Revert to original copying scheme

manimlib/mobject/mobject.py

@@ -464,40 +464,76 @@ class Mobject(object):
 
     # Copying and serialization
 
+    def stash_mobject_pointers(func):
+        @wraps(func)
+        def wrapper(self, *args, **kwargs):
+            uncopied_attrs = ["parents", "target", "saved_state"]
+            stash = dict()
+            for attr in uncopied_attrs:
+                if hasattr(self, attr):
+                    value = getattr(self, attr)
+                    stash[attr] = value
+                    null_value = [] if isinstance(value, Iterable) else None
+                    setattr(self, attr, null_value)
+            result = func(self, *args, **kwargs)
+            self.__dict__.update(stash)
+            return result
+        return wrapper
+
+    @stash_mobject_pointers
     def serialize(self):
-        pre, self.parents = self.parents, []
-        result = pickle.dumps(self)
-        self.parents = pre
-        return result
+        return pickle.dumps(self)
 
     def deserialize(self, data: bytes):
         self.become(pickle.loads(data))
         return self
 
-    def copy(self):
-        try:
-            serial = self.serialize()
-            return pickle.loads(serial)
-        except AttributeError:
-            return copy.deepcopy(self)
+    @stash_mobject_pointers
+    def copy(self, deep: bool = False):
+        if deep:
+            try:
+                # Often faster than deepcopy
+                return pickle.loads(self.serialize())
+            except AttributeError:
+                return copy.deepcopy(self)
+
+        result = copy.copy(self)
+
+        # The line above is only a shallow copy, so the internal
+        # data which are numpyu arrays or other mobjects still
+        # need to be further copied.
+        result.data = dict(self.data)
+        for key in result.data:
+            result.data[key] = result.data[key].copy()
+
+        result.uniforms = dict(self.uniforms)
+        for key in result.uniforms:
+            if isinstance(result.uniforms[key], np.ndarray):
+                result.uniforms[key] = result.uniforms[key].copy()
+
+        result.submobjects = []
+        result.add(*(sm.copy() for sm in self.submobjects))
+        result.match_updaters(self)
+
+        family = self.get_family()
+        for attr, value in list(self.__dict__.items()):
+            if isinstance(value, Mobject) and value in family and value is not self:
+                setattr(result, attr, result.family[self.family.index(value)])
+            if isinstance(value, np.ndarray):
+                setattr(result, attr, value.copy())
+            if isinstance(value, ShaderWrapper):
+                setattr(result, attr, value.copy())
         return result
 
     def deepcopy(self):
-        # This used to be different from copy, so is now just here for backward compatibility
-        return self.copy()
+        return self.copy(deep=True)
 
     def generate_target(self, use_deepcopy: bool = False):
-        # TODO, remove now pointless use_deepcopy arg
-        self.target = None  # Prevent exponential explosion
-        self.target = self.copy()
+        self.target = self.copy(deep=use_deepcopy)
         return self.target
 
     def save_state(self, use_deepcopy: bool = False):
-        # TODO, remove now pointless use_deepcopy arg
-        if hasattr(self, "saved_state"):
-            # Prevent exponential growth of data
-            self.saved_state = None
-        self.saved_state = self.copy()
+        self.saved_state = self.copy(deep=use_deepcopy)
         return self
 
     def restore(self):
@@ -540,9 +576,8 @@ class Mobject(object):
     # Creating new Mobjects from this one
 
     def replicate(self, n: int) -> Group:
-        serial = self.serialize()
         group_class = self.get_group_class()
-        return group_class(*(pickle.loads(serial) for _ in range(n)))
+        return group_class(*(self.copy() for _ in range(n)))
 
     def get_grid(self, n_rows: int, n_cols: int, height: float | None = None, **kwargs) -> Group:
         """