty
// * Stmts(SmallVec<[ast::Stmt; 1]>) -> token_tree_to_stmts
// * Items(SmallVec<[P<ast::Item>; 1]>) -> token_tree_to_items
//
// * TraitItems(SmallVec<[ast::TraitItem; 1]>)
// * ImplItems(SmallVec<[ast::ImplItem; 1]>)
// * ForeignItems(SmallVec<[ast::ForeignItem; 1]>
2019-05-25 08:55:46 -05:00
pub fn token_tree_to_syntax_node(
2019-09-02 10:51:03 -05:00
tt: &tt::Subtree,
fragment_kind: FragmentKind,
2019-11-07 19:57:01 -06:00
) -> Result<(Parse<SyntaxNode>, RevTokenMap), ExpandError> {
let tmp;
let tokens = match tt {
tt::Subtree { delimiter: tt::Delimiter::None, token_trees } => token_trees.as_slice(),
_ => {
tmp = [tt.clone().into()];
&tmp[..]
}
};
2019-09-02 10:51:03 -05:00
let buffer = TokenBuffer::new(&tokens);
let mut token_source = SubtreeTokenSource::new(&buffer);
2019-11-03 13:12:19 -06:00
let mut tree_sink = TtTreeSink::new(buffer.begin());
2019-09-02 10:51:03 -05:00
ra_parser::parse_fragment(&mut token_source, &mut tree_sink, fragment_kind);
if tree_sink.roots.len() != 1 {
return Err(ExpandError::ConversionError);
}
//FIXME: would be cool to report errors
2019-11-03 13:12:19 -06:00
let (parse, range_map) = tree_sink.finish();
2019-11-03 08:46:12 -06:00
Ok((parse, range_map))
2019-09-02 10:51:03 -05:00
}
2019-02-11 12:12:06 -06:00
impl TokenMap {
2019-11-16 07:49:26 -06:00
pub fn token_by_offset(&self, relative_offset: TextUnit) -> Option<tt::TokenId> {
let (idx, _) =
self.tokens.iter().enumerate().find(|(_, range)| range.contains(relative_offset))?;
Some(tt::TokenId(idx as u32))
}
2019-02-11 12:31:54 -06:00
pub fn relative_range_of(&self, tt: tt::TokenId) -> Option<TextRange> {
let idx = tt.0 as usize;
2019-07-04 12:26:44 -05:00
self.tokens.get(idx).copied()
2019-02-11 12:31:54 -06:00
}
2019-02-11 12:12:06 -06:00
fn alloc(&mut self, relative_range: TextRange) -> tt::TokenId {
2019-02-26 11:59:32 -06:00
let id = self.tokens.len();
self.tokens.push(relative_range);
tt::TokenId(id as u32)
2019-02-11 12:12:06 -06:00
}
}
2019-11-07 19:57:01 -06:00
impl RevTokenMap {
fn add(&mut self, relative_range: TextRange, token_id: tt::TokenId) {
2019-11-03 08:46:12 -06:00
self.ranges.push((relative_range, token_id.clone()))
}
}
2019-05-04 02:00:16 -05:00
/// Returns the textual content of a doc comment block as a quoted string
/// That is, strips leading `///` (or `/**`, etc)
/// and strips the ending `*/`
/// And then quote the string, which is needed to convert to `tt::Literal`
fn doc_comment_text(comment: &ast::Comment) -> SmolStr {
let prefix_len = comment.prefix().len();
let mut text = &comment.text()[prefix_len..];
// Remove ending "*/"
if comment.kind().shape == ast::CommentShape::Block {
text = &text[0..text.len() - 2];
}
// Quote the string
// Note that `tt::Literal` expect an escaped string
2019-05-04 06:34:02 -05:00
let text = format!("{:?}", text.escape_default().to_string());
2019-05-04 02:00:16 -05:00
text.into()
}
2019-07-18 12:09:50 -05:00
