rust-analyzer/crates/ra_mbe/src/subtree_source.rs

522 lines
16 KiB
Rust
Raw Normal View History

2019-04-07 13:42:53 +00:00
use ra_parser::{TokenSource};
use ra_syntax::{classify_literal, SmolStr, SyntaxKind, SyntaxKind::*};
use std::cell::{RefCell};
2019-04-07 13:42:53 +00:00
#[derive(Debug, Clone, Eq, PartialEq)]
2019-04-07 13:42:53 +00:00
struct TtToken {
pub kind: SyntaxKind,
pub is_joint_to_next: bool,
pub text: SmolStr,
pub n_tokens: usize,
}
#[derive(Debug, Clone, Eq, PartialEq)]
2019-04-08 12:32:21 +00:00
enum WalkCursor {
DelimiterBegin(Option<TtToken>),
Token(usize, Option<TtToken>),
DelimiterEnd(Option<TtToken>),
Eof,
2019-04-07 13:42:53 +00:00
}
2019-04-08 10:21:48 +00:00
#[derive(Debug)]
struct SubTreeWalker<'a> {
pos: usize,
stack: Vec<(&'a tt::Subtree, Option<usize>)>,
2019-04-08 12:32:21 +00:00
cursor: WalkCursor,
2019-04-08 10:21:48 +00:00
last_steps: Vec<usize>,
subtree: &'a tt::Subtree,
}
impl<'a> SubTreeWalker<'a> {
fn new(subtree: &tt::Subtree) -> SubTreeWalker {
let mut res = SubTreeWalker {
pos: 0,
stack: vec![],
2019-04-08 12:32:21 +00:00
cursor: WalkCursor::Eof,
last_steps: vec![],
subtree,
};
res.reset();
res
}
2019-04-08 12:32:21 +00:00
fn is_eof(&self) -> bool {
self.cursor == WalkCursor::Eof
}
fn reset(&mut self) {
self.pos = 0;
self.stack = vec![(self.subtree, None)];
2019-04-08 12:32:21 +00:00
self.cursor = WalkCursor::DelimiterBegin(convert_delim(self.subtree.delimiter, false));
self.last_steps = vec![];
while self.is_empty_delimiter() {
self.forward_unchecked();
}
}
2019-04-08 12:32:21 +00:00
// This funciton will fast forward the cursor,
// Such that backward will stop at `start_pos` point
fn start_from_nth(&mut self, start_pos: usize) {
self.reset();
self.pos = start_pos;
2019-04-08 12:32:21 +00:00
self.cursor = self.walk_token(start_pos, 0, false);
while self.is_empty_delimiter() {
self.forward_unchecked();
}
}
fn current(&self) -> Option<&TtToken> {
2019-04-08 12:32:21 +00:00
match &self.cursor {
WalkCursor::DelimiterBegin(t) => t.as_ref(),
WalkCursor::Token(_, t) => t.as_ref(),
WalkCursor::DelimiterEnd(t) => t.as_ref(),
WalkCursor::Eof => None,
}
}
fn is_empty_delimiter(&self) -> bool {
2019-04-08 12:32:21 +00:00
match &self.cursor {
WalkCursor::DelimiterBegin(None) => true,
WalkCursor::DelimiterEnd(None) => true,
_ => false,
}
}
2019-04-08 12:32:21 +00:00
/// Move cursor backward by 1 step with empty checking
fn backward(&mut self) {
if self.last_steps.is_empty() {
return;
}
self.pos -= 1;
loop {
self.backward_unchecked();
// Skip Empty delimiter
if self.last_steps.is_empty() || !self.is_empty_delimiter() {
break;
}
}
2019-04-08 10:21:48 +00:00
2019-04-08 12:32:21 +00:00
// Move forward if it is empty delimiter
2019-04-08 10:21:48 +00:00
if self.last_steps.is_empty() {
while self.is_empty_delimiter() {
self.forward_unchecked();
}
}
}
2019-04-08 12:32:21 +00:00
/// Move cursor backward by 1 step without empty check
///
/// Depends on the current state of cursor:
///
/// * Delimiter Begin => Pop the stack, goto last walking token (`walk_token`)
/// * Token => Goto prev token (`walk_token`)
/// * Delimiter End => Goto the last child token (`walk_token`)
/// * Eof => push the root subtree, and set it as Delimiter End
fn backward_unchecked(&mut self) {
if self.last_steps.is_empty() {
return;
}
let last_step = self.last_steps.pop().unwrap();
2019-04-08 12:32:21 +00:00
let do_walk_token = match self.cursor {
WalkCursor::DelimiterBegin(_) => None,
WalkCursor::Token(u, _) => Some(u),
WalkCursor::DelimiterEnd(_) => {
