Only treat plain literal patterns as short

[oli-obk]

See #134228 (comment) and #134228 (comment) for context. We never wanted to treat const blocks and paths as short, only plain literals.

I don't know how to write a test for this, it.s not clear to me how the short pattern check actually affects the formatting

[oli-obk]

r? @ytmimi

[ytmimi]

As you mentioned in #134228 (comment), ConstBlock and Path patterns snuck into rustfmt unexpectedly. Let me know if I'm misunderstanding anything, but assuming that's the case then we'll need to gate these changes on the style edition:

Accessing the style_edition requires us to add some additional context to is_short_pattern and is_short_pattern_inner. We can either pass the entire &RewriteContext<'_>, or just the &Config (context.config). The &RewriteContext<'_> is available at the call site so this shouldn't be too difficult.

 ///     - `[small, ntp]`
 ///     - unary tuple constructor `([small, ntp])`
 ///     - `&[small]`
-pub(crate) fn is_short_pattern(pat: &ast::Pat, pat_str: &str) -> bool {
+pub(crate) fn is_short_pattern(context: &RewriteContext<'_>, pat: &ast::Pat, pat_str: &str) -> bool {
     // We also require that the pattern is reasonably 'small' with its literal width.
-    pat_str.len() <= 20 && !pat_str.contains('\n') && is_short_pattern_inner(pat)
+    pat_str.len() <= 20 && !pat_str.contains('\n') && is_short_pattern_inner(context, pat)
 }
 
-fn is_short_pattern_inner(pat: &ast::Pat) -> bool {
+fn is_short_pattern_inner(context: &RewriteContext<'_>, pat: &ast::Pat) -> bool {

Then we can gate the change:

Suggested change

	ast::ExprKind::ConstBlock(_) => false,
	ast::ExprKind::Path(..) => false,
	ast::ExprKind::ConstBlock(_) | ast::ExprKind::Path(..) => {
	// Let's leave a comment explaining why this needs to be gated
	context.config.style_edition() <= StyleEdition::Edition2024
	}

[ytmimi]

I don't know how to write a test for this, it.s not clear to me how the short pattern check actually affects the formatting

Yeah, I think this will be tougher to test. Especially because I think we need to gate this change on the style edition as to not change any behavior with stable formatting. Assuming we don't change any behavior for style editions <= 2024, then I think we should be good.