Unnamed repository; edit this file 'description' to name the repository.
-rw-r--r--crates/mbe/src/lib.rs2
-rw-r--r--crates/mbe/src/syntax_bridge.rs21
-rw-r--r--crates/mbe/src/to_parser_input.rs (renamed from crates/mbe/src/to_parser_tokens.rs)4
-rw-r--r--crates/mbe/src/tt_iter.rs51
-rw-r--r--crates/parser/src/event.rs18
-rw-r--r--crates/parser/src/input.rs (renamed from crates/parser/src/tokens.rs)22
-rw-r--r--crates/parser/src/lexed_str.rs4
-rw-r--r--crates/parser/src/lib.rs50
-rw-r--r--crates/parser/src/output.rs76
-rw-r--r--crates/parser/src/parser.rs33
-rw-r--r--crates/syntax/src/parsing.rs38
-rw-r--r--crates/syntax/src/parsing/reparsing.rs10
-rw-r--r--crates/syntax/src/parsing/text_tree_sink.rs111
-rw-r--r--crates/syntax/src/syntax_node.rs4
14 files changed, 259 insertions, 185 deletions
diff --git a/crates/mbe/src/lib.rs b/crates/mbe/src/lib.rs
index 1a56878fdb..5e14a3fb59 100644
--- a/crates/mbe/src/lib.rs
+++ b/crates/mbe/src/lib.rs
@@ -10,7 +10,7 @@ mod parser;
mod expander;
mod syntax_bridge;
mod tt_iter;
-mod to_parser_tokens;
+mod to_parser_input;
#[cfg(test)]
mod benchmark;
diff --git a/crates/mbe/src/syntax_bridge.rs b/crates/mbe/src/syntax_bridge.rs
index 109842b0cd..f0c1f806ff 100644
--- a/crates/mbe/src/syntax_bridge.rs
+++ b/crates/mbe/src/syntax_bridge.rs
@@ -1,6 +1,5 @@
//! Conversions between [`SyntaxNode`] and [`tt::TokenTree`].
-use parser::{ParseError, TreeSink};
use rustc_hash::{FxHashMap, FxHashSet};
use syntax::{
ast::{self, make::tokens::doc_comment},
@@ -11,7 +10,7 @@ use syntax::{
use tt::buffer::{Cursor, TokenBuffer};
use crate::{
- to_parser_tokens::to_parser_tokens, tt_iter::TtIter, ExpandError, ParserEntryPoint, TokenMap,
+ to_parser_input::to_parser_input, tt_iter::TtIter, ExpandError, ParserEntryPoint, TokenMap,
};
/// Convert the syntax node to a `TokenTree` (what macro
@@ -55,9 +54,19 @@ pub fn token_tree_to_syntax_node(
}
_ => TokenBuffer::from_subtree(tt),
};
- let parser_tokens = to_parser_tokens(&buffer);
+ let parser_input = to_parser_input(&buffer);
+ let parser_output = parser::parse(&parser_input, entry_point);
let mut tree_sink = TtTreeSink::new(buffer.begin());
- parser::parse(&parser_tokens, &mut tree_sink, entry_point);
+ for event in parser_output.iter() {
+ match event {
+ parser::Step::Token { kind, n_input_tokens: n_raw_tokens } => {
+ tree_sink.token(kind, n_raw_tokens)
+ }
+ parser::Step::Enter { kind } => tree_sink.start_node(kind),
+ parser::Step::Exit => tree_sink.finish_node(),
+ parser::Step::Error { msg } => tree_sink.error(msg.to_string()),
+ }
+ }
if tree_sink.roots.len() != 1 {
return Err(ExpandError::ConversionError);
}
@@ -643,7 +652,7 @@ fn delim_to_str(d: tt::DelimiterKind, closing: bool) -> &'static str {
&texts[idx..texts.len() - (1 - idx)]
}
-impl<'a> TreeSink for TtTreeSink<'a> {
+impl<'a> TtTreeSink<'a> {
fn token(&mut self, kind: SyntaxKind, mut n_tokens: u8) {
if kind == LIFETIME_IDENT {
n_tokens = 2;
@@ -741,7 +750,7 @@ impl<'a> TreeSink for TtTreeSink<'a> {
*self.roots.last_mut().unwrap() -= 1;
}
- fn error(&mut self, error: ParseError) {
+ fn error(&mut self, error: String) {
self.inner.error(error, self.text_pos)
}
}
diff --git a/crates/mbe/src/to_parser_tokens.rs b/crates/mbe/src/to_parser_input.rs
index f419c78d46..6faa147218 100644
--- a/crates/mbe/src/to_parser_tokens.rs
+++ b/crates/mbe/src/to_parser_input.rs
@@ -4,8 +4,8 @@
use syntax::{SyntaxKind, SyntaxKind::*, T};
use tt::buffer::TokenBuffer;
-pub(crate) fn to_parser_tokens(buffer: &TokenBuffer) -> parser::Tokens {
- let mut res = parser::Tokens::default();
+pub(crate) fn to_parser_input(buffer: &TokenBuffer) -> parser::Input {
+ let mut res = parser::Input::default();
let mut current = buffer.begin();
diff --git a/crates/mbe/src/tt_iter.rs b/crates/mbe/src/tt_iter.rs
index d05e84b0f0..2d2dbd8994 100644
--- a/crates/mbe/src/tt_iter.rs
+++ b/crates/mbe/src/tt_iter.rs
@@ -1,11 +1,10 @@
//! A "Parser" structure for token trees. We use this when parsing a declarative
//! macro definition into a list of patterns and templates.
