Reorganize Serializer class layout

dumper.py

@@ -1,52 +1,341 @@
 from __future__ import annotations
 
+from abc import ABC, abstractmethod
 from copy import copy
+from dataclasses import dataclass
 from os.path import abspath
+from typing import Any
 
 import bpy
 
 from . import utils
+from . utils import BlenderProperty
 
 
-def get_dumper(bl_object: bpy.types.bpy_struct) -> type[Dumper]:
-    """Get the closest corresponding dumper for a given Blender object using its MRO"""
-    for cls in bl_object.__class__.mro():
-        dumper_map = DumperRegistry().dumper_map
-        if cls in dumper_map:
-            return dumper_map[cls]
+def serialize_selected_nodes_from_node_tree(node_tree: bpy.types.NodeTree):
+    """Serialize the selected nodes from a node tree"""
+    selected_nodes = [node for node in node_tree.nodes if node.select]
+    selected_links = [link for link in node_tree.links if link.from_node.select
+                                                       and link.to_node.select]
 
-    # Fallback to base Dumper if no matches are found
-    return Dumper
+    bl_pointers = {}
+    nodes_data = [Serializer.serialize(node, bl_pointers) for node in selected_nodes]
+    links_data = [Serializer.serialize(link, bl_pointers) for link in selected_links]
+
+    # Only serialize selected nodes and their links
+    ntree_data = {
+        "nodes": nodes_data,
+        "links": links_data,
+    }
+
+    return ntree_data
 
 
-def dump_nodes(nodes: list[bpy.types.Node]):
-    """Generic recursive dump, convert nodes into a dict"""
-    Dumper.pointers.clear()  # TODO: Bad global
+def deserialize_nodes_into_node_tree(data: dict, node_tree: bpy.types.NodeTree):
+    """Deserialize node data into a specific node tree"""
+    bl_pointers = {}
+    Serializer.deserialize(data, node_tree, bl_pointers)
 
-    data = [dump_node(node) for node in nodes]
 
-    Dumper.pointers.clear()
+# TODO: Sub serialize function where the isinstance is set to the default number of things.
+
+# TODO: Collection is not handled as a class anymore, handle it manually
+
+class Serializer(ABC):
+    """
+    Base Serializer class.
+    `bl_pointers_ref` corresponds to a mutable dict passed through serialize/deserialize
+    functions, containing a map of Blender pointers IDs and their corresponding objects.
+    """
+    # Whitelisted properties, applied after the blacklist
+    prop_whitelist = None
+    # Properties that are excluded from (de)serialization, in addition to any bl_* properties
+    prop_blacklist = ("rna_type", "id_data", "depsgraph")
+
+    serializer_map = {}
+
+    @classmethod
+    @abstractmethod
+    def construct_bl_object(cls, data: dict):
+        """Abstract method to construct a Serializer's specific Blender Object"""
+        print("DEBUG: construct_bl_object called on Base Serializer, shouldn't happen")
+        return None
+
+    # --- Serialization ---
+
+    @classmethod
+    @abstractmethod
+    def serialize(cls, obj: bpy.types.bpy_struct | Any, bl_pointers_ref: dict) -> dict:
+        """Base serialization method, overridden by subclasses"""
+
+        # Early recursive return case
+        # TODO: Ported as is, check the heuristics, (if there are more or attributes type to add)
+        if isinstance(obj, (str, int, float, dict, list, type(None))):
+            return obj
+
+        # Returned data, tracks the pointer for later re-assignments during deserialization
+        data = {"_kit_ptr": obj.as_pointer()}
+        bl_pointers_ref[obj.as_pointer()] = obj
+
+        # Iterate over all *filtered* properties found in the object
+        for bl_prop in cls.get_serialized_properties(obj):
+            # Do not store default values nor read-only properties
+            if (array := getattr(bl_prop.rep, "default_array", None)) and bl_prop.attr == array:
+                continue
+            if isinstance(bl_prop.attr, (str, int, float)) and bl_prop.attr == bl_prop.rep.default:
+                continue
+            if obj.is_property_readonly(bl_prop.rep.identifier):
+                continue
+            
+            print(type(bl_prop.attr))
+                
+            # Serialize each property
+            data[bl_prop.rep.identifier] = cls.serialize_property(bl_prop, bl_pointers_ref)
+
+        return data
+
+    @classmethod
+    def serialize_property(cls, bl_prop: BlenderProperty, bl_pointers_ref: dict) -> Any:
+        """Serialize node property, special cases for arrays/collections/pointers"""
+        # Property array case
+        if getattr(bl_prop.rep, "is_array", False):
+            # Contained in BoolProperty, IntProperty and FloatProperty
+            prop_array = []
+            for item in bl_prop.attr:
+                assert isinstance(item, (bool, int, float)) # TODO: For development, replace by list comprehension later
+                prop_array.append(item)
+
+            return prop_array
+        
+        # Collection case
+        if isinstance(bl_prop.attr, bpy.types.bpy_prop_collection):
+            collection = bl_prop.attr
+            if not collection:
+                return []
+            
+            values = [cls.sub_serialize(sub_prop) for sub_prop in collection]
+
+            # TODO: Check why the original code has a None check
+            return [v for v in values if v is not None]
+
+        # Pointer case
+        if bl_prop.rep.type == "POINTER" and bl_prop.attr:
+            # Property points to another object, stores it ptr/deref value in our pointer table
+            ptr = bl_prop.attr.as_pointer()
+
+            if ptr in bl_pointers_ref:
+                return ptr
+            
+            bl_pointers_ref[ptr] = bl_prop.attr
+
+        return cls.sub_serialize(bl_prop.attr, bl_pointers_ref)
     
