Merge #1224

1224: Remove unused multchar puncts code and add space between puncts r=matklad a=edwin0cheng

After #1213 , parser only need single char punct, this PR do the following things: 

* Remove code which handles multi char puncts
* Remove code which handle traversal backward in `SubtreeSource` , because we cached the result in #1195
* Add space between two consecutive puncts while `tt` to `SyntaxNode` conversion . 

Note that the spaces should only be added if both puncts are not delimiters. 

Co-authored-by: Edwin Cheng <edwin0cheng@gmail.com>

crates/ra_mbe/src/subtree_source.rs

@@ -45,20 +45,6 @@ impl<'a> TokenSeq<'a> {
             }
         }
     }
-
-    fn len(&self) -> usize {
-        match self {
-            TokenSeq::Subtree(subtree) => subtree.token_trees.len() + 2,
-            TokenSeq::Seq(tokens) => tokens.len(),
-        }
-    }
-
-    fn child_slice(&self, pos: usize) -> &[tt::TokenTree] {
-        match self {
-            TokenSeq::Subtree(subtree) => &subtree.token_trees[pos - 1..],
-            TokenSeq::Seq(tokens) => &tokens[pos..],
-        }
-    }
 }
 
 #[derive(Debug, Clone, Eq, PartialEq)]
@@ -66,7 +52,6 @@ struct TtToken {
     pub kind: SyntaxKind,
     pub is_joint_to_next: bool,
     pub text: SmolStr,
-    pub n_tokens: usize,
 }
 
 #[derive(Debug, Clone, Eq, PartialEq)]
@@ -80,19 +65,12 @@ struct SubTreeWalker<'a> {
     pos: usize,
     stack: Vec<(TokenSeq<'a>, usize)>,
     cursor: WalkCursor,
-    last_steps: Vec<usize>,
     ts: TokenSeq<'a>,
 }
 
 impl<'a> SubTreeWalker<'a> {
     fn new(ts: TokenSeq<'a>) -> SubTreeWalker {
-        let mut res = SubTreeWalker {
-            pos: 0,
-            stack: vec![],
-            cursor: WalkCursor::Eof,
-            last_steps: vec![],
-            ts,
-        };
+        let mut res = SubTreeWalker { pos: 0, stack: vec![], cursor: WalkCursor::Eof, ts };
 
         res.reset();
         res
@@ -105,7 +83,6 @@ impl<'a> SubTreeWalker<'a> {
     fn reset(&mut self) {
         self.pos = 0;
         self.stack = vec![];
-        self.last_steps = vec![];
 
         self.cursor = match self.ts.get(0) {
             DelimToken::Token(token) => match token {
@@ -114,10 +91,7 @@ impl<'a> SubTreeWalker<'a> {
                     self.stack.push((ts, 0));
                     WalkCursor::Token(0, convert_delim(subtree.delimiter, false))
                 }
-                tt::TokenTree::Leaf(leaf) => {
-                    let next_tokens = self.ts.child_slice(0);
-                    WalkCursor::Token(0, convert_leaf(&next_tokens, leaf))
-                }
+                tt::TokenTree::Leaf(leaf) => WalkCursor::Token(0, convert_leaf(leaf)),
             },
             DelimToken::Delim(delim, is_end) => {
                 assert!(!is_end);
@@ -138,24 +112,6 @@ impl<'a> SubTreeWalker<'a> {
         self.stack.last().map(|(t, _)| t).unwrap_or(&self.ts)
     }
 
-    /// Move cursor backward by 1 step
-    fn backward(&mut self) {
-        if self.last_steps.is_empty() {
-            return;
-        }
-
-        self.pos -= 1;
-        let last_step = self.last_steps.pop().unwrap();
-
-        self.cursor = match self.cursor {
-            WalkCursor::Token(idx, _) => self.walk_token(idx, last_step, true),
-            WalkCursor::Eof => {
-                let len = self.top().len();
-                self.walk_token(len, last_step, true)
-            }
-        }
-    }
-
     /// Move cursor forward by 1 step        
     fn forward(&mut self) {
         if self.is_eof() {
@@ -163,37 +119,24 @@ impl<'a> SubTreeWalker<'a> {
         }
         self.pos += 1;
 