let (top, _) = self.stack.last().unwrap();
Some(top.token_trees.len())
}
2019-04-08 12:32:21 +00:00
WalkCursor::Eof => None,
};
2019-04-08 12:32:21 +00:00
self.cursor = match do_walk_token {
Some(u) => self.walk_token(u, last_step, true),
None => match self.cursor {
WalkCursor::Eof => {
self.stack.push((self.subtree, None));
2019-04-08 12:32:21 +00:00
WalkCursor::DelimiterEnd(convert_delim(
self.stack.last().unwrap().0.delimiter,
true,
))
}
_ => {
2019-04-08 12:32:21 +00:00
let (_, last_top_cursor) = self.stack.pop().unwrap();
assert!(!self.stack.is_empty());
2019-04-08 12:32:21 +00:00
self.walk_token(last_top_cursor.unwrap(), last_step, true)
}
},
};
}
2019-04-08 12:32:21 +00:00
/// Move cursor forward by 1 step with empty checking
fn forward(&mut self) {
2019-04-08 12:32:21 +00:00
if self.is_eof() {
2019-04-08 10:21:48 +00:00
return;
}
self.pos += 1;
loop {
self.forward_unchecked();
if !self.is_empty_delimiter() {
break;
}
}
}
2019-04-08 12:32:21 +00:00
/// Move cursor forward by 1 step without empty checking
///
/// Depends on the current state of cursor:
///
/// * Delimiter Begin => Goto the first child token (`walk_token`)
/// * Token => Goto next token (`walk_token`)
/// * Delimiter End => Pop the stack, goto last walking token (`walk_token`)
///
fn forward_unchecked(&mut self) {
2019-04-08 12:32:21 +00:00
if self.is_eof() {
return;
}
let step = self.current().map(|x| x.n_tokens).unwrap_or(1);
self.last_steps.push(step);
2019-04-08 12:32:21 +00:00
let do_walk_token = match self.cursor {
WalkCursor::DelimiterBegin(_) => Some((0, 0)),
WalkCursor::Token(u, _) => Some((u, step)),
WalkCursor::DelimiterEnd(_) => None,
_ => unreachable!(),
};
2019-04-08 12:32:21 +00:00
self.cursor = match do_walk_token {
Some((u, step)) => self.walk_token(u, step, false),
None => {
let (_, last_top_idx) = self.stack.pop().unwrap();
match self.stack.last() {
2019-04-08 12:32:21 +00:00
Some(_) => self.walk_token(last_top_idx.unwrap(), 1, false),
None => WalkCursor::Eof,
}
}
};
}
2019-04-08 12:32:21 +00:00
/// Traversal child token
/// Depends on the new position, it returns:
///
/// * new position < 0 => DelimiterBegin
/// * new position > token_tree.len() => DelimiterEnd
/// * if new position is a subtree, depends on traversal direction:
/// ** backward => DelimiterEnd
/// ** forward => DelimiterBegin
/// * if new psoition is a leaf, return walk_leaf()
fn walk_token(&mut self, pos: usize, offset: usize, backward: bool) -> WalkCursor {
let (top, _) = self.stack.last().unwrap();
2019-04-08 12:32:21 +00:00
if backward && pos < offset {
return WalkCursor::DelimiterBegin(convert_delim(
self.stack.last().unwrap().0.delimiter,
false,
));
}
if !backward && pos + offset >= top.token_trees.len() {
return WalkCursor::DelimiterEnd(convert_delim(
self.stack.last().unwrap().0.delimiter,
true,
));
}
let pos = if backward { pos - offset } else { pos + offset };
match &top.token_trees[pos] {
tt::TokenTree::Subtree(subtree) => {
self.stack.push((subtree, Some(pos)));
let delim = convert_delim(self.stack.last().unwrap().0.delimiter, backward);
if backward {
2019-04-08 12:32:21 +00:00
WalkCursor::DelimiterEnd(delim)
} else {
2019-04-08 12:32:21 +00:00
WalkCursor::DelimiterBegin(delim)
}
}
2019-04-08 12:32:21 +00:00
tt::TokenTree::Leaf(leaf) => WalkCursor::Token(pos, Some(self.walk_leaf(leaf, pos))),
}
}
fn walk_leaf(&mut self, leaf: &tt::Leaf, pos: usize) -> TtToken {
match leaf {
tt::Leaf::Literal(l) => convert_literal(l),
tt::Leaf::Ident(ident) => convert_ident(ident),
tt::Leaf::Punct(punct) => {