fn convert_doc_comment(token: &ra_syntax::SyntaxToken) -> Option<Vec<tt::TokenTree>> {
let comment = ast::Comment::cast(token.clone())?;
2019-05-04 02:00:16 -05:00
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));
// Make `#![]`
let mut token_trees = Vec::new();
token_trees.push(mk_punct('#'));
if let ast::CommentPlacement::Inner = doc {
token_trees.push(mk_punct('!'));
}
2019-06-03 09:21:08 -05:00
token_trees.push(tt::TokenTree::from(tt::Subtree {
delimiter: tt::Delimiter::Bracket,
token_trees: meta_tkns,
}));
2019-05-04 02:00:16 -05:00
return Some(token_trees);
// Helper functions
fn mk_ident(s: &str) -> tt::TokenTree {
tt::TokenTree::from(tt::Leaf::from(tt::Ident {
text: s.into(),
id: tt::TokenId::unspecified(),
}))
}
fn mk_punct(c: char) -> tt::TokenTree {
tt::TokenTree::from(tt::Leaf::from(tt::Punct { char: c, spacing: tt::Spacing::Alone }))
}
fn mk_doc_literal(comment: &ast::Comment) -> tt::TokenTree {
let lit = tt::Literal { text: doc_comment_text(comment) };
tt::TokenTree::from(tt::Leaf::from(lit))
}
}
2019-02-11 12:12:06 -06:00
fn convert_tt(
token_map: &mut TokenMap,
global_offset: TextUnit,
tt: &SyntaxNode,
) -> Option<tt::Subtree> {
2019-04-25 10:12:57 -05:00
// This tree is empty
if tt.first_child_or_token().is_none() {
return Some(tt::Subtree { token_trees: vec![], delimiter: tt::Delimiter::None });
}
2019-03-30 05:25:53 -05:00
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()) {
2019-05-15 07:35:47 -05:00
(T!['('], T![')']) => (tt::Delimiter::Parenthesis, true),
(T!['{'], T!['}']) => (tt::Delimiter::Brace, true),
(T!['['], T![']']) => (tt::Delimiter::Bracket, true),
_ => (tt::Delimiter::None, false),
2019-01-31 12:29:04 -06:00
};
2019-01-31 12:29:04 -06:00
let mut token_trees = Vec::new();
2019-04-24 10:01:32 -05:00
let mut child_iter = tt.children_with_tokens().skip(skip_first as usize).peekable();
while let Some(child) = child_iter.next() {
2019-05-04 02:00:16 -05:00
if skip_first && (child == first_child || child == last_child) {
2019-01-31 12:29:04 -06:00
continue;
}
2019-05-04 02:00:16 -05:00
2019-03-30 05:25:53 -05:00
match child {
2019-07-20 12:04:34 -05:00
NodeOrToken::Token(token) => {
2019-05-04 02:00:16 -05:00
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() {
2019-07-20 12:04:34 -05:00
Some(NodeOrToken::Token(token)) => {
if token.kind().is_punct() {
tt::Spacing::Joint
} else {
tt::Spacing::Alone
2019-04-24 10:01:32 -05:00
}
}
_ => tt::Spacing::Alone,
};
2019-04-24 10:01:32 -05:00
token_trees.push(tt::Leaf::from(tt::Punct { char, spacing }).into());
2019-03-30 05:25:53 -05:00
} else {
2019-05-15 07:35:47 -05:00
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
{
2019-07-20 04:58:27 -05:00
let relative_range = token.text_range() - global_offset;
2019-05-15 07:35:47 -05:00
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;
};
2019-03-30 05:25:53 -05:00
token_trees.push(child);
2019-01-31 12:29:04 -06:00
}
}
2019-07-20 12:04:34 -05:00
NodeOrToken::Node(node) => {