-    return data
+    @classmethod
+    def sub_serialize(cls, obj: bpy.types.bpy_struct | Any, bl_pointers_ref: dict) -> Any:
+        if not isinstance(obj, bpy.types.bpy_struct):
+            # Primitive type, return directly
+            return obj
 
+        """Resolve which Serializer class to use"""
+        serializer = cls.get_serializer(obj)
 
-def dump_node(node: bpy.types.Node):
-    dumper = get_dumper(node)
-    return dumper.dump(node)
+        return serializer.serialize(obj, bl_pointers_ref)
+    
+    # --- Deserialization ---
+    
+    @classmethod
+    @abstractmethod
+    def deserialize(cls, data: dict, target_obj: bpy.types.bpy_struct, bl_pointers_ref: dict):
+        """
+        Base deserialization method.
+        Deserialize data into a specific Blender object, creating sub-objects as needed.
+        Partial data may be provided, in which case, fields not specified will be left to default.
+        """
 
+        if (kit_ptr := data.get("_kit_ptr", None)):
+            bl_pointers_ref[kit_ptr] = target_obj
+        
+        data_to_deserialize = cls.get_data_to_deserialize(data, target_obj)
+        
+        for stored_key, stored_value in data_to_deserialize:
+            if stored_key.startswith("_kit") or stored_key not in target_obj.bl_rna.properties:
+                continue
 
-def load_nodes(data, node_tree):
-    """Load/Dump nodes into a specific node tree"""
-    Dumper.pointers.clear()
+            target_bl_prop = BlenderProperty(rep=target_obj.bl_rna.properties[stored_key],
+                                             attr=getattr(target_obj, stored_key))
 
-    dumper = get_dumper(node_tree)
-    dumper.load(data, node_tree)
+            # Collection case
+            # Unlike serialization, there's no property array case, as they are just directly assigned
+            if isinstance(target_bl_prop.attr, bpy.types.bpy_prop_collection):
+                cls.deserialize_collection(stored_value, target_bl_prop.attr, bl_pointers_ref)
+                continue
 
