internal: replace TreeSink with a data structure

The general theme of this is to make parser a better independent
library.

The specific thing we do here is replacing callback based TreeSink with
a data structure. That is, rather than calling user-provided tree
construction methods, the parser now spits out a very bare-bones tree,
effectively a log of a DFS traversal.

This makes the parser usable without any *specifc* tree sink, and allows
us to, eg, move tests into this crate.

Now, it's also true that this is a distinction without a difference, as
the old and the new interface are equivalent in expressiveness. Still,
this new thing seems somewhat simpler. But yeah, I admit I don't have a
suuper strong motivation here, just a hunch that this is better.

crates/parser/src/event.rs

@@ -10,9 +10,8 @@
 use std::mem;
 
 use crate::{
-    ParseError,
+    tree_traversal::TreeTraversal,
     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>) -> TreeTraversal {
+    let mut res = TreeTraversal::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
 }

crates/parser/src/lib.rs

@@ -25,31 +25,19 @@ mod event;
 mod parser;
 mod grammar;
 mod tokens;
+mod tree_traversal;
 
 #[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::{
+    lexed_str::LexedStr,
+    syntax_kind::SyntaxKind,
+    tokens::Tokens,
+    tree_traversal::{TraversalStep, TreeTraversal},
+};
 
 /// rust-analyzer parser allows you to choose one of the possible entry points.
 ///
@@ -74,11 +62,11 @@ 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(tokens: &Tokens) -> TreeTraversal {
+    parse(tokens, ParserEntryPoint::SourceFile)
 }
 
-pub fn parse(tokens: &Tokens, tree_sink: &mut dyn TreeSink, entry_point: ParserEntryPoint) {
+pub fn parse(tokens: &Tokens, entry_point: ParserEntryPoint) -> TreeTraversal {
     let entry_point: fn(&'_ mut parser::Parser) = match entry_point {
         ParserEntryPoint::SourceFile => grammar::entry_points::source_file,
         ParserEntryPoint::Path => grammar::entry_points::path,
@@ -99,7 +87,7 @@ pub fn parse(tokens: &Tokens, tree_sink: &mut dyn TreeSink, entry_point: ParserE
     let mut p = parser::Parser::new(tokens);
     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 +107,11 @@ 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: &Tokens) -> TreeTraversal {
         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)
     }
 }

crates/parser/src/parser.rs

@@ -8,7 +8,6 @@ use limit::Limit;
 use crate::{
     event::Event,
     tokens::Tokens,
-    ParseError,
     SyntaxKind::{self, EOF, ERROR, TOMBSTONE},
     TokenSet, T,
 };
@@ -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 });
     }
 

crates/parser/src/tree_traversal.rs

@@ -0,0 +1,67 @@
+//! TODO
+use crate::SyntaxKind;
+
+/// Output of the parser.
+#[derive(Default)]
+pub struct TreeTraversal {
+    /// 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_raw_tokens|4 bit tag|4 bit leftover|
+    ///
+    event: Vec<u32>,
+    error: Vec<String>,
+}
+
+pub enum TraversalStep<'a> {
+    Token { kind: SyntaxKind, n_raw_tokens: u8 },
+    EnterNode { kind: SyntaxKind },
+    LeaveNode,
+    Error { msg: &'a str },
+}
+
+impl TreeTraversal {
+    pub fn iter(&self) -> impl Iterator<Item = TraversalStep<'_>> {
+        self.event.iter().map(|&event| {
+            if event & 0b1 == 0 {
+                return TraversalStep::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_raw_tokens = ((event & 0x0000_FF00) >> 8) as u8;
+                    TraversalStep::Token { kind, n_raw_tokens }
+                }
+                1 => {
+                    let kind: SyntaxKind = (((event & 0xFFFF_0000) >> 16) as u16).into();
+                    TraversalStep::EnterNode { kind }
+                }
+                2 => TraversalStep::LeaveNode,
+                _ => 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);
+    }
+}