Fix ndjson serialization / deserialization

Author: Val Brodsky

libs/labelbox/src/labelbox/data/annotation_types/label.py

@@ -205,16 +205,14 @@ def _assign_option(self, classification: ClassificationAnnotation,
 
     @field_validator("annotations", mode='before')
     def validate_union(cls, value):
-        supported = tuple([
-            field.type_ for field in
-            cls.__model_fields__['annotations'].sub_fields[0].sub_fields
-        ])
         if not isinstance(value, list):
             raise TypeError(f"Annotations must be a list. Found {type(value)}")
 
+        supported_types = cls.model_fields['annotations'].annotation.__args__[
+            0].__args__
         for v in value:
-            if not isinstance(v, supported):
+            if not isinstance(v, supported_types):
                 raise TypeError(
-                    f"Annotations should be a list containing the following classes : {supported}. Found {type(v)}"
+                    f"Annotations should be a list containing the following classes : {supported_types}. Found {type(v)}"
                 )
         return value

libs/labelbox/src/labelbox/data/serialization/ndjson/base.py

@@ -10,12 +10,12 @@ class DataRow(_CamelCaseMixin):
     id: str = None
     global_key: str = None
 
-    @model_validator(mode='before')
-    @classmethod
-    def must_set_one(cls, values):
-        if not is_exactly_one_set(values.get('id'), values.get('global_key')):
+    @model_validator(mode='after')
+    def must_set_one(self):
+        if not is_exactly_one_set(self.id, self.global_key):
             raise ValueError("Must set either id or global_key")
-        return values
+
+        return self
 
 
 class NDJsonBase(_CamelCaseMixin):
@@ -44,13 +44,13 @@ class NDAnnotation(NDJsonBase):
     page: Optional[int] = None
     unit: Optional[str] = None
 
-    @model_validator(mode='before')
-    @classmethod
-    def must_set_one(cls, values):
-        if ('schema_id' not in values or values['schema_id']
-                is None) and ('name' not in values or values['name'] is None):
+    @model_validator(mode='after')
+    def must_set_one(self):
+        if self.schema_id is None and self.name is None:
             raise ValueError("Schema id or name are not set. Set either one.")
-        return values
+        if self.schema_id is not None and self.name is not None:
+            raise ValueError("Schema id and name are both set. Set only one.")
+        return self
 
     def dict(self, *args, **kwargs):
         res = super().dict(*args, **kwargs)

libs/labelbox/src/labelbox/data/serialization/ndjson/classification.py

@@ -17,13 +17,14 @@ class NDAnswer(ConfidenceMixin, CustomMetricsMixin):
     schema_id: Optional[Cuid] = None
     classifications: Optional[List['NDSubclassificationType']] = []
 
-    @model_validator(mode='before')
-    @classmethod
-    def must_set_one(cls, values):
-        if ('schema_id' not in values or values['schema_id']
-                is None) and ('name' not in values or values['name'] is None):
+    @model_validator(mode='after')
+    def must_set_one(self):
+        if self.schema_id is None and self.name is None:
             raise ValueError("Schema id or name are not set. Set either one.")
-        return values
+        if self.schema_id is not None and self.name is not None:
+            raise ValueError("Schema id and name are both set. Set only one.")
+
+        return self
 
     def dict(self, *args, **kwargs):
         res = super().dict(*args, **kwargs)

libs/labelbox/src/labelbox/schema/data_row_metadata.py

@@ -35,8 +35,8 @@ class DataRowMetadataSchema(BaseModel):
         StringConstraints(strip_whitespace=True, min_length=1, max_length=100)]
     reserved: bool
     kind: DataRowMetadataKind
-    options: Optional[List["DataRowMetadataSchema"]]
-    parent: Optional[SchemaId]
+    options: Optional[List["DataRowMetadataSchema"]] = []
+    parent: Optional[SchemaId] = None
 
 
 DataRowMetadataSchema.update_forward_refs()