Don't expose hir::Path out of hir

Conjecture: it's impossible to use hir::Path *correctly* from an IDE.

I am not entirely sure about this, and we might need to add it back at
some point, but I have to arguments that convince me that we probably
won't:

* `hir::Path` has to know about hygiene, which an IDE can't set up
  properly.

* `hir::Path` lacks identity, but you actually have to know identity
  to resolve it correctly

1 files changed, 6 insertions, 1 deletions

diff --git a/crates/assists/src/handlers/extract_struct_from_enum_variant.rs b/crates/assists/src/handlers/extract_struct_from_enum_variant.rs
index 4bcdae7ba..d62e06b4a 100644
--- a/crates/assists/src/handlers/extract_struct_from_enum_variant.rs
+++ b/crates/assists/src/handlers/extract_struct_from_enum_variant.rs
@@ -106,7 +106,12 @@ fn insert_import(
     if let Some(mut mod_path) = mod_path {
         mod_path.segments.pop();
         mod_path.segments.push(variant_hir_name.clone());
-        insert_use_statement(path.syntax(), &mod_path, ctx, builder.text_edit_builder());
+        insert_use_statement(
+            path.syntax(),
+            &mod_path.to_string(),
+            ctx,
+            builder.text_edit_builder(),
+        );
     }
     Some(())
 }