2019-07-18 12:09:50 -05:00
let child = convert_tt(token_map, global_offset, &node)?.into();
2019-03-30 05:25:53 -05:00
token_trees.push(child);
2019-01-31 12:29:04 -06:00
}
2019-03-30 05:25:53 -05:00
};
2019-01-31 12:29:04 -06:00
}
2019-02-08 05:49:43 -06:00
let res = tt::Subtree { delimiter, token_trees };
2019-01-31 12:29:04 -06:00
Some(res)
}
2019-02-23 05:18:32 -06:00
2019-05-27 09:56:21 -05:00
struct TtTreeSink<'a> {
2019-02-23 08:21:56 -06:00
buf: String,
2019-05-27 09:56:21 -05:00
cursor: Cursor<'a>,
2019-02-23 08:21:56 -06:00
text_pos: TextUnit,
inner: SyntaxTreeBuilder,
2019-11-07 19:57:01 -06:00
range_map: RevTokenMap,
// Number of roots
// Use for detect ill-form tree which is not single root
roots: smallvec::SmallVec<[usize; 1]>,
2019-02-23 08:21:56 -06:00
}
2019-05-27 09:56:21 -05:00
impl<'a> TtTreeSink<'a> {
2019-11-03 13:12:19 -06:00
fn new(cursor: Cursor<'a>) -> Self {
2019-02-23 08:21:56 -06:00
TtTreeSink {
buf: String::new(),
2019-05-27 09:56:21 -05:00
cursor,
2019-02-23 08:21:56 -06:00
text_pos: 0.into(),
inner: SyntaxTreeBuilder::default(),
roots: smallvec::SmallVec::new(),
2019-11-07 19:57:01 -06:00
range_map: RevTokenMap::default(),
2019-02-23 08:21:56 -06:00
}
}
2019-11-03 13:12:19 -06:00
2019-11-07 19:57:01 -06:00
fn finish(self) -> (Parse<SyntaxNode>, RevTokenMap) {
2019-11-03 13:12:19 -06:00
(self.inner.finish(), self.range_map)
}
2019-02-23 08:21:56 -06:00
}
2019-05-27 09:56:21 -05:00
fn delim_to_str(d: tt::Delimiter, closing: bool) -> SmolStr {
let texts = match d {
tt::Delimiter::Parenthesis => "()",
tt::Delimiter::Brace => "{}",
tt::Delimiter::Bracket => "[]",
tt::Delimiter::None => "",
};
let idx = closing as usize;
2019-06-04 01:28:22 -05:00
let text = if !texts.is_empty() { &texts[idx..texts.len() - (1 - idx)] } else { "" };
2019-05-27 09:56:21 -05:00
text.into()
}
2019-05-27 09:56:21 -05:00
impl<'a> TreeSink for TtTreeSink<'a> {
2019-03-30 05:25:53 -05:00
fn token(&mut self, kind: SyntaxKind, n_tokens: u8) {
if kind == L_DOLLAR || kind == R_DOLLAR {
2019-05-27 11:28:46 -05:00
self.cursor = self.cursor.bump_subtree();
return;
}
2019-02-23 08:21:56 -06:00
for _ in 0..n_tokens {
2019-05-27 09:56:21 -05:00
if self.cursor.eof() {
break;
}
match self.cursor.token_tree() {
Some(tt::TokenTree::Leaf(leaf)) => {
2019-11-03 11:45:58 -06:00
// Mark the range if needed
2019-11-03 08:46:12 -06:00
if let tt::Leaf::Ident(ident) = leaf {
2019-11-03 11:45:58 -06:00
if kind == IDENT {
2019-11-07 19:57:01 -06:00
let range =
TextRange::offset_len(self.text_pos, TextUnit::of_str(&ident.text));
self.range_map.add(range, ident.id);
2019-11-03 11:45:58 -06:00
}
2019-11-03 08:46:12 -06:00
}
2019-11-03 11:45:58 -06:00
self.cursor = self.cursor.bump();
self.buf += &format!("{}", leaf);
2019-05-27 09:56:21 -05:00
}
Some(tt::TokenTree::Subtree(subtree)) => {
self.cursor = self.cursor.subtree().unwrap();
self.buf += &delim_to_str(subtree.delimiter, false);
}
