rust/crates/ra_syntax/src/parsing/reparsing.rs

359 lines
8.1 KiB
Rust
Raw Normal View History

2019-02-21 06:24:42 -06:00
//! Implementation of incremental re-parsing.
//!
//! We use two simple strategies for this:
//! - if the edit modifies only a single token (like changing an identifier's
//! letter), we replace only this token.
//! - otherwise, we search for the nearest `{}` block which contains the edit
//! and try to parse only this block.
2019-02-21 04:37:32 -06:00
use ra_text_edit::AtomTextEdit;
use ra_parser::Reparser;
use crate::{
2019-02-21 06:24:42 -06:00
SyntaxKind::*, TextRange, TextUnit, SyntaxError,
algo,
2019-02-20 07:16:14 -06:00
syntax_node::{GreenNode, SyntaxNode},
parsing::{
2019-02-23 07:07:29 -06:00
text_token_source::TextTokenSource,
text_tree_sink::TextTreeSink,
lexer::{tokenize, Token},
}
};
pub(crate) fn incremental_reparse(
2019-01-07 07:15:47 -06:00
node: &SyntaxNode,
edit: &AtomTextEdit,
errors: Vec<SyntaxError>,
) -> Option<(GreenNode, Vec<SyntaxError>)> {
let (node, green, new_errors) =
reparse_leaf(node, &edit).or_else(|| reparse_block(node, &edit))?;
let green_root = node.replace_with(green);
let errors = merge_errors(errors, new_errors, node, edit);
Some((green_root, errors))
}
fn reparse_leaf<'node>(
2019-01-07 07:15:47 -06:00
node: &'node SyntaxNode,
edit: &AtomTextEdit,
2019-01-07 07:15:47 -06:00
) -> Option<(&'node SyntaxNode, GreenNode, Vec<SyntaxError>)> {
let node = algo::find_covering_node(node, edit.delete);
match node.kind() {
2018-10-31 16:38:18 -05:00
WHITESPACE | COMMENT | IDENT | STRING | RAW_STRING => {
let text = get_text_after_edit(node, &edit);
let tokens = tokenize(&text);
let token = match tokens[..] {
[token] if token.kind == node.kind() => token,
_ => return None,
};
if token.kind == IDENT && is_contextual_kw(&text) {
return None;
}
2018-10-02 09:07:12 -05:00
let green = GreenNode::new_leaf(node.kind(), text.into());
let new_errors = vec![];
Some((node, green, new_errors))
}
_ => None,
}
}
fn reparse_block<'node>(
2019-01-07 07:15:47 -06:00
node: &'node SyntaxNode,
edit: &AtomTextEdit,
2019-01-07 07:15:47 -06:00
) -> Option<(&'node SyntaxNode, GreenNode, Vec<SyntaxError>)> {
let (node, reparser) = find_reparsable_node(node, edit.delete)?;
let text = get_text_after_edit(node, &edit);
let tokens = tokenize(&text);
if !is_balanced(&tokens) {
return None;
}
2019-02-23 07:07:29 -06:00
let token_source = TextTokenSource::new(&text, &tokens);
let mut tree_sink = TextTreeSink::new(&text, &tokens);
2019-02-21 04:37:32 -06:00
reparser.parse(&token_source, &mut tree_sink);
let (green, new_errors) = tree_sink.finish();
Some((node, green, new_errors))
}
2019-01-07 07:15:47 -06:00
fn get_text_after_edit(node: &SyntaxNode, edit: &AtomTextEdit) -> String {
2019-01-08 12:59:55 -06:00
let edit = AtomTextEdit::replace(edit.delete - node.range().start(), edit.insert.clone());
edit.apply(node.text().to_string())
}
fn is_contextual_kw(text: &str) -> bool {
match text {
"auto" | "default" | "union" => true,
_ => false,
}
}
2019-02-21 04:37:32 -06:00
fn find_reparsable_node(node: &SyntaxNode, range: TextRange) -> Option<(&SyntaxNode, Reparser)> {
let node = algo::find_covering_node(node, range);
node.ancestors().find_map(|node| {
let first_child = node.first_child().map(|it| it.kind());
let parent = node.parent().map(|it| it.kind());
2019-02-21 04:37:32 -06:00
Reparser::for_node(node.kind(), first_child, parent).map(|r| (node, r))
})
}
fn is_balanced(tokens: &[Token]) -> bool {
2018-10-17 18:25:37 -05:00
if tokens.is_empty()
|| tokens.first().unwrap().kind != L_CURLY
|| tokens.last().unwrap().kind != R_CURLY
{
return false;
}
let mut balance = 0usize;
for t in tokens.iter() {
match t.kind {
L_CURLY => balance += 1,
R_CURLY => {
balance = match balance.checked_sub(1) {
Some(b) => b,
None => return false,
}
}
_ => (),
}
}
balance == 0
}
fn merge_errors(
old_errors: Vec<SyntaxError>,
new_errors: Vec<SyntaxError>,