-use crate::{to_parser_tokens::to_parser_tokens, ExpandError, ExpandResult, ParserEntryPoint};
+use crate::{to_parser_input::to_parser_input, ExpandError, ExpandResult, ParserEntryPoint};
-use parser::TreeSink;
use syntax::SyntaxKind;
-use tt::buffer::{Cursor, TokenBuffer};
+use tt::buffer::TokenBuffer;
macro_rules! err {
() => {
@@ -94,34 +93,28 @@ impl<'a> TtIter<'a> {
&mut self,
entry_point: ParserEntryPoint,
) -> ExpandResult<Option<tt::TokenTree>> {
- struct OffsetTokenSink<'a> {
- cursor: Cursor<'a>,
- error: bool,
- }
-
- impl<'a> TreeSink for OffsetTokenSink<'a> {
- fn token(&mut self, kind: SyntaxKind, mut n_tokens: u8) {
- if kind == SyntaxKind::LIFETIME_IDENT {
- n_tokens = 2;
- }
- for _ in 0..n_tokens {
- self.cursor = self.cursor.bump_subtree();
+ let buffer = TokenBuffer::from_tokens(self.inner.as_slice());
+ let parser_input = to_parser_input(&buffer);
+ let tree_traversal = parser::parse(&parser_input, entry_point);
+
+ let mut cursor = buffer.begin();
+ let mut error = false;
+ for step in tree_traversal.iter() {
+ match step {
+ parser::Step::Token { kind, mut n_input_tokens } => {
+ if kind == SyntaxKind::LIFETIME_IDENT {
+ n_input_tokens = 2;
+ }
+ for _ in 0..n_input_tokens {
+ cursor = cursor.bump_subtree();
+ }
}
- }
- fn start_node(&mut self, _kind: SyntaxKind) {}
- fn finish_node(&mut self) {}
- fn error(&mut self, _error: parser::ParseError) {
- self.error = true;
+ parser::Step::Enter { .. } | parser::Step::Exit => (),
+ parser::Step::Error { .. } => error = true,
}
}
- let buffer = TokenBuffer::from_tokens(self.inner.as_slice());
- let parser_tokens = to_parser_tokens(&buffer);
- let mut sink = OffsetTokenSink { cursor: buffer.begin(), error: false };
-
- parser::parse(&parser_tokens, &mut sink, entry_point);
-
- let mut err = if !sink.cursor.is_root() || sink.error {
+ let mut err = if !cursor.is_root() || error {
Some(err!("expected {:?}", entry_point))
} else {
None
@@ -130,8 +123,8 @@ impl<'a> TtIter<'a> {
let mut curr = buffer.begin();
let mut res = vec![];
- if sink.cursor.is_root() {
- while curr != sink.cursor {
+ if cursor.is_root() {
+ while curr != cursor {
if let Some(token) = curr.token_tree() {
res.push(token);
}
diff --git a/crates/parser/src/event.rs b/crates/parser/src/event.rs
index 41b0328027..b0e70e7943 100644
--- a/crates/parser/src/event.rs
+++ b/crates/parser/src/event.rs
@@ -10,9 +10,8 @@
use std::mem;
use crate::{
- ParseError,
+ output::Output,
SyntaxKind::{self, *},
- TreeSink,
};
/// `Parser` produces a flat list of `Event`s.