-    Dumper.pointers.clear()
+            value_to_set = stored_value
+
+            # Pointer case
+            # Dereference the value if its already present in the pointers_ref map
+            if target_bl_prop.rep.type == "POINTER":
+                value_to_set = cls.deserialize_pointer(stored_value, target_bl_prop.attr, bl_pointers_ref)
+
+            # Skip setting the property if it's read-only
+            if target_bl_prop.rep.is_readonly:
+                continue
+
+            # Assign the property
+            setattr(target_obj, stored_key, value_to_set)
+
+            # If supported, update the Blender property after setting it
+            if hasattr(target_bl_prop.attr, "update"):
+                target_bl_prop.attr.update()
+
+    @classmethod
+    def deserialize_collection(cls, stored_value: Any, bl_coll: bpy.types.bpy_prop_collection, bl_pointers_ref: dict):
+        # Static collection case
+        if not hasattr(bl_coll, "new"):
+            cls.sub_deserialize(stored_value, bl_coll, bl_pointers_ref)
+            return
+
+        # We need to call the collection "new" function, parse and construct its parameters
+        new_func = bl_coll.bl_rna.functions["new"]
+
+        for i, value in enumerate(stored_value):
+            # Using a dictionary of {parameter: parameter_value}
+            default_new_func_params = {
+                    k: value.get(k, utils.get_bl_default(v))
+                    for k, v in new_func.parameters.items()[:-1]
+                }
+            new_func_params = value.get("_kit_new_params", default_new_func_params)
+
+            solved_all_pointers = True
+            for param in bl_coll.bl_rna.functions["new"].parameters:
+                if param.identifier not in new_func_params or param.type != "POINTER":
+                    continue
+
+                pointer_id = param[param.identifier]
+                if bl_object := bl_pointers_ref.get(pointer_id):
+                    new_func_params[param.identifier] = bl_object
+                else:
+                    print(f"No pointer found for param {param.identifier} of new function of {bl_coll}")
+                    solved_all_pointers = False
+
+            # Bail out if we fail to solve all pointers (TODO: I'm guessing this causes a runtimerror, but double check)
+            if not solved_all_pointers:
+                continue
+
+            print("Calling BL collection new with the following parameters")
+            print(new_func_params)
+
+            # Creates a collection item, type from the collection type, no need to manually construct
+            collection_item = bl_coll.new(**new_func_params)
+            deserializer = cls.get_serializer(collection_item)
+
+            # Recursively deserialize into the newly constructured object
+            cls.deserialize(value, collection_item, bl_pointers_ref)
+
+            # Old code guarded by a RuntimeError before, investigate later
+            # Static collection case, would need to check how to detect this.
+            collection_item = bl_coll[i]
+
+    # TODO: The target_bl_prop_attr terminology is unclear
+    @classmethod
+    def deserialize_pointer(cls, stored_value: Any, target_bl_prop_attr: bpy.types.bpy_struct, bl_pointers_ref: dict):
+        if stored_value is None:
+            return None
+
+        # Actual existing pointer, dereference and return
+        if isinstance(stored_value, int):
+            if stored_value not in bl_pointers_ref:
+                print("DEBUG: Pointer reference hasn't been loaded yet")
+            # Obtain a reference to a previously dereferenced object
+            return bl_pointers_ref[stored_value]
+        
+        # Pointer doesn't exist yet, create it if it doesn't exist yet, store it, and return its object
+        deserializer = cls.get_serializer(target_bl_prop_attr)
+
+        # Create the Blender object if it doesn't exist yet
+        if target_bl_prop_attr is None:
+            target_bl_prop_attr = cls.construct_bl_object(stored_value)
+        
+        # Recursively deserialize into the target object
+        deserializer.deserialize(stored_value, target_bl_prop_attr, bl_pointers_ref)
+
+        bl_pointers_ref[stored_value["_kit_ptr"]] = target_bl_prop_attr
+
+        return target_bl_prop_attr
+    
+    @classmethod 
+    def get_data_to_deserialize(cls, data: dict, target_obj: bpy.types.bpy_struct=None):
+        props_to_deserialize = cls.get_serialized_properties(target_obj)
+        
+        sorted_data = sorted(
+            data.items(), key=lambda x: props_to_deserialize.index(x[0]) if x[0] in props_to_deserialize else 0
+        )
+
+        return sorted_data
+
+    # --- Getters for sub-serializers ---
+
+    @classmethod
+    def get_serializer_map(cls) -> dict[type[bpy.types.bpy_struct], type[Serializer]]:
+        """Get the serializer map, stored in a class variable for simple caching"""
+        if not cls.serializer_map:
+            for subclass in utils.all_subclasses(Serializer):
+                assert hasattr(subclass, "bl_type")
+                cls.serializer_map[subclass.bl_type] = subclass
+
+        return cls.serializer_map
+
+    @classmethod
+    def get_serializer(cls, bl_object: bpy.types.bpy_struct) -> type[Serializer]:
+        """Get the closest corresponding serializer for a given Blender object using its MRO"""
+        serializer_map = cls.get_serializer_map()
+
+        bl_type = type(bl_object.bl_rna.type_recast())
+
+        for bl_parents in bl_type.mro():
+            if bl_parents in serializer_map:
+                return serializer_map[bl_parents]
+
+        # Fallback to base Serializer if no matches are found
+        return Serializer 
+    
+    # --- Properties to (de)serialize ---
+    
+    @classmethod
+    def get_serialized_properties(cls, obj: bpy.types.bpy_struct | Any):
+        serialized_properties: list[BlenderProperty] = [
+            BlenderProperty(rep=prop, attr=getattr(obj, prop.identifier))
+            for prop in obj.bl_rna.properties
+            if not prop.identifier.startswith("bl_")  # Exclude internal Blender properties
+            and prop.identifier not in cls.prop_blacklist  # Additional blacklist filtering
+        ]
+
+        if cls.prop_whitelist:  # Additional whitelist, applied after the blacklist
+            serialized_properties: list[BlenderProperty] = [
+                prop for prop in serialized_properties
+                if prop.rep.identifier in cls.prop_whitelist
+            ]
+        
+        return serialized_properties
+
+# class NodeSocket(Serializer):
+#     bl_type = bpy.types.NodeSocket
+#     prop_blacklist = Serializer.prop_blacklist + (
+#         "node",
+#         "links",
+#         "display_shape",
+#         "link_limit",
+#     )
+
+#     @classmethod
+#     def serialize(cls, socket_obj: bpy.types.NodeSocket, _: dict) -> dict:
+#         if socket_obj.is_unavailable:
+#             return None
+        
+#         return super().serialize(socket_obj)
 