-        let step = self.current().map(|x| x.n_tokens).unwrap_or(1);
-        self.last_steps.push(step);
-
         if let WalkCursor::Token(u, _) = self.cursor {
-            self.cursor = self.walk_token(u, step, false)
+            self.cursor = self.walk_token(u)
         }
     }
 
     /// Traversal child token
-    fn walk_token(&mut self, pos: usize, offset: usize, backward: bool) -> WalkCursor {
+    fn walk_token(&mut self, pos: usize) -> WalkCursor {
         let top = self.stack.last().map(|(t, _)| t).unwrap_or(&self.ts);
-
-        if backward && pos < offset {
-            let (_, last_idx) = self.stack.pop().unwrap();
-            return self.walk_token(last_idx, offset, backward);
-        }
-
-        let pos = if backward { pos - offset } else { pos + offset };
+        let pos = pos + 1;
 
         match top.get(pos) {
             DelimToken::Token(token) => match token {
                 tt::TokenTree::Subtree(subtree) => {
                     let ts = TokenSeq::from(subtree);
-                    let new_idx = if backward { ts.len() - 1 } else { 0 };
                     self.stack.push((ts, pos));
-                    WalkCursor::Token(new_idx, convert_delim(subtree.delimiter, backward))
-                }
-                tt::TokenTree::Leaf(leaf) => {
-                    let next_tokens = top.child_slice(pos);
-                    WalkCursor::Token(pos, convert_leaf(&next_tokens, leaf))
+                    WalkCursor::Token(0, convert_delim(subtree.delimiter, false))
                 }
+                tt::TokenTree::Leaf(leaf) => WalkCursor::Token(pos, convert_leaf(leaf)),
             },
             DelimToken::Delim(delim, is_end) => {
                 WalkCursor::Token(pos, convert_delim(*delim, is_end))
@@ -201,8 +144,7 @@ impl<'a> SubTreeWalker<'a> {
             DelimToken::End => {
                 // it is the top level
                 if let Some((_, last_idx)) = self.stack.pop() {
-                    assert!(!backward);
-                    self.walk_token(last_idx, offset, backward)
+                    self.walk_token(last_idx)
                 } else {
                     WalkCursor::Eof
                 }
@@ -237,12 +179,9 @@ impl<'a> WalkerOwner<'a> {
         }
 
         while pos >= cached.len() {
-            let len = cached.len();
-            cached.push({
-                self.set_pos(len);
-                let walker = self.walker.borrow();
-                walker.current().cloned()
-            });
+            self.set_pos(cached.len());
+            let walker = self.walker.borrow();
+            cached.push(walker.current().cloned());
         }
 
         return cached[pos].clone();
@@ -250,12 +189,11 @@ impl<'a> WalkerOwner<'a> {
 
     fn set_pos(&self, pos: usize) {
         let mut walker = self.walker.borrow_mut();
+        assert!(walker.pos <= pos);
+
         while pos > walker.pos && !walker.is_eof() {
             walker.forward();
         }
-        while pos < walker.pos {
-            walker.backward();
-        }
     }
 
     fn collect_token_trees(&mut self, n: usize) -> Vec<&tt::TokenTree> {
@@ -264,15 +202,16 @@ impl<'a> WalkerOwner<'a> {
         walker.reset();
 