let (top, _) = self.stack.last().unwrap();
convert_punct(punct, top, pos)
}
}
2019-04-07 13:42:53 +00:00
}
}
pub(crate) trait Querier {
fn token(&self, uidx: usize) -> (SyntaxKind, SmolStr);
2019-04-07 13:42:53 +00:00
}
// A wrapper class for ref cell
2019-04-08 10:21:48 +00:00
#[derive(Debug)]
pub(crate) struct WalkerOwner<'a> {
walker: RefCell<SubTreeWalker<'a>>,
offset: usize,
}
impl<'a> WalkerOwner<'a> {
2019-04-08 12:32:21 +00:00
fn new(subtree: &'a tt::Subtree) -> Self {
WalkerOwner { walker: RefCell::new(SubTreeWalker::new(subtree)), offset: 0 }
}
fn get<'b>(&self, pos: usize) -> Option<TtToken> {
self.set_walker_pos(pos);
2019-04-08 10:21:48 +00:00
let walker = self.walker.borrow();
2019-04-08 11:21:07 +00:00
walker.current().cloned()
2019-04-07 13:42:53 +00:00
}
fn start_from_nth(&mut self, pos: usize) {
self.offset = pos;
self.walker.borrow_mut().start_from_nth(pos);
}
2019-04-07 13:42:53 +00:00
fn set_walker_pos(&self, mut pos: usize) {
pos += self.offset;
let mut walker = self.walker.borrow_mut();
2019-04-08 12:32:21 +00:00
while pos > walker.pos && !walker.is_eof() {
walker.forward();
}
while pos < walker.pos {
walker.backward();
2019-04-07 13:42:53 +00:00
}
}
2019-04-08 12:32:21 +00:00
fn collect_token_trees(&mut self, n: usize) -> Vec<&tt::TokenTree> {
self.start_from_nth(self.offset);
2019-04-07 13:42:53 +00:00
let mut res = vec![];
let mut walker = self.walker.borrow_mut();
while walker.pos - self.offset < n {
2019-04-08 12:32:21 +00:00
if let WalkCursor::Token(u, tt) = &walker.cursor {
if walker.stack.len() == 1 {
// We only collect the topmost child
res.push(&walker.stack[0].0.token_trees[*u]);
if let Some(tt) = tt {
for i in 0..tt.n_tokens - 1 {
res.push(&walker.stack[0].0.token_trees[u + i]);
}
}
}
}
walker.forward();
2019-04-07 13:42:53 +00:00
}
res
}
}
impl<'a> Querier for WalkerOwner<'a> {
fn token(&self, uidx: usize) -> (SyntaxKind, SmolStr) {
2019-04-08 12:32:21 +00:00
let tkn = self.get(uidx).unwrap();
(tkn.kind, tkn.text)
}
}
pub(crate) struct SubtreeTokenSource<'a> {
walker: WalkerOwner<'a>,
}
impl<'a> SubtreeTokenSource<'a> {
pub fn new(subtree: &tt::Subtree) -> SubtreeTokenSource {
SubtreeTokenSource { walker: WalkerOwner::new(subtree) }
}
pub fn start_from_nth(&mut self, n: usize) {
self.walker.start_from_nth(n);
}
pub fn querier<'b>(&'a self) -> &'b WalkerOwner<'a>
where
'a: 'b,
{
&self.walker
}
2019-04-08 12:32:21 +00:00
pub(crate) fn bump_n(&mut self, parsed_tokens: usize) -> Vec<&tt::TokenTree> {
let res = self.walker.collect_token_trees(parsed_tokens);
res
2019-04-07 13:42:53 +00:00
}
}
impl<'a> TokenSource for SubtreeTokenSource<'a> {
fn token_kind(&self, pos: usize) -> SyntaxKind {
2019-04-08 12:32:21 +00:00
if let Some(tok) = self.walker.get(pos) {
2019-04-07 13:42:53 +00:00
tok.kind
} else {
SyntaxKind::EOF
}
}
fn is_token_joint_to_next(&self, pos: usize) -> bool {
2019-04-08 12:32:21 +00:00
self.walker.get(pos).unwrap().is_joint_to_next
2019-04-07 13:42:53 +00:00
}
fn is_keyword(&self, pos: usize, kw: &str) -> bool {
2019-04-08 12:32:21 +00:00
self.walker.get(pos).unwrap().text == *kw
2019-04-07 13:42:53 +00:00
}
}
struct TokenPeek<'a, I>
where
I: Iterator<Item = &'a tt::TokenTree>,
{
iter: itertools::MultiPeek<I>,
}
// helper function
fn to_punct(tt: &tt::TokenTree) -> Option<&tt::Punct> {
if let tt::TokenTree::Leaf(tt::Leaf::Punct(pp)) = tt {
return Some(pp);
}
None
}
impl<'a, I> TokenPeek<'a, I>
where
I: Iterator<Item = &'a tt::TokenTree>,
{
pub fn new(iter: I) -> Self {
TokenPeek { iter: itertools::multipeek(iter) }
