fix(formatter): keep source parens around html-in-js placeholders past comments (#28381)

It's a rare case, but just to be safe.

apps/oxfmt/conformance/fixtures/edge-cases/html-in-js/placeholder-parens-with-comment.js

@@ -0,0 +1,3 @@
+html`<script>f((${a}) /* c */)</script>`;
+html`<script>f((${a}))</script>`;
+html`<script>f(/* c */ (${a}))</script>`;

apps/oxfmt/conformance/snapshots/conformance.snap.md

@@ -74,7 +74,7 @@
 
 ## html-in-js
 
-### Option 1: 172/200 (86.00%)
+### Option 1: 173/201 (86.07%)
 
 ```json
 {"printWidth":80}
@@ -137,7 +137,7 @@
 - [externals/webawesome/textarea/textarea.ts](diffs/html-in-js/externals__webawesome__textarea__textarea.ts.md)
   - crates/oxc_formatter/DIVERGENCES.md#union-annotation-flat-retry
 
-### Option 2: 186/200 (93.00%)
+### Option 2: 187/201 (93.03%)
 
 ```json
 {"printWidth":100,"htmlWhitespaceSensitivity":"ignore"}

apps/oxfmt/conformance/snapshots/diffs/html-in-js/edge-cases__html-in-js__placeholder-case-in-child.js.md

@@ -74,7 +74,7 @@ const d = html`<style>
  const a = html`
    <style>
      a {
--      prettier_html_placeholder_0_2989_in_js: ${c};
+-      prettier_html_placeholder_0_2992_in_js: ${c};
 +      ${b}: ${c};
      }
    </style>
@@ -110,7 +110,7 @@ const d = html`
 const a = html`
   <style>
     a {
-      prettier_html_placeholder_0_2989_in_js: ${c};
+      prettier_html_placeholder_0_2992_in_js: ${c};
     }
   </style>
 `;

crates/oxc_formatter/src/parentheses/expression.rs

@@ -144,23 +144,33 @@ impl NeedsParentheses<'_> for AstNode<'_, '_, IdentifierReference<'_>> {
 
 /// Whether the source wraps `node` in parentheses of its own (the AST keeps none),
 /// besides those of the syntax around it (a sole call argument, an `if` test, ...).
+/// Comments between them are skipped (`f((a) /* c */)`).
+#[expect(clippy::cast_possible_truncation)] // Offsets fit in `u32`, source length does
 fn is_parenthesized_in_source(
     node: &AstNode<'_, '_, IdentifierReference<'_>>,
     f: &JsFormatter<'_, '_>,
 ) -> bool {
     let bytes = f.source_text().as_str().as_bytes();
-    let (mut start, mut end) = (node.span.start as usize, node.span.end as usize);
-    let mut pairs = 0;
-    loop {
-        let before = bytes[..start].iter().rposition(|b| !b.is_ascii_whitespace());
-        let after = bytes[end..].iter().position(|b| !b.is_ascii_whitespace()).map(|i| end + i);
-        match (before, after) {
-            (Some(before), Some(after)) if bytes[before] == b'(' && bytes[after] == b')' => {
-                pairs += 1;
-                (start, end) = (before, after + 1);
+    let comments = f.context().comments();
+    // The last byte before `pos` that is neither whitespace nor in a comment
+    let before = |mut pos: u32| -> Option<u32> {
+        loop {
+            let i = bytes[..pos as usize].iter().rposition(|b| !b.is_ascii_whitespace())? as u32;
+            match comments.all_comments_before(pos).last() {
+                Some(comment) if comment.span.end > i => pos = comment.span.start,
+                _ => return Some(i),
             }
-            _ => break,
         }
+    };
+    let (mut start, mut end) = (node.span.start, node.span.end);
+    let mut pairs = 0;
+    while let Some(open) = before(start)
+        && bytes[open as usize] == b'('
+        && comments.is_followed_by_closing_paren(end)
+    {
+        pairs += 1;
+        start = open;
+        end = comments.position_after_character(end, b')');
     }
     let span = node.span;
     let syntax_parens = match node.parent() {