 
 class Dumper:
-    pointers = {}
     includes = []
     excludes = ["rna_type", "bl_rna", "id_data", "depsgraph"]
 
@@ -67,7 +356,7 @@ class Dumper:
         print(f"New not implemented for data {data}")
 
     @classmethod
-    def load(cls, data, bl_object=None):
+    def load(cls, data, bl_pointers_ref, bl_object=None):
         if bl_object is None:
             bl_object = cls.new(data)
 
@@ -75,7 +364,7 @@ class Dumper:
             return
 
         if bl_pointer := data.get("bl_pointer"):
-            cls.pointers[bl_pointer] = bl_object
+            bl_pointers_ref[bl_pointer] = bl_object
 
         props = cls.properties(bl_object)
         for key, value in sorted(
@@ -98,9 +387,9 @@ class Dumper:
 
             elif prop.type == "POINTER":
                 if isinstance(value, int):  # It's a pointer
-                    if value not in cls.pointers:
+                    if value not in bl_pointers_ref:
                         print(bl_object, "not loaded yet", prop)
-                    value = cls.pointers[value]
+                    value = bl_pointers_ref[value]
 
                 elif value is None:
                     utils.set_bl_attribute(bl_object, key, value)
@@ -114,7 +403,7 @@ class Dumper:
                         attr = dumper.new(value)
 
                     dumper.load(value, attr)
-                    cls.pointers[value["bl_pointer"]] = attr
+                    bl_pointers_ref[value["bl_pointer"]] = attr
 
                     if hasattr(attr, "update"):
                         attr.update()
@@ -135,12 +424,12 @@ class Dumper:
         # return bl_object
 
     @classmethod
-    def dump(cls, bl_object):
+    def dump(cls, bl_object, bl_pointers_ref):
         if isinstance(bl_object, (str, int, float, dict, list, type(None))):
             return bl_object
 
         data = {"bl_pointer": bl_object.as_pointer()}
-        cls.pointers[bl_object.as_pointer()] = bl_object
+        bl_pointers_ref[bl_object.as_pointer()] = bl_object
 
         for prop in cls.properties(bl_object):
             if not hasattr(bl_object, prop.identifier):
@@ -163,10 +452,10 @@ class Dumper:
                 value = PropCollection.dump(value)
 
             elif prop.type == "POINTER" and value:
-                if value.as_pointer() in cls.pointers:
+                if value.as_pointer() in bl_pointers_ref:
                     value = value.as_pointer()
                 else:
-                    cls.pointers[value.as_pointer()] = value
+                    bl_pointers_ref[value.as_pointer()] = value
                     dumper = get_dumper(value)
                     value = dumper.dump(value)
 