2019-01-07 07:15:47 -06:00
old_node: &SyntaxNode,
edit: &AtomTextEdit,
) -> Vec<SyntaxError> {
let mut res = Vec::new();
for e in old_errors {
if e.offset() <= old_node.range().start() {
res.push(e)
} else if e.offset() >= old_node.range().end() {
res.push(e.add_offset(TextUnit::of_str(&edit.insert) - edit.delete.len()));
}
}
for e in new_errors {
res.push(e.add_offset(old_node.range().start()));
}
res
}
2018-09-15 07:35:30 -05:00
#[cfg(test)]
mod tests {
2018-12-06 12:16:37 -06:00
use test_utils::{extract_range, assert_eq_text};
2019-02-21 06:51:22 -06:00
use crate::{SourceFile, AstNode};
2018-12-06 12:16:37 -06:00
use super::*;
2018-09-15 07:35:30 -05:00
fn do_check<F>(before: &str, replace_with: &str, reparser: F)
where
2018-11-24 12:52:49 -06:00
for<'a> F: Fn(
2019-01-07 07:15:47 -06:00
&'a SyntaxNode,
&AtomTextEdit,
2019-01-07 07:15:47 -06:00
) -> Option<(&'a SyntaxNode, GreenNode, Vec<SyntaxError>)>,
{
let (range, before) = extract_range(before);
2019-01-08 12:59:55 -06:00
let edit = AtomTextEdit::replace(range, replace_with.to_owned());
let after = edit.apply(before.clone());
2018-09-15 07:35:30 -05:00
2019-01-07 07:15:47 -06:00
let fully_reparsed = SourceFile::parse(&after);
let incrementally_reparsed = {
2019-01-07 07:15:47 -06:00
let f = SourceFile::parse(&before);
2019-02-08 05:49:43 -06:00
let edit = AtomTextEdit { delete: range, insert: replace_with.to_string() };
let (node, green, new_errors) =
reparser(f.syntax(), &edit).expect("cannot incrementally reparse");
let green_root = node.replace_with(green);
let errors = super::merge_errors(f.errors(), new_errors, node, &edit);
2019-01-07 07:15:47 -06:00
SourceFile::new(green_root, errors)
};
assert_eq_text!(
2019-02-21 06:51:22 -06:00
&fully_reparsed.syntax().debug_dump(),
&incrementally_reparsed.syntax().debug_dump(),
)
}
#[test]
fn reparse_block_tests() {
let do_check = |before, replace_to| do_check(before, replace_to, reparse_block);
2018-09-15 07:35:30 -05:00
do_check(
r"
2018-09-15 07:35:30 -05:00
fn foo() {
let x = foo + <|>bar<|>
}
",
"baz",
);
do_check(
r"
2018-09-15 07:35:30 -05:00
fn foo() {
let x = foo<|> + bar<|>
}
",
"baz",
);
do_check(
r"
2018-09-15 07:35:30 -05:00
struct Foo {
f: foo<|><|>
}
",
",\n g: (),",
);
do_check(
r"
2018-09-15 07:35:30 -05:00
fn foo {
let;
1 + 1;
<|>92<|>;
}
",
"62",
);
do_check(
r"
2018-09-15 07:35:30 -05:00
mod foo {
fn <|><|>
}
",
"bar",
);
do_check(
r"
2018-09-15 07:35:30 -05:00
trait Foo {
type <|>Foo<|>;
}
",
"Output",
);
do_check(
r"
2018-09-15 07:35:30 -05:00
impl IntoIterator<Item=i32> for Foo {
f<|><|>
}
",
"n next(",
);
do_check(
r"
2018-09-15 07:35:30 -05:00
use a::b::{foo,<|>,bar<|>};
",
"baz",
);
do_check(
r"
2018-09-15 07:35:30 -05:00
pub enum A {
Foo<|><|>
}
",
"\nBar;\n",
);
do_check(
r"
2018-09-15 07:35:30 -05:00
foo!{a, b<|><|> d}
",
", c[3]",
);
do_check(
r"
2018-09-15 07:35:30 -05:00
fn foo() {
vec![<|><|>]
}
",
"123",
);
do_check(
r"
2018-09-15 07:35:30 -05:00
extern {
fn<|>;<|>
}
",
" exit(code: c_int)",
);
}
#[test]
fn reparse_leaf_tests() {
let do_check = |before, replace_to| do_check(before, replace_to, reparse_leaf);
do_check(
r"<|><|>
fn foo() -> i32 { 1 }
",
"\n\n\n \n",
);
do_check(
r"
2018-09-15 07:35:30 -05:00
fn foo() -> <|><|> {}
",
" \n",
);
do_check(
r"
fn <|>foo<|>() -> i32 { 1 }
",
"bar",
);
do_check(
r"
fn foo<|><|>foo() { }
",
"bar",
);
do_check(
r"
fn foo /* <|><|> */ () {}
",
"some comment",
);
do_check(
r"
fn baz <|><|> () {}
",
" \t\t\n\n",
);
do_check(
r"
fn baz <|><|> () {}
",
" \t\t\n\n",
);
do_check(
r"
/// foo <|><|>omment
mod { }
",
"c",
);
do_check(
r#"
2018-09-15 07:35:30 -05:00
fn -> &str { "Hello<|><|>" }
"#,
", world",
);
do_check(
r#"
2018-09-15 07:35:30 -05:00
fn -> &str { // "Hello<|><|>"
"#,
", world",
);
do_check(
r##"
fn -> &str { r#"Hello<|><|>"#
"##,
", world",
);
do_check(
r"
#[derive(<|>Copy<|>)]
enum Foo {
}
",
"Clone",
);
2018-09-15 07:35:30 -05:00
}
2018-10-02 09:07:12 -05:00
}