rust/crates/ra_mbe/src/syntax_bridge.rs

232 lines
7.6 KiB
Rust
Raw Normal View History

2019-02-23 08:21:56 -06:00
use ra_parser::{TokenSource, TreeSink, ParseError};
2019-02-11 12:12:06 -06:00
use ra_syntax::{
2019-02-23 08:21:56 -06:00
AstNode, SyntaxNode, TextRange, SyntaxKind, SmolStr, SyntaxTreeBuilder, TreeArc,
2019-02-11 12:12:06 -06:00
ast, SyntaxKind::*, TextUnit
};
2019-01-31 12:29:04 -06:00
2019-02-11 12:31:54 -06:00
/// Maps `tt::TokenId` to the relative range of the original token.
2019-02-11 12:12:06 -06:00
#[derive(Default)]
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
}
2019-02-11 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();
let tt = convert_tt(&mut token_map, node.range().start(), node)?;
Some((tt, token_map))
}
2019-02-23 05:06:10 -06:00
/// Parses the token tree (result of macro expansion) as a sequence of items
2019-02-23 08:21:56 -06:00
pub fn token_tree_to_ast_item_list(tt: &tt::Subtree) -> TreeArc<ast::SourceFile> {
let token_source = TtTokenSource::new(tt);
let mut tree_sink = TtTreeSink::new(&token_source.tokens);
ra_parser::parse(&token_source, &mut tree_sink);
let syntax = tree_sink.inner.finish();
ast::SourceFile::cast(&syntax).unwrap().to_owned()
2019-02-23 05:06:10 -06:00
}
2019-02-11 12:12:06 -06:00
impl TokenMap {
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;
2019-02-26 11:59:32 -06:00
self.tokens.get(idx).map(|&it| it)
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);
2019-02-11 12:12:06 -06:00
tt::TokenId(id as u32)
}
}
fn convert_tt(
token_map: &mut TokenMap,
global_offset: TextUnit,
tt: &SyntaxNode,
) -> Option<tt::Subtree> {
2019-01-31 12:29:04 -06:00
let first_child = tt.first_child()?;
let last_child = tt.last_child()?;
let delimiter = match (first_child.kind(), last_child.kind()) {
(L_PAREN, R_PAREN) => tt::Delimiter::Parenthesis,
(L_CURLY, R_CURLY) => tt::Delimiter::Brace,
(L_BRACK, R_BRACK) => tt::Delimiter::Bracket,
_ => return None,
};
let mut token_trees = Vec::new();
for child in tt.children().skip(1) {
if child == first_child || child == last_child || child.kind().is_trivia() {
continue;
}
if child.kind().is_punct() {
let mut prev = None;
for char in child.leaf_text().unwrap().chars() {
if let Some(char) = prev {
token_trees.push(
2019-02-08 05:49:43 -06:00
tt::Leaf::from(tt::Punct { char, spacing: tt::Spacing::Joint }).into(),
2019-01-31 12:29:04 -06:00
);
}
prev = Some(char)
}
if let Some(char) = prev {
2019-02-08 05:49:43 -06:00
token_trees
.push(tt::Leaf::from(tt::Punct { char, spacing: tt::Spacing::Alone }).into());
2019-01-31 12:29:04 -06:00
}
} else {
let child: tt::TokenTree = if child.kind() == TOKEN_TREE {
2019-02-11 12:12:06 -06:00
convert_tt(token_map, global_offset, child)?.into()
2019-01-31 12:29:04 -06:00
} else if child.kind().is_keyword() || child.kind() == IDENT {
2019-02-11 12:12:06 -06:00
let relative_range = child.range() - global_offset;
let id = token_map.alloc(relative_range);
2019-01-31 12:29:04 -06:00
let text = child.leaf_text().unwrap().clone();
2019-02-11 12:12:06 -06:00
tt::Leaf::from(tt::Ident { text, id }).into()
2019-01-31 12:29:04 -06:00
} else if child.kind().is_literal() {
2019-02-08 05:49:43 -06:00
tt::Leaf::from(tt::Literal { text: child.leaf_text().unwrap().clone() }).into()
2019-01-31 12:29:04 -06:00
} else {
return None;
};
token_trees.push(child)
}
}
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
2019-02-23 06:29:06 -06:00
struct TtTokenSource {
2019-02-24 04:14:14 -06:00
tokens: Vec<TtToken>,
2019-02-23 06:29:06 -06:00
}
2019-02-24 04:14:14 -06:00
struct TtToken {
2019-02-23 06:29:06 -06:00