@@ -204,7 +493,7 @@ class PropCollection(Dumper):
         dumper = None
 
         if not hasattr(coll, "new"):  # Static collection
-            for item, value in zip(coll, values):
+            for item, value in zip(coll, values):  # TODO: That zip doesn't make sense, or does it?
                 dumper = dumper or get_dumper(item)
                 dumper.load(value, item)
 
@@ -227,7 +516,7 @@ class PropCollection(Dumper):
                     continue
 
                 pointer_id = params[param.identifier]
-                if bl_object := cls.pointers.get(pointer_id):
+                if bl_object := bl_pointers_ref.get(pointer_id):
                     params[param.identifier] = bl_object
                 else:
                     print(f"No Pointer found for param {param.identifier} of {coll}")
@@ -277,7 +566,7 @@ class NodeSocket(Dumper):
         if socket.is_unavailable:
             return None
 
-        # cls.pointers[socket.as_pointer()] = socket
+        # bl_pointers_ref[socket.as_pointer()] = socket
 
         data = super().dump(socket)
 
@@ -316,7 +605,7 @@ class NodeTreeInterfaceSocket(Dumper):
 
     @classmethod
     def dump(cls, socket):
-        # cls.pointers[socket.as_pointer()] = socket
+        # bl_pointers_ref[socket.as_pointer()] = socket
 
         data = super().dump(socket)
         # data["_id"] = socket.as_pointer()
@@ -340,7 +629,7 @@ class NodeSockets(PropCollection):
 
         node_sockets = [s for s in coll if not s.is_unavailable]
         for socket, value in zip(node_sockets, values):
-            cls.pointers[value["bl_pointer"]] = socket
+            bl_pointers_ref[value["bl_pointer"]] = socket
             Dumper.load(value, socket)
             # for k, v in value.items():
             #     if k not in socket.bl_rna.properties:
@@ -353,7 +642,7 @@ class NodeSockets(PropCollection):
         if len(node_sockets) == len(inputs): # Match by index
             super().load({"inputs": inputs}, node)
             for socket, value in zip(node_sockets, coll):
-                cls.pointers[value['_id']] = socket
+                bl_pointers_ref[value['_id']] = socket
         else: # Match by name
             print(f'Match Inputs by Name for node {node}')
             for socket in node_sockets:
@@ -363,7 +652,7 @@ class NodeSockets(PropCollection):
 
                 value = inputs[index]
                 print(socket, value)
-                cls.pointers[value['_id']] = socket
+                bl_pointers_ref[value['_id']] = socket
 
                 Dumper.load(value, socket)
                 del inputs[index]
@@ -389,7 +678,7 @@ class Node(Dumper):
 
     @classmethod
     def dump(cls, node=None):
-        # cls.pointers[node.as_pointer()] = node
+        # bl_pointers_ref[node.as_pointer()] = node
 
         data = super().dump(node)
         # data["_id"] = node.as_pointer()
@@ -409,7 +698,7 @@ class Node(Dumper):
     def load(cls, data, node):
         if node is None:
             return
-        # cls.pointers[data['bl_pointer']] = node
+        # bl_pointers_ref[data['bl_pointer']] = node
 
         inputs = copy(data.pop("inputs", []))
         outputs = copy(data.pop("outputs", []))
@@ -465,8 +754,8 @@ class Nodes(PropCollection):
         # Pair zone input and output
         for node_data in values:
             if paired_output_id := node_data.get("_pair_with_output", None):
-                node = cls.pointers[node_data["bl_pointer"]]
-                node.pair_with_output(cls.pointers[paired_output_id])
+                node = bl_pointers_ref[node_data["bl_pointer"]]
+                node.pair_with_output(bl_pointers_ref[paired_output_id])
 
                 Dumper.load(
                     {"inputs": node_data["inputs"], "outputs": node_data["outputs"]},
@@ -701,7 +990,7 @@ class CompositorNodeRLayers(Node):
                 view_layer_data[prop.identifier]
                 """
 