None => {
if let Some(parent) = self.cursor.end() {
self.cursor = self.cursor.bump();
self.buf += &delim_to_str(parent.delimiter, true);
}
}
};
2019-02-23 08:21:56 -06:00
}
2019-05-27 09:56:21 -05:00
2019-02-23 08:21:56 -06:00
self.text_pos += TextUnit::of_str(&self.buf);
let text = SmolStr::new(self.buf.as_str());
self.buf.clear();
2019-04-25 10:12:57 -05:00
self.inner.token(kind, text);
2019-05-27 09:56:21 -05:00
// Add whitespace between adjoint puncts
let next = self.cursor.bump();
if let (
Some(tt::TokenTree::Leaf(tt::Leaf::Punct(curr))),
Some(tt::TokenTree::Leaf(tt::Leaf::Punct(_))),
) = (self.cursor.token_tree(), next.token_tree())
{
if curr.spacing == tt::Spacing::Alone {
self.inner.token(WHITESPACE, " ".into());
2019-11-03 08:46:12 -06:00
self.text_pos += TextUnit::of_char(' ');
}
}
2019-02-23 08:21:56 -06:00
}
2019-03-30 05:25:53 -05:00
fn start_node(&mut self, kind: SyntaxKind) {
self.inner.start_node(kind);
match self.roots.last_mut() {
None | Some(0) => self.roots.push(1),
Some(ref mut n) => **n += 1,
};
2019-02-23 08:21:56 -06:00
}
2019-03-30 05:25:53 -05:00
fn finish_node(&mut self) {
self.inner.finish_node();
*self.roots.last_mut().unwrap() -= 1;
2019-02-23 08:21:56 -06:00
}
fn error(&mut self, error: ParseError) {
self.inner.error(error, self.text_pos)
}
}
#[cfg(test)]
mod tests {
use super::*;
use crate::tests::{create_rules, expand};
2019-05-27 09:56:21 -05:00
use ra_parser::TokenSource;
#[test]
fn convert_tt_token_source() {
let rules = create_rules(
r#"
macro_rules! literals {
($i:ident) => {
{
let a = 'c';
let c = 1000;
let f = 12E+99_f64;
let s = "rust1";
}
}
}
"#,
);
2019-05-28 10:46:11 -05:00
let expansion = expand(&rules, "literals!(foo);");
2019-10-30 12:36:37 -05:00
let tts = &[expansion.into()];
2019-06-02 11:54:33 -05:00
let buffer = tt::buffer::TokenBuffer::new(tts);
2019-05-27 09:56:21 -05:00
let mut tt_src = SubtreeTokenSource::new(&buffer);
let mut tokens = vec![];
while tt_src.current().kind != EOF {
tokens.push((tt_src.current().kind, tt_src.text()));
tt_src.bump();
}
// [${]
// [let] [a] [=] ['c'] [;]
2019-05-27 09:56:21 -05:00
assert_eq!(tokens[2 + 3].1, "'c'");
assert_eq!(tokens[2 + 3].0, CHAR);
// [let] [c] [=] [1000] [;]
2019-05-27 09:56:21 -05:00
assert_eq!(tokens[2 + 5 + 3].1, "1000");
assert_eq!(tokens[2 + 5 + 3].0, INT_NUMBER);
// [let] [f] [=] [12E+99_f64] [;]
2019-05-27 09:56:21 -05:00
assert_eq!(tokens[2 + 10 + 3].1, "12E+99_f64");
assert_eq!(tokens[2 + 10 + 3].0, FLOAT_NUMBER);
// [let] [s] [=] ["rust1"] [;]
2019-05-27 09:56:21 -05:00
assert_eq!(tokens[2 + 15 + 3].1, "\"rust1\"");
assert_eq!(tokens[2 + 15 + 3].0, STRING);
}
#[test]
fn stmts_token_trees_to_expr_is_err() {
let rules = create_rules(
r#"
macro_rules! stmts {
() => {
let a = 0;
let b = 0;
let c = 0;
let d = 0;
}
}
"#,
);
2019-05-28 10:46:11 -05:00
let expansion = expand(&rules, "stmts!();");
assert!(token_tree_to_syntax_node(&expansion, FragmentKind::Expr).is_err());
}
}