kind: SyntaxKind,
is_joint_to_next: bool,
2019-02-23 06:43:45 -06:00
text: SmolStr,
2019-02-23 06:29:06 -06:00
}
2019-02-23 05:18:32 -06:00
impl TtTokenSource {
fn new(tt: &tt::Subtree) -> TtTokenSource {
2019-02-23 06:29:06 -06:00
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);
sub.token_trees.iter().for_each(|tt| self.convert_tt(tt));
self.push_delim(sub.delimiter, true)
}
fn convert_tt(&mut self, tt: &tt::TokenTree) {
match tt {
tt::TokenTree::Leaf(leaf) => self.convert_leaf(leaf),
tt::TokenTree::Subtree(sub) => self.convert_subtree(sub),
}
}
fn convert_leaf(&mut self, leaf: &tt::Leaf) {
let tok = match leaf {
2019-02-24 04:14:14 -06:00
tt::Leaf::Literal(l) => TtToken {
2019-02-23 06:29:06 -06:00
kind: SyntaxKind::INT_NUMBER, // FIXME
is_joint_to_next: false,
2019-02-23 06:43:45 -06:00
text: l.text.clone(),
2019-02-23 06:29:06 -06:00
},
2019-02-23 08:51:23 -06:00
tt::Leaf::Punct(p) => {
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 = {
2019-02-23 06:43:45 -06:00
let mut buf = [0u8; 4];
let s: &str = p.char.encode_utf8(&mut buf);
SmolStr::new(s)
2019-02-23 08:51:23 -06:00
};
2019-02-24 04:14:14 -06:00
TtToken { kind, is_joint_to_next: p.spacing == tt::Spacing::Joint, text }
2019-02-23 08:51:23 -06:00
}
2019-02-23 06:29:06 -06:00
tt::Leaf::Ident(ident) => {
2019-02-23 08:51:23 -06:00
let kind = SyntaxKind::from_keyword(ident.text.as_str()).unwrap_or(IDENT);
2019-02-24 04:14:14 -06:00
TtToken { kind, is_joint_to_next: false, text: ident.text.clone() }
2019-02-23 06:29:06 -06:00
}
};
self.tokens.push(tok)
}
fn push_delim(&mut self, d: tt::Delimiter, closing: bool) {
2019-02-23 06:43:45 -06:00
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], "[]"),
2019-02-23 06:29:06 -06:00
tt::Delimiter::None => return,
};
2019-02-23 06:43:45 -06:00
let idx = closing as usize;
let kind = kinds[idx];
let text = &texts[idx..texts.len() - (1 - idx)];
2019-02-24 04:14:14 -06:00
let tok = TtToken { kind, is_joint_to_next: false, text: SmolStr::new(text) };
2019-02-23 06:29:06 -06:00
self.tokens.push(tok)
2019-02-23 05:18:32 -06:00
}
}
impl TokenSource for TtTokenSource {
fn token_kind(&self, pos: usize) -> SyntaxKind {
2019-02-23 08:51:23 -06:00
if let Some(tok) = self.tokens.get(pos) {
tok.kind
} else {
SyntaxKind::EOF
}
2019-02-23 05:18:32 -06:00
}
fn is_token_joint_to_next(&self, pos: usize) -> bool {
2019-02-23 06:29:06 -06:00
self.tokens[pos].is_joint_to_next
2019-02-23 05:18:32 -06:00
}
fn is_keyword(&self, pos: usize, kw: &str) -> bool {
2019-02-23 06:43:45 -06:00
self.tokens[pos].text == *kw
2019-02-23 05:18:32 -06:00
}
}
2019-02-23 08:21:56 -06:00
#[derive(Default)]
struct TtTreeSink<'a> {
buf: String,
2019-02-24 04:14:14 -06:00
tokens: &'a [TtToken],
2019-02-23 08:21:56 -06:00
text_pos: TextUnit,
token_pos: usize,
inner: SyntaxTreeBuilder,
}
impl<'a> TtTreeSink<'a> {
2019-02-24 04:14:14 -06:00
fn new(tokens: &'a [TtToken]) -> TtTreeSink {
2019-02-23 08:21:56 -06:00
TtTreeSink {
buf: String::new(),
tokens,
text_pos: 0.into(),
token_pos: 0,
inner: SyntaxTreeBuilder::default(),
}
}
}
impl<'a> TreeSink for TtTreeSink<'a> {
fn leaf(&mut self, kind: SyntaxKind, n_tokens: u8) {
for _ in 0..n_tokens {
self.buf += self.tokens[self.token_pos].text.as_str();
self.token_pos += 1;
}
self.text_pos += TextUnit::of_str(&self.buf);
let text = SmolStr::new(self.buf.as_str());
self.buf.clear();
self.inner.leaf(kind, text)
}
fn start_branch(&mut self, kind: SyntaxKind) {
self.inner.start_branch(kind);
}
fn finish_branch(&mut self) {
self.inner.finish_branch();
}
fn error(&mut self, error: ParseError) {
self.inner.error(error, self.text_pos)
}
}