-        # cls.pointers[bl_object.as_pointer()] = bl_object
+        # bl_pointers_ref[bl_object.as_pointer()] = bl_object
 
         data["scene"] = {
             "bl_pointer": node.scene.as_pointer(),
@@ -742,22 +1031,4 @@ class ViewLayers(PropCollection):
             if view_layer is None:
                 view_layer = coll.new(value["name"])
 
-            Dumper.load(value, view_layer)
-
-
-class DumperRegistry:
-    """Singleton-like class that holds a map of all parsers, constructed on first instantiation"""
-    dumper_map = None
-    
-    def __init__(self):
-        if self.dumper_map is None:
-            self.construct_dumper_map()
-
-    @classmethod
-    def construct_dumper_map(cls):
-        cls.dumper_map = {}
-
-        for subclass in utils.all_subclasses(Dumper):
-            assert hasattr(subclass, "bl_type")
-            cls.dumper_map[subclass.bl_type] = subclass
-            print(cls.dumper_map)
+            Dumper.load(value, view_layer)

operators.py

@@ -14,7 +14,7 @@ import bpy
 from bpy.props import BoolProperty, EnumProperty
 from bpy.types import Operator
 
-from .dumper import dump_nodes, load_nodes
+from .dumper import serialize_selected_nodes_from_node_tree, deserialize_nodes_into_node_tree
 from .node_utils import remap_node_group_duplicates
 from .pack_nodes import combine_objects, extract_objects
 from .formats import dump_nkit_format, parse_nkit_format
@@ -28,18 +28,12 @@ class NODEKIT_OT_copy(Operator):
 
     def execute(self, context):
         ntree = context.space_data.edit_tree
-        selected_nodes = [node for node in ntree.nodes if node.select]
+        serialized_nodes_data = serialize_selected_nodes_from_node_tree(ntree)
 
-        ntree_data = {
-            "nodes": dump_nodes(selected_nodes),
-            "links": dump_nodes(
-                [l for l in ntree.links if l.from_node.select and l.to_node.select]
-            ),
-        }
+        context.window_manager.clipboard = dump_nkit_format(serialized_nodes_data)
 
-        context.window_manager.clipboard = dump_nkit_format(ntree_data)
-
-        self.report({"INFO"}, f"Copied {len(selected_nodes)} selected nodes to system clipboard")
+        num_selected_nodes = len([n for n in ntree.nodes if n.select])
+        self.report({"INFO"}, f"Copied {num_selected_nodes} selected nodes to system clipboard")
         return {"FINISHED"}
 
 
@@ -55,7 +49,8 @@ class NODEKIT_OT_copy_tree(Operator):
 
         context.window_manager.clipboard = dump_nkit_format(ntree_data)
 
-        self.report({"INFO"}, f"Copied {len(ntree.nodes)} selected nodes to system clipboard")
+        num_nodes = len(ntree.nodes)
+        self.report({"INFO"}, f"Copied {num_nodes} selected nodes to system clipboard")
         return {"FINISHED"}
 
 
@@ -67,7 +62,7 @@ class NODEKIT_OT_paste(Operator):
 
     def execute(self, context):
         ntree_data = parse_nkit_format(context.window_manager.clipboard)
-        load_nodes(ntree_data, context.space_data.edit_tree)
+        deserialize_nodes_into_node_tree(ntree_data, context.space_data.edit_tree)
 
         self.report({"INFO"}, f"X node(s) pasted from system clipboard")  # TODO: Ge the number of parsed nodes returned
         return {"FINISHED"}

utils.py

@@ -1,6 +1,21 @@
+from dataclasses import dataclass
+from typing import Any
+
 import bpy
 
 
+
+@dataclass
+class BlenderProperty:
+    """
+    Blender Property abstraction, used since a Blender property value isn't
+    directly accessible from its Property object representation
+    NOTE: Do not rely on value being up-to-date, data will get stall
+    """
+    rep: bpy.types.Property
+    attr: Any
+
+
 def all_subclasses(cls):
     return set(cls.__subclasses__()).union(
         [s for c in cls.__subclasses__() for s in all_subclasses(c)]