@@ -77,7 +76,7 @@ pub(crate) enum Event {
},
Error {
- msg: ParseError,
+ msg: String,
},
}
@@ -88,7 +87,8 @@ impl Event {
}
/// Generate the syntax tree with the control of events.
-pub(super) fn process(sink: &mut dyn TreeSink, mut events: Vec<Event>) {
+pub(super) fn process(mut events: Vec<Event>) -> Output {
+ let mut res = Output::default();
let mut forward_parents = Vec::new();
for i in 0..events.len() {
@@ -117,15 +117,17 @@ pub(super) fn process(sink: &mut dyn TreeSink, mut events: Vec<Event>) {
for kind in forward_parents.drain(..).rev() {
if kind != TOMBSTONE {
- sink.start_node(kind);
+ res.enter_node(kind);
}
}
}
- Event::Finish => sink.finish_node(),
+ Event::Finish => res.leave_node(),
Event::Token { kind, n_raw_tokens } => {
- sink.token(kind, n_raw_tokens);
+ res.token(kind, n_raw_tokens);
}
- Event::Error { msg } => sink.error(msg),
+ Event::Error { msg } => res.error(msg),
}
}
+
+ res
}
diff --git a/crates/parser/src/tokens.rs b/crates/parser/src/input.rs
index 4fc2361add..9504bd4d9e 100644
--- a/crates/parser/src/tokens.rs
+++ b/crates/parser/src/input.rs
@@ -1,26 +1,26 @@
-//! Input for the parser -- a sequence of tokens.
-//!
-//! As of now, parser doesn't have access to the *text* of the tokens, and makes
-//! decisions based solely on their classification. Unlike `LexerToken`, the
-//! `Tokens` doesn't include whitespace and comments.
+//! See [`Input`].
use crate::SyntaxKind;
#[allow(non_camel_case_types)]
type bits = u64;
-/// Main input to the parser.
+/// Input for the parser -- a sequence of tokens.
///
-/// A sequence of tokens represented internally as a struct of arrays.
+/// As of now, parser doesn't have access to the *text* of the tokens, and makes
+/// decisions based solely on their classification. Unlike `LexerToken`, the
+/// `Tokens` doesn't include whitespace and comments. Main input to the parser.
+///
+/// Struct of arrays internally, but this shouldn't really matter.
#[derive(Default)]
-pub struct Tokens {
+pub struct Input {
kind: Vec<SyntaxKind>,
joint: Vec<bits>,
contextual_kind: Vec<SyntaxKind>,
}
/// `pub` impl used by callers to create `Tokens`.
-impl Tokens {
+impl Input {
#[inline]
pub fn push(&mut self, kind: SyntaxKind) {
self.push_impl(kind, SyntaxKind::EOF)
@@ -63,7 +63,7 @@ impl Tokens {
}
/// pub(crate) impl used by the parser to consume `Tokens`.
-impl Tokens {
+impl Input {
pub(crate) fn kind(&self, idx: usize) -> SyntaxKind {
self.kind.get(idx).copied().unwrap_or(SyntaxKind::EOF)
}
@@ -76,7 +76,7 @@ impl Tokens {
}
}
-impl Tokens {
+impl Input {
fn bit_index(&self, n: usize) -> (usize, usize) {
let idx = n / (bits::BITS as usize);
let b_idx = n % (bits::BITS as usize);
diff --git a/crates/parser/src/lexed_str.rs b/crates/parser/src/lexed_str.rs
index b8936c3440..f17aae1d31 100644
--- a/crates/parser/src/lexed_str.rs
+++ b/crates/parser/src/lexed_str.rs
@@ -122,8 +122,8 @@ impl<'a> LexedStr<'a> {
self.error.iter().map(|it| (it.token as usize, it.msg.as_str()))
}
- pub fn to_tokens(&self) -> crate::Tokens {
- let mut res = crate::Tokens::default();
+ pub fn to_input(&self) -> crate::Input {
+ let mut res = crate::Input::default();
let mut was_joint = false;
for i in 0..self.len() {
let kind = self.kind(i);
diff --git a/crates/parser/src/lib.rs b/crates/parser/src/lib.rs
index dc02ae6e83..da78889f35 100644
--- a/crates/parser/src/lib.rs
+++ b/crates/parser/src/lib.rs
@@ -24,32 +24,20 @@ mod syntax_kind;
mod event;
mod parser;
mod grammar;
-mod tokens;
+mod input;
+mod output;
#[cfg(test)]
mod tests;
pub(crate) use token_set::TokenSet;
-pub use crate::{lexed_str::LexedStr, syntax_kind::SyntaxKind, tokens::Tokens};
-
-#[derive(Debug, Clone, PartialEq, Eq, Hash)]
-pub struct ParseError(pub Box<String>);
-
-/// `TreeSink` abstracts details of a particular syntax tree implementation.