         while walker.pos < n {
-            if let WalkCursor::Token(u, tt) = &walker.cursor {
+            if let WalkCursor::Token(u, _) = &walker.cursor {
                 // We only collect the topmost child
                 if walker.stack.len() == 0 {
-                    for i in 0..tt.n_tokens {
-                        if let DelimToken::Token(token) = walker.ts.get(u + i) {
-                            res.push(token);
-                        }
+                    if let DelimToken::Token(token) = walker.ts.get(*u) {
+                        res.push(token);
                     }
-                } else if walker.stack.len() == 1 {
+                }
+                // Check whether the second level is a subtree
+                // if so, collect its parent which is topmost child
+                else if walker.stack.len() == 1 {
                     if let DelimToken::Delim(_, is_end) = walker.top().get(*u) {
                         if !is_end {
                             let (_, last_idx) = &walker.stack[0];
@@ -343,78 +282,6 @@ impl<'a> TokenSource for SubtreeTokenSource<'a> {
     }
 }
 
-pub(crate) 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) }
-    }
-
-    pub 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))
-    }
-
-    pub 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))
-            }
-        })
-    }
-}
-
-// FIXME: Remove this function
-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 {
-            _ => None,
-        } {
-            return Some((kind, is_joint_to_next, text, 3));
-        }
-    }
-
-    if let Some((m, is_joint_to_next)) = iter.current_punct2(p) {
-        if let Some((kind, text)) = match m {
-            _ => None,
-        } {
-            return Some((kind, is_joint_to_next, text, 2));
-        }
-    }
-
-    None
-}
-
 fn convert_delim(d: tt::Delimiter, closing: bool) -> TtToken {
     let (kinds, texts) = match d {
         tt::Delimiter::Parenthesis => ([L_PAREN, R_PAREN], "()"),
@@ -426,7 +293,7 @@ fn convert_delim(d: tt::Delimiter, closing: bool) -> TtToken {
     let idx = closing as usize;
     let kind = kinds[idx];
     let text = if texts.len() > 0 { &texts[idx..texts.len() - (1 - idx)] } else { "" };
-    TtToken { kind, is_joint_to_next: false, text: SmolStr::new(text), n_tokens: 1 }
+    TtToken { kind, is_joint_to_next: false, text: SmolStr::new(text) }
 }
 
 fn convert_literal(l: &tt::Literal) -> TtToken {
@@ -437,7 +304,7 @@ fn convert_literal(l: &tt::Literal) -> TtToken {
             _ => panic!("Fail to convert given literal {:#?}", &l),
         });
 
-    TtToken { kind, is_joint_to_next: false, text: l.text.clone(), n_tokens: 1 }
+    TtToken { kind, is_joint_to_next: false, text: l.text.clone() }
 }
 
 fn convert_ident(ident: &tt::Ident) -> TtToken {
@@ -447,39 +314,31 @@ fn convert_ident(ident: &tt::Ident) -> TtToken {
         SyntaxKind::from_keyword(ident.text.as_str()).unwrap_or(IDENT)
     };
 
-    TtToken { kind, is_joint_to_next: false, text: ident.text.clone(), n_tokens: 1 }
+    TtToken { kind, is_joint_to_next: false, text: ident.text.clone() }
 }
 
-fn convert_punct(p: &tt::Punct, next_tokens: &[tt::TokenTree]) -> TtToken {
-    let mut iter = next_tokens.iter();
-    iter.next();
-    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 }
-    }
+fn convert_punct(p: &tt::Punct) -> TtToken {
+    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 }
 }
 
-fn convert_leaf(tokens: &[tt::TokenTree], leaf: &tt::Leaf) -> TtToken {
+fn convert_leaf(leaf: &tt::Leaf) -> TtToken {
     match leaf {
         tt::Leaf::Literal(l) => convert_literal(l),
         tt::Leaf::Ident(ident) => convert_ident(ident),
-        tt::Leaf::Punct(punct) => convert_punct(punct, tokens),
+        tt::Leaf::Punct(punct) => convert_punct(punct),
     }
 }