}
fn current_punct2(&mut self, p: &tt::Punct) -> Option<((char, char), bool)> {
if p.spacing != tt::Spacing::Joint {
return None;
}
self.iter.reset_peek();
let p1 = to_punct(self.iter.peek()?)?;
Some(((p.char, p1.char), p1.spacing == tt::Spacing::Joint))
}
fn current_punct3(&mut self, p: &tt::Punct) -> Option<((char, char, char), bool)> {
self.current_punct2(p).and_then(|((p0, p1), last_joint)| {
if !last_joint {
None
} else {
let p2 = to_punct(*self.iter.peek()?)?;
Some(((p0, p1, p2.char), p2.spacing == tt::Spacing::Joint))
}
})
}
}
fn convert_multi_char_punct<'b, I>(
p: &tt::Punct,
iter: &mut TokenPeek<'b, I>,
) -> Option<(SyntaxKind, bool, &'static str, usize)>
where
I: Iterator<Item = &'b tt::TokenTree>,
{
if let Some((m, is_joint_to_next)) = iter.current_punct3(p) {
if let Some((kind, text)) = match m {
('<', '<', '=') => Some((SHLEQ, "<<=")),
('>', '>', '=') => Some((SHREQ, ">>=")),
('.', '.', '.') => Some((DOTDOTDOT, "...")),
('.', '.', '=') => Some((DOTDOTEQ, "..=")),
_ => None,
} {
return Some((kind, is_joint_to_next, text, 3));
2019-04-07 13:42:53 +00:00
}
}
if let Some((m, is_joint_to_next)) = iter.current_punct2(p) {
if let Some((kind, text)) = match m {
('<', '<') => Some((SHL, "<<")),
('>', '>') => Some((SHR, ">>")),
('|', '|') => Some((PIPEPIPE, "||")),
('&', '&') => Some((AMPAMP, "&&")),
('%', '=') => Some((PERCENTEQ, "%=")),
('*', '=') => Some((STAREQ, "*=")),
('/', '=') => Some((SLASHEQ, "/=")),
('^', '=') => Some((CARETEQ, "^=")),
('&', '=') => Some((AMPEQ, "&=")),
('|', '=') => Some((PIPEEQ, "|=")),
('-', '=') => Some((MINUSEQ, "-=")),
('+', '=') => Some((PLUSEQ, "+=")),
('>', '=') => Some((GTEQ, ">=")),
('<', '=') => Some((LTEQ, "<=")),
('-', '>') => Some((THIN_ARROW, "->")),
('!', '=') => Some((NEQ, "!=")),
('=', '>') => Some((FAT_ARROW, "=>")),
('=', '=') => Some((EQEQ, "==")),
('.', '.') => Some((DOTDOT, "..")),
(':', ':') => Some((COLONCOLON, "::")),
_ => None,
} {
return Some((kind, is_joint_to_next, text, 2));
2019-04-07 13:42:53 +00:00
}
}
None
}
2019-04-07 13:42:53 +00:00
fn convert_delim(d: tt::Delimiter, closing: bool) -> Option<TtToken> {
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], "[]"),
tt::Delimiter::None => return None,
};
let idx = closing as usize;
let kind = kinds[idx];
let text = &texts[idx..texts.len() - (1 - idx)];
Some(TtToken { kind, is_joint_to_next: false, text: SmolStr::new(text), n_tokens: 1 })
}
2019-04-07 13:42:53 +00:00
fn convert_literal(l: &tt::Literal) -> TtToken {
TtToken {
kind: classify_literal(&l.text).unwrap().kind,
is_joint_to_next: false,
text: l.text.clone(),
n_tokens: 1,
2019-04-07 13:42:53 +00:00
}
}
2019-04-07 13:42:53 +00:00
fn convert_ident(ident: &tt::Ident) -> TtToken {
let kind = SyntaxKind::from_keyword(ident.text.as_str()).unwrap_or(IDENT);
TtToken { kind, is_joint_to_next: false, text: ident.text.clone(), n_tokens: 1 }
}
2019-04-07 13:42:53 +00:00
fn convert_punct(p: &tt::Punct, parent: &tt::Subtree, next: usize) -> TtToken {
2019-04-08 11:21:07 +00:00
let iter = parent.token_trees[next + 1..].iter();
let mut peek = TokenPeek::new(iter);
if let Some((kind, is_joint_to_next, text, size)) = convert_multi_char_punct(p, &mut peek) {
TtToken { kind, is_joint_to_next, text: text.into(), n_tokens: size }
} else {
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 = {
let mut buf = [0u8; 4];
let s: &str = p.char.encode_utf8(&mut buf);
SmolStr::new(s)
};
TtToken { kind, is_joint_to_next: p.spacing == tt::Spacing::Joint, text, n_tokens: 1 }
2019-04-07 13:42:53 +00:00
}
}