-pub trait TreeSink {
- /// Adds new token to the current branch.
- fn token(&mut self, kind: SyntaxKind, n_tokens: u8);
-
- /// Start new branch and make it current.
- fn start_node(&mut self, kind: SyntaxKind);
-
- /// Finish current branch and restore previous
- /// branch as current.
- fn finish_node(&mut self);
-
- fn error(&mut self, error: ParseError);
-}
+pub use crate::{
+ input::Input,
+ lexed_str::LexedStr,
+ output::{Output, Step},
+ syntax_kind::SyntaxKind,
+};
/// rust-analyzer parser allows you to choose one of the possible entry points.
///
@@ -74,11 +62,19 @@ pub enum ParserEntryPoint {
}
/// Parse given tokens into the given sink as a rust file.
-pub fn parse_source_file(tokens: &Tokens, tree_sink: &mut dyn TreeSink) {
- parse(tokens, tree_sink, ParserEntryPoint::SourceFile);
+pub fn parse_source_file(inp: &Input) -> Output {
+ parse(inp, ParserEntryPoint::SourceFile)
}
-pub fn parse(tokens: &Tokens, tree_sink: &mut dyn TreeSink, entry_point: ParserEntryPoint) {
+/// Parses the given [`Input`] into [`Output`] assuming that the top-level
+/// syntactic construct is the given [`ParserEntryPoint`].
+///
+/// Both input and output here are fairly abstract. The overall flow is that the
+/// caller has some "real" tokens, converts them to [`Input`], parses them to
+/// [`Output`], and then converts that into a "real" tree. The "real" tree is
+/// made of "real" tokens, so this all hinges on rather tight coordination of
+/// indices between the four stages.
+pub fn parse(inp: &Input, entry_point: ParserEntryPoint) -> Output {
let entry_point: fn(&'_ mut parser::Parser) = match entry_point {
ParserEntryPoint::SourceFile => grammar::entry_points::source_file,
ParserEntryPoint::Path => grammar::entry_points::path,
@@ -96,10 +92,10 @@ pub fn parse(tokens: &Tokens, tree_sink: &mut dyn TreeSink, entry_point: ParserE
ParserEntryPoint::Attr => grammar::entry_points::attr,
};
- let mut p = parser::Parser::new(tokens);
+ let mut p = parser::Parser::new(inp);
entry_point(&mut p);
let events = p.finish();
- event::process(tree_sink, events);
+ event::process(events)
}
/// A parsing function for a specific braced-block.
@@ -119,11 +115,11 @@ impl Reparser {
///
/// Tokens must start with `{`, end with `}` and form a valid brace
/// sequence.
- pub fn parse(self, tokens: &Tokens, tree_sink: &mut dyn TreeSink) {
+ pub fn parse(self, tokens: &Input) -> Output {
let Reparser(r) = self;
let mut p = parser::Parser::new(tokens);
r(&mut p);
let events = p.finish();
- event::process(tree_sink, events);
+ event::process(events)
}
}
diff --git a/crates/parser/src/output.rs b/crates/parser/src/output.rs
new file mode 100644
index 0000000000..b613df029f
--- /dev/null
+++ b/crates/parser/src/output.rs
@@ -0,0 +1,76 @@
+//! See [`Output`]
+
+use crate::SyntaxKind;
+
+/// Output of the parser -- a DFS traversal of a concrete syntax tree.
+///
+/// Use the [`Output::iter`] method to iterate over traversal steps and consume
+/// a syntax tree.