crates/ra_mbe/src/syntax_bridge.rs

@@ -148,30 +148,21 @@ fn convert_tt(
         match child {
             SyntaxElement::Token(token) => {
                 if token.kind().is_punct() {
-                    let mut prev = None;
-                    for char in token.text().chars() {
-                        if let Some(char) = prev {
-                            token_trees.push(
-                                tt::Leaf::from(tt::Punct { char, spacing: tt::Spacing::Joint })
-                                    .into(),
-                            );
-                        }
-                        prev = Some(char)
-                    }
-                    if let Some(char) = prev {
-                        let spacing = match child_iter.peek() {
-                            Some(SyntaxElement::Token(token)) => {
-                                if token.kind().is_punct() {
-                                    tt::Spacing::Joint
-                                } else {
-                                    tt::Spacing::Alone
-                                }
-                            }
-                            _ => tt::Spacing::Alone,
-                        };
+                    assert!(token.text().len() == 1, "Input ast::token punct must be single char.");
+                    let char = token.text().chars().next().unwrap();
 
-                        token_trees.push(tt::Leaf::from(tt::Punct { char, spacing }).into());
-                    }
+                    let spacing = match child_iter.peek() {
+                        Some(SyntaxElement::Token(token)) => {
+                            if token.kind().is_punct() {
+                                tt::Spacing::Joint
+                            } else {
+                                tt::Spacing::Alone
+                            }
+                        }
+                        _ => tt::Spacing::Alone,
+                    };
+
+                    token_trees.push(tt::Leaf::from(tt::Punct { char, spacing }).into());
                 } else {
                     let child: tt::TokenTree = if token.kind() == SyntaxKind::TRUE_KW
                         || token.kind() == SyntaxKind::FALSE_KW
@@ -224,6 +215,15 @@ impl<'a, Q: Querier> TtTreeSink<'a, Q> {
     }
 }
 
+fn is_delimiter(kind: SyntaxKind) -> bool {
+    use SyntaxKind::*;
+
+    match kind {
+        L_PAREN | L_BRACK | L_CURLY | R_PAREN | R_BRACK | R_CURLY => true,
+        _ => false,
+    }
+}
+
 impl<'a, Q: Querier> TreeSink for TtTreeSink<'a, Q> {
     fn token(&mut self, kind: SyntaxKind, n_tokens: u8) {
         if kind == L_DOLLAR || kind == R_DOLLAR {
@@ -240,14 +240,18 @@ impl<'a, Q: Querier> TreeSink for TtTreeSink<'a, Q> {
         self.buf.clear();
         self.inner.token(kind, text);
 
-        // // Add a white space to token
-        // let (last_kind, _, last_joint_to_next ) = self.src_querier.token(self.token_pos-n_tokens as usize);
-        // if !last_joint_to_next && last_kind.is_punct() {
-        //     let (cur_kind, _, _ ) = self.src_querier.token(self.token_pos);
-        //     if cur_kind.is_punct() {
-        //         self.inner.token(WHITESPACE, " ".into());
-        //     }
-        // }
+        // Add a white space between tokens, only if both are not delimiters
+        if !is_delimiter(kind) {
+            let (last_kind, _, last_joint_to_next) = self.src_querier.token(self.token_pos - 1);
+            if !last_joint_to_next && last_kind.is_punct() {
+                let (cur_kind, _, _) = self.src_querier.token(self.token_pos);
+                if !is_delimiter(cur_kind) {
+                    if cur_kind.is_punct() {
+                        self.inner.token(WHITESPACE, " ".into());
+                    }
+                }
+            }
+        }
     }
 
     fn start_node(&mut self, kind: SyntaxKind) {

crates/ra_mbe/src/tt_cursor.rs

@@ -1,6 +1,5 @@
 use crate::ParseError;
 use crate::subtree_parser::Parser;
-use crate::subtree_source::TokenPeek;
 use smallvec::{SmallVec, smallvec};
 
 #[derive(Debug, Clone)]
@@ -262,3 +261,48 @@ impl<'a> TtCursor<'a> {
         self.pos = memento.pos;
     }
 }
+
+pub(crate) 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) }
+    }
+
+    pub 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))
+    }
+
+    pub 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))
+            }
+        })
+    }
+}