internal: Lay basic ground work for standalone mbe tests

crates/syntax-bridge/src/insert_whitespace_into_node.rs

@@ -0,0 +1,134 @@
+//! Utilities for formatting macro expanded nodes until we get a proper formatter.
+use syntax::{
+    ast::make,
+    ted::{self, Position},
+    NodeOrToken,
+    SyntaxKind::{self, *},
+    SyntaxNode, SyntaxToken, WalkEvent, T,
+};
+
+/// Renders a [`SyntaxNode`] with whitespace inserted between tokens that require them.
+pub fn insert_ws_into(syn: SyntaxNode) -> SyntaxNode {
+    let mut indent = 0;
+    let mut last: Option<SyntaxKind> = None;
+    let mut mods = Vec::new();
+    let syn = syn.clone_subtree().clone_for_update();
+
+    let before = Position::before;
+    let after = Position::after;
+
+    let do_indent = |pos: fn(_) -> Position, token: &SyntaxToken, indent| {
+        (pos(token.clone()), make::tokens::whitespace(&" ".repeat(4 * indent)))
+    };
+    let do_ws = |pos: fn(_) -> Position, token: &SyntaxToken| {
+        (pos(token.clone()), make::tokens::single_space())
+    };
+    let do_nl = |pos: fn(_) -> Position, token: &SyntaxToken| {
+        (pos(token.clone()), make::tokens::single_newline())
+    };
+
+    for event in syn.preorder_with_tokens() {
+        let token = match event {
+            WalkEvent::Enter(NodeOrToken::Token(token)) => token,
+            WalkEvent::Leave(NodeOrToken::Node(node))
+                if matches!(
+                    node.kind(),
+                    ATTR | MATCH_ARM | STRUCT | ENUM | UNION | FN | IMPL | MACRO_RULES
+                ) =>
+            {
+                if indent > 0 {
+                    mods.push((
+                        Position::after(node.clone()),
+                        make::tokens::whitespace(&" ".repeat(4 * indent)),
+                    ));
+                }
+                if node.parent().is_some() {
+                    mods.push((Position::after(node), make::tokens::single_newline()));
+                }
+                continue;
+            }
+            _ => continue,
+        };
+        let tok = &token;
+
+        let is_next = |f: fn(SyntaxKind) -> bool, default| -> bool {
+            tok.next_token().map(|it| f(it.kind())).unwrap_or(default)
+        };
+        let is_last =
+            |f: fn(SyntaxKind) -> bool, default| -> bool { last.map(f).unwrap_or(default) };
+
+        match tok.kind() {
+            k if is_text(k)
+                && is_next(|it| !it.is_punct() || matches!(it, T![_] | T![#]), false) =>
+            {
+                mods.push(do_ws(after, tok));
+            }
+            L_CURLY if is_next(|it| it != R_CURLY, true) => {
+                indent += 1;
+                if is_last(is_text, false) {
+                    mods.push(do_ws(before, tok));
+                }
+
+                mods.push(do_indent(after, tok, indent));
+                mods.push(do_nl(after, tok));
+            }
+            R_CURLY if is_last(|it| it != L_CURLY, true) => {
+                indent = indent.saturating_sub(1);
+
+                if indent > 0 {
+                    mods.push(do_indent(before, tok, indent));
+                }
+                mods.push(do_nl(before, tok));
+            }
+            R_CURLY => {
+                if indent > 0 {
+                    mods.push(do_indent(after, tok, indent));
+                }
+                mods.push(do_nl(after, tok));
+            }
+            LIFETIME_IDENT if is_next(is_text, true) => {
+                mods.push(do_ws(after, tok));
+            }
+            AS_KW | DYN_KW | IMPL_KW | CONST_KW | MUT_KW => {
+                mods.push(do_ws(after, tok));
+            }
+            T![;] if is_next(|it| it != R_CURLY, true) => {
+                if indent > 0 {
+                    mods.push(do_indent(after, tok, indent));
+                }
+                mods.push(do_nl(after, tok));
+            }
+            T![=] if is_next(|it| it == T![>], false) => {
+                // FIXME: this branch is for `=>` in macro_rules!, which is currently parsed as
+                // two separate symbols.
+                mods.push(do_ws(before, tok));
+                mods.push(do_ws(after, &tok.next_token().unwrap()));
+            }
+            T![->] | T![=] | T![=>] => {
+                mods.push(do_ws(before, tok));
+                mods.push(do_ws(after, tok));
+            }
+            T![!] if is_last(|it| it == MACRO_RULES_KW, false) && is_next(is_text, false) => {
+                mods.push(do_ws(after, tok));
+            }
+            _ => (),
+        }
+
+        last = Some(tok.kind());
+    }
+
+    for (pos, insert) in mods {
+        ted::insert(pos, insert);
+    }
+
+    if let Some(it) = syn.last_token().filter(|it| it.kind() == SyntaxKind::WHITESPACE) {
+        ted::remove(it);
+    }
+
+    syn
+}
+
+fn is_text(k: SyntaxKind) -> bool {
+    // Consider all keywords in all editions.
+    k.is_any_identifier() || k.is_literal() || k == UNDERSCORE
+}

crates/syntax-bridge/src/lib.rs

@@ -17,6 +17,7 @@ use tt::{
     token_to_literal,
 };
 
+pub mod insert_whitespace_into_node;
 mod to_parser_input;
 pub use to_parser_input::to_parser_input;
 // FIXME: we probably should re-think  `token_tree_to_syntax_node` interfaces