+///
+/// In a sense, this is just a sequence of [`SyntaxKind`]-colored parenthesis
+/// interspersed into the original [`crate::Input`]. The output is fundamentally
+/// coordinated with the input and `n_input_tokens` refers to the number of
+/// times [`crate::Input::push`] was called.
+#[derive(Default)]
+pub struct Output {
+ /// 32-bit encoding of events. If LSB is zero, then that's an index into the
+ /// error vector. Otherwise, it's one of the thee other variants, with data encoded as
+ ///
+ /// |16 bit kind|8 bit n_input_tokens|4 bit tag|4 bit leftover|
+ ///
+ event: Vec<u32>,
+ error: Vec<String>,
+}
+
+pub enum Step<'a> {
+ Token { kind: SyntaxKind, n_input_tokens: u8 },
+ Enter { kind: SyntaxKind },
+ Exit,
+ Error { msg: &'a str },
+}
+
+impl Output {
+ pub fn iter(&self) -> impl Iterator<Item = Step<'_>> {
+ self.event.iter().map(|&event| {
+ if event & 0b1 == 0 {
+ return Step::Error { msg: self.error[(event as usize) >> 1].as_str() };
+ }
+ let tag = ((event & 0x0000_00F0) >> 4) as u8;
+ match tag {
+ 0 => {
+ let kind: SyntaxKind = (((event & 0xFFFF_0000) >> 16) as u16).into();
+ let n_input_tokens = ((event & 0x0000_FF00) >> 8) as u8;
+ Step::Token { kind, n_input_tokens }
+ }
+ 1 => {
+ let kind: SyntaxKind = (((event & 0xFFFF_0000) >> 16) as u16).into();
+ Step::Enter { kind }
+ }
+ 2 => Step::Exit,
+ _ => unreachable!(),
+ }
+ })
+ }
+
+ pub(crate) fn token(&mut self, kind: SyntaxKind, n_tokens: u8) {
+ let e = ((kind as u16 as u32) << 16) | ((n_tokens as u32) << 8) | (0 << 4) | 1;
+ self.event.push(e)
+ }
+
+ pub(crate) fn enter_node(&mut self, kind: SyntaxKind) {
+ let e = ((kind as u16 as u32) << 16) | (1 << 4) | 1;
+ self.event.push(e)
+ }
+
+ pub(crate) fn leave_node(&mut self) {
+ let e = 2 << 4 | 1;
+ self.event.push(e)
+ }
+
+ pub(crate) fn error(&mut self, error: String) {
+ let idx = self.error.len();
+ self.error.push(error);
+ let e = (idx as u32) << 1;
+ self.event.push(e);
+ }
+}
diff --git a/crates/parser/src/parser.rs b/crates/parser/src/parser.rs
index 4c891108a6..d4aecf9b44 100644
--- a/crates/parser/src/parser.rs
+++ b/crates/parser/src/parser.rs
@@ -7,8 +7,7 @@ use limit::Limit;
use crate::{
event::Event,
- tokens::Tokens,
- ParseError,
+ input::Input,
SyntaxKind::{self, EOF, ERROR, TOMBSTONE},
TokenSet, T,
};
@@ -23,7 +22,7 @@ use crate::{
/// "start expression, consume number literal,
/// finish expression". See `Event` docs for more.
pub(crate) struct Parser<'t> {
- tokens: &'t Tokens,
+ inp: &'t Input,
pos: usize,
events: Vec<Event>,
steps: Cell<u32>,
@@ -32,8 +31,8 @@ pub(crate) struct Parser<'t> {
static PARSER_STEP_LIMIT: Limit = Limit::new(15_000_000);
impl<'t> Parser<'t> {
- pub(super) fn new(tokens: &'t Tokens) -> Parser<'t> {
- Parser { tokens, pos: 0, events: Vec::new(), steps: Cell::new(0) }
+ pub(super) fn new(inp: &'t Input) -> Parser<'t> {
+ Parser { inp, pos: 0, events: Vec::new(), steps: Cell::new(0) }
}
pub(crate) fn finish(self) -> Vec<Event> {
@@ -56,7 +55,7 @@ impl<'t> Parser<'t> {
assert!(PARSER_STEP_LIMIT.check(steps as usize).is_ok(), "the parser seems stuck");
self.steps.set(steps + 1);
- self.tokens.kind(self.pos + n)
+ self.inp.kind(self.pos + n)
}
/// Checks if the current token is `kind`.
