Summary

This PR fixes ReferenceError: Cannot access before initialization errors that occur when using methods like .check(), .catchall(), .passthrough(), etc. on recursive object schemas.

Problem

When defining recursive objects and using certain methods, the object spread syntax ({...shape}) causes initialization errors:

const Category = z.object({
  id: z.string(),
  get subcategories() {
    return z.array(Category).optional();
  },
}).check((ctx) => {
  // Custom validation logic
});
// ReferenceError: Cannot access 'Category' before initialization

The issue occurs because object spread syntax immediately evaluates all properties, including getters. When the getter references a recursive type that's still being initialized, it triggers a ReferenceError.

Solution

Replace object spread with a new util.objectClone() function that uses Object.create() with Object.getOwnPropertyDescriptors() to properly clone objects while preserving property descriptors (including getters) without evaluating them immediately.

This approach:

• Preserves all property descriptors exactly as they are
• Defers getter evaluation until after initialization is complete
• Maintains the same behavior for non-recursive objects
• Works with the existing shape getter pattern

Changes

• Added util.objectClone() function that properly clones objects with descriptors
• Updated shape getter implementations in object(), strictObject(), and looseObject() to use objectClone() instead of object spread
• Added test case demonstrating recursive objects with .check() method

Testing

All existing tests pass, plus new test case for recursive objects with .check() method.

🤖 Generated with Claude Code