@@ -92,7 +91,7 @@ impl<'t> Parser<'t> {
T![<<=] => self.at_composite3(n, T![<], T![<], T![=]),
T![>>=] => self.at_composite3(n, T![>], T![>], T![=]),
- _ => self.tokens.kind(self.pos + n) == kind,
+ _ => self.inp.kind(self.pos + n) == kind,
}
}
@@ -131,17 +130,17 @@ impl<'t> Parser<'t> {
}
fn at_composite2(&self, n: usize, k1: SyntaxKind, k2: SyntaxKind) -> bool {
- self.tokens.kind(self.pos + n) == k1
- && self.tokens.kind(self.pos + n + 1) == k2
- && self.tokens.is_joint(self.pos + n)
+ self.inp.kind(self.pos + n) == k1
+ && self.inp.kind(self.pos + n + 1) == k2
+ && self.inp.is_joint(self.pos + n)
}
fn at_composite3(&self, n: usize, k1: SyntaxKind, k2: SyntaxKind, k3: SyntaxKind) -> bool {
- self.tokens.kind(self.pos + n) == k1
- && self.tokens.kind(self.pos + n + 1) == k2
- && self.tokens.kind(self.pos + n + 2) == k3
- && self.tokens.is_joint(self.pos + n)
- && self.tokens.is_joint(self.pos + n + 1)
+ self.inp.kind(self.pos + n) == k1
+ && self.inp.kind(self.pos + n + 1) == k2
+ && self.inp.kind(self.pos + n + 2) == k3
+ && self.inp.is_joint(self.pos + n)
+ && self.inp.is_joint(self.pos + n + 1)
}
/// Checks if the current token is in `kinds`.
@@ -151,7 +150,7 @@ impl<'t> Parser<'t> {
/// Checks if the current token is contextual keyword with text `t`.
pub(crate) fn at_contextual_kw(&self, kw: SyntaxKind) -> bool {
- self.tokens.contextual_kind(self.pos) == kw
+ self.inp.contextual_kind(self.pos) == kw
}
/// Starts a new node in the syntax tree. All nodes and tokens
@@ -196,7 +195,7 @@ impl<'t> Parser<'t> {
/// structured errors with spans and notes, like rustc
/// does.
pub(crate) fn error<T: Into<String>>(&mut self, message: T) {
- let msg = ParseError(Box::new(message.into()));
+ let msg = message.into();
self.push_event(Event::Error { msg });
}
diff --git a/crates/syntax/src/parsing.rs b/crates/syntax/src/parsing.rs
index cba1ddde85..20c7101a03 100644
--- a/crates/syntax/src/parsing.rs
+++ b/crates/syntax/src/parsing.rs
@@ -4,24 +4,18 @@
mod text_tree_sink;
mod reparsing;
-use parser::SyntaxKind;
-use text_tree_sink::TextTreeSink;
-
-use crate::{syntax_node::GreenNode, AstNode, SyntaxError, SyntaxNode};
+use crate::{
+ parsing::text_tree_sink::build_tree, syntax_node::GreenNode, AstNode, SyntaxError, SyntaxNode,
+};
pub(crate) use crate::parsing::reparsing::incremental_reparse;
pub(crate) fn parse_text(text: &str) -> (GreenNode, Vec<SyntaxError>) {
let lexed = parser::LexedStr::new(text);
- let parser_tokens = lexed.to_tokens();
-
- let mut tree_sink = TextTreeSink::new(lexed);
-
- parser::parse_source_file(&parser_tokens, &mut tree_sink);
-
- let (tree, parser_errors) = tree_sink.finish();
-
- (tree, parser_errors)
+ let parser_input = lexed.to_input();
+ let parser_output = parser::parse_source_file(&parser_input);
+ let (node, errors, _eof) = build_tree(lexed, parser_output, false);
+ (node, errors)
}
/// Returns `text` parsed as a `T` provided there are no parse errors.
@@ -33,21 +27,13 @@ pub(crate) fn parse_text_as<T: AstNode>(
if lexed.errors().next().is_some() {
return Err(());
}
- let parser_tokens = lexed.to_tokens();
-
- let mut tree_sink = TextTreeSink::new(lexed);
-
- // TextTreeSink assumes that there's at least some root node to which it can attach errors and
- // tokens. We arbitrarily give it a SourceFile.
- use parser::TreeSink;
- tree_sink.start_node(SyntaxKind::SOURCE_FILE);
- parser::parse(&parser_tokens, &mut tree_sink, entry_point);
- tree_sink.finish_node();
+ let parser_input = lexed.to_input();
+ let parser_output = parser::parse(&parser_input, entry_point);
+ let (node, errors, eof) = build_tree(lexed, parser_output, true);
- let (tree, parser_errors, eof) = tree_sink.finish_eof();
- if !parser_errors.is_empty() || !eof {
+ if !errors.is_empty() || !eof {
return Err(());
}
- SyntaxNode::new_root(tree).first_child().and_then(T::cast).ok_or(())
+ SyntaxNode::new_root(node).first_child().and_then(T::cast).ok_or(())
}
diff --git a/crates/syntax/src/parsing/reparsing.rs b/crates/syntax/src/parsing/reparsing.rs
index e9567a838c..a6abe3cccf 100644
--- a/crates/syntax/src/parsing/reparsing.rs
+++ b/crates/syntax/src/parsing/reparsing.rs
@@ -10,7 +10,7 @@ use parser::Reparser;
use text_edit::Indel;
use crate::{
- parsing::text_tree_sink::TextTreeSink,
+ parsing::text_tree_sink::build_tree,
syntax_node::{GreenNode, GreenToken, NodeOrToken, SyntaxElement, SyntaxNode},
SyntaxError,
SyntaxKind::*,
@@ -89,16 +89,14 @@ fn reparse_block(
let text = get_text_after_edit(node.clone().into(), edit);
let lexed = parser::LexedStr::new(text.as_str());
- let parser_tokens = lexed.to_tokens();
+ let parser_input = lexed.to_input();
if !is_balanced(&lexed) {
return None;
}
- let mut tree_sink = TextTreeSink::new(lexed);
+ let tree_traversal = reparser.parse(&parser_input);
- reparser.parse(&parser_tokens, &mut tree_sink);
-
- let (green, new_parser_errors) = tree_sink.finish();
+ let (green, new_parser_errors, _eof) = build_tree(lexed, tree_traversal, false);
Some((node.replace_with(green), new_parser_errors, node.text_range()))
}
diff --git a/crates/syntax/src/parsing/text_tree_sink.rs b/crates/syntax/src/parsing/text_tree_sink.rs
index c9e7feb965..f40c549e3d 100644
--- a/crates/syntax/src/parsing/text_tree_sink.rs
+++ b/crates/syntax/src/parsing/text_tree_sink.rs
@@ -2,7 +2,7 @@
use std::mem;
-use parser::{LexedStr, ParseError, TreeSink};
+use parser::LexedStr;
use crate::{
ast,
@@ -12,10 +12,37 @@ use crate::{
SyntaxTreeBuilder, TextRange,
};
-/// Bridges the parser with our specific syntax tree representation.
-///
-/// `TextTreeSink` also handles attachment of trivia (whitespace) to nodes.
-pub(crate) struct TextTreeSink<'a> {
+pub(crate) fn build_tree(
+ lexed: LexedStr<'_>,
+ parser_output: parser::Output,
+ synthetic_root: bool,
+) -> (GreenNode, Vec<SyntaxError>, bool) {
+ let mut builder = Builder::new(lexed);
+
+ if synthetic_root {
+ builder.enter(SyntaxKind::SOURCE_FILE);
+ }
+
+ for event in parser_output.iter() {
+ match event {
+ parser::Step::Token { kind, n_input_tokens: n_raw_tokens } => {
+ builder.token(kind, n_raw_tokens)
+ }
+ parser::Step::Enter { kind } => builder.enter(kind),
+ parser::Step::Exit => builder.exit(),
+ parser::Step::Error { msg } => {
+ let text_pos = builder.lexed.text_start(builder.pos).try_into().unwrap();
+ builder.inner.error(msg.to_string(), text_pos);
+ }
+ }
+ }
+ if synthetic_root {
+ builder.exit()
+ }
+ builder.build()
+}
+
+struct Builder<'a> {
lexed: LexedStr<'a>,
pos: usize,
state: State,
@@ -28,7 +55,35 @@ enum State {
PendingFinish,
}
-impl<'a> TreeSink for TextTreeSink<'a> {
+impl<'a> Builder<'a> {
+ fn new(lexed: parser::LexedStr<'a>) -> Self {
+ Self { lexed, pos: 0, state: State::PendingStart, inner: SyntaxTreeBuilder::default() }
+ }
+
+ fn build(mut self) -> (GreenNode, Vec<SyntaxError>, bool) {
+ match mem::replace(&mut self.state, State::Normal) {
+ State::PendingFinish => {
+ self.eat_trivias();
+ self.inner.finish_node();
+ }
+ State::PendingStart | State::Normal => unreachable!(),
+ }
+
+ let (node, mut errors) = self.inner.finish_raw();
+ for (i, err) in self.lexed.errors() {
+ let text_range = self.lexed.text_range(i);
+ let text_range = TextRange::new(
+ text_range.start.try_into().unwrap(),
+ text_range.end.try_into().unwrap(),
+ );
+ errors.push(SyntaxError::new(err, text_range))
+ }
+
+ let is_eof = self.pos == self.lexed.len();
+
+ (node, errors, is_eof)
+ }
+
fn token(&mut self, kind: SyntaxKind, n_tokens: u8) {
match mem::replace(&mut self.state, State::Normal) {
State::PendingStart => unreachable!(),
@@ -39,7 +94,7 @@ impl<'a> TreeSink for TextTreeSink<'a> {
self.do_token(kind, n_tokens as usize);
}
- fn start_node(&mut self, kind: SyntaxKind) {
+ fn enter(&mut self, kind: SyntaxKind) {
match mem::replace(&mut self.state, State::Normal) {
State::PendingStart => {
self.inner.start_node(kind);
@@ -63,7 +118,7 @@ impl<'a> TreeSink for TextTreeSink<'a> {
self.eat_n_trivias(n_attached_trivias);
}
- fn finish_node(&mut self) {
+ fn exit(&mut self) {
match mem::replace(&mut self.state, State::PendingFinish) {
State::PendingStart => unreachable!(),
State::PendingFinish => self.inner.finish_node(),
@@ -71,46 +126,6 @@ impl<'a> TreeSink for TextTreeSink<'a> {
}
}
- fn error(&mut self, error: ParseError) {
- let text_pos = self.lexed.text_start(self.pos).try_into().unwrap();
- self.inner.error(error, text_pos);
- }
-}
-
-impl<'a> TextTreeSink<'a> {
- pub(super) fn new(lexed: parser::LexedStr<'a>) -> Self {
- Self { lexed, pos: 0, state: State::PendingStart, inner: SyntaxTreeBuilder::default() }
- }
-
- pub(super) fn finish_eof(mut self) -> (GreenNode, Vec<SyntaxError>, bool) {
- match mem::replace(&mut self.state, State::Normal) {
- State::PendingFinish => {
- self.eat_trivias();
- self.inner.finish_node();
- }
- State::PendingStart | State::Normal => unreachable!(),
- }
-
- let (node, mut errors) = self.inner.finish_raw();
- for (i, err) in self.lexed.errors() {
- let text_range = self.lexed.text_range(i);
- let text_range = TextRange::new(
- text_range.start.try_into().unwrap(),
- text_range.end.try_into().unwrap(),
- );
- errors.push(SyntaxError::new(err, text_range))
- }
-
- let is_eof = self.pos == self.lexed.len();
-
- (node, errors, is_eof)
- }
-
- pub(super) fn finish(self) -> (GreenNode, Vec<SyntaxError>) {
- let (node, errors, _eof) = self.finish_eof();
- (node, errors)
- }
-
fn eat_trivias(&mut self) {
while self.pos < self.lexed.len() {
let kind = self.lexed.kind(self.pos);
diff --git a/crates/syntax/src/syntax_node.rs b/crates/syntax/src/syntax_node.rs
index c95c76c0a8..b96f10c173 100644
--- a/crates/syntax/src/syntax_node.rs
+++ b/crates/syntax/src/syntax_node.rs
@@ -69,7 +69,7 @@ impl SyntaxTreeBuilder {
self.inner.finish_node();
}
- pub fn error(&mut self, error: parser::ParseError, text_pos: TextSize) {
- self.errors.push(SyntaxError::new_at_offset(*error.0, text_pos));
+ pub fn error(&mut self, error: String, text_pos: TextSize) {
+ self.errors.push(SyntaxError::new_at_offset(error, text_pos));
}
}