V0.30.1 - bugfix tag key is field (#150)

* Fix: dataclasses in `Union` when `Meta.tag_key` is a dataclass field

* Fix: dataclasses in `Union` when `Meta.tag_key` is a dataclass field

* Update HISTORY.rst

Author: rnag

HISTORY.rst

@@ -2,6 +2,14 @@
 History
 =======
 
+0.30.1 (2024-11-25)
+-------------------
+
+**Bugfixes**
+
+* Resolved inconsistent behavior with dataclasses in ``Union`` when ``Meta`` :attr:`tag_key`
+  is also defined as a dataclass field (:issue:`148`).
+
 0.30.0 (2024-11-25)
 -------------------
 

dataclass_wizard/class_helper.pyi

@@ -7,6 +7,7 @@ from .bases import META
 from .models import Condition
 from .type_def import ExplicitNullType, T
 from .utils.dict_helper import DictWithLowerStore
+from .utils.object_path import PathType
 
 
 # A cached mapping of dataclass to the list of fields, as returned by
@@ -31,21 +32,18 @@ CLASS_TO_DUMPER: dict[type, type[AbstractDumper]] = {}
 
 # A cached mapping of a dataclass to each of its case-insensitive field names
 # and load hook.
-FIELD_NAME_TO_LOAD_PARSER: dict[
-    type, DictWithLowerStore[str, AbstractParser]] = {}
+FIELD_NAME_TO_LOAD_PARSER: dict[type, DictWithLowerStore[str, AbstractParser]] = {}
 
 # Since the dump process doesn't use Parsers currently, we use a sentinel
 # mapping to confirm if we need to setup the dump config for a dataclass
 # on an initial run.
 IS_DUMP_CONFIG_SETUP: dict[type, bool] = {}
 
 # A cached mapping, per dataclass, of JSON field to instance field name
-JSON_FIELD_TO_DATACLASS_FIELD: dict[
-    type, dict[str, str | ExplicitNullType]] = defaultdict(dict)
+JSON_FIELD_TO_DATACLASS_FIELD: dict[type, dict[str, str | ExplicitNullType]] = defaultdict(dict)
 
 # A cached mapping, per dataclass, of instance field name to JSON path
-DATACLASS_FIELD_TO_JSON_PATH: dict[
-    type, dict[str, list[str | int | bool | float]]] = defaultdict(dict)
+DATACLASS_FIELD_TO_JSON_PATH: dict[type, dict[str, PathType]] = defaultdict(dict)
 
 # A cached mapping, per dataclass, of instance field name to JSON field
 DATACLASS_FIELD_TO_JSON_FIELD: dict[type, dict[str, str]] = defaultdict(dict)
@@ -56,22 +54,20 @@ DATACLASS_FIELD_TO_SKIP_IF: dict[type, dict[str, Condition]] = defaultdict(dict)
 # A mapping of dataclass name to its Meta initializer (defined in
 # :class:`bases.BaseJSONWizardMeta`), which is only set when the
 # :class:`JSONSerializable.Meta` is sub-classed.
-META_INITIALIZER: dict[
-    str, Callable[[type[W]], None]] = {}
-
+META_INITIALIZER: dict[str, Callable[[type[W]], None]] = {}
 
 # Mapping of dataclass to its Meta inner class, which will only be set when
 # the :class:`JSONSerializable.Meta` is sub-classed.
 _META: dict[type, META] = {}
 
 
-def dataclass_to_loader(cls: type):
+def dataclass_to_loader(cls: type) -> type[AbstractLoader]:
     """
     Returns the loader for a dataclass.
     """
 
 
-def dataclass_to_dumper(cls: type):
+def dataclass_to_dumper(cls: type) -> type[AbstractDumper]:
     """
     Returns the dumper for a dataclass.
     """
@@ -89,13 +85,13 @@ def set_class_dumper(cls: type, dumper: type[AbstractDumper]):
     """
 
 
-def json_field_to_dataclass_field(cls: type):
+def json_field_to_dataclass_field(cls: type) -> dict[str, str | ExplicitNullType]:
     """
     Returns a mapping of JSON field to dataclass field.
     """
 
 
-def dataclass_field_to_json_path(cls: type):
+def dataclass_field_to_json_path(cls: type) -> dict[str, PathType]:
     """
     Returns a mapping of dataclass field to JSON path.
     """
@@ -151,7 +147,7 @@ def _setup_load_config_for_cls(cls_loader: type[AbstractLoader],
     """
 
 
-def setup_dump_config_for_cls_if_needed(cls: type):
+def setup_dump_config_for_cls_if_needed(cls: type) -> None:
     """
     This function processes a class `cls` on an initial run, and sets up the
     dump process for `cls` by iterating over each dataclass field. For each
@@ -172,7 +168,7 @@ def setup_dump_config_for_cls_if_needed(cls: type):
     """
 
 
-def call_meta_initializer_if_needed(cls: type[W]):
+def call_meta_initializer_if_needed(cls: type[W]) -> None:
     """
     Calls the Meta initializer when the inner :class:`Meta` is sub-classed.
     """
@@ -186,31 +182,31 @@ def get_meta(cls: type) -> META:
     """
 
 
-def dataclass_fields(cls) -> tuple[Field, ...]:
+def dataclass_fields(cls: type) -> tuple[Field, ...]:
     """
     Cache the `dataclasses.fields()` call for each class, as overall that
     ends up around 5x faster than making a fresh call each time.
 
     """
 
 
-def dataclass_init_fields(cls) -> tuple[Field, ...]:
+def dataclass_init_fields(cls: type) -> tuple[Field, ...]:
     """Get only the dataclass fields that would be passed into the constructor."""
 
 
-def dataclass_field_names(cls) -> tuple[str, ...]:
+def dataclass_field_names(cls: type) -> tuple[str, ...]:
     """Get the names of all dataclass fields"""
 
 
-def dataclass_field_to_default(cls) -> dict[str, Any]:
+def dataclass_field_to_default(cls: type) -> dict[str, Any]:
     """Get default values for the (optional) dataclass fields."""
 
 
-def is_builtin_class(cls):
+def is_builtin_class(cls: type) -> bool:
     """Check if a class is a builtin in Python."""
 
 
-def is_builtin(o: Any):
+def is_builtin(o: Any) -> bool:
     """Check if an object/singleton/class is a builtin in Python."""
 
 
@@ -227,7 +223,7 @@ def get_class_name(class_or_instance) -> str:
     """Return the fully qualified name of a class."""
 
 
-def get_outer_class_name(inner_cls, default=None, raise_=True):
+def get_outer_class_name(inner_cls, default=None, raise_: bool = True) -> str:
     """
     Attempt to return the fully qualified name of the outer (enclosing) class,
     given a reference to the inner class.
@@ -239,11 +235,11 @@ def get_outer_class_name(inner_cls, default=None, raise_=True):
     """
 
 
-def get_class(obj):
+def get_class(obj: Any) -> type:
     """Get the class for an object `obj`"""
 
 
-def is_subclass(obj, base_cls: type) -> bool:
+def is_subclass(obj: Any, base_cls: type) -> bool:
     """Check if `obj` is a sub-class of `base_cls`"""
 
 

dataclass_wizard/dumpers.py

@@ -431,6 +431,8 @@ def dump_func_for_dataclass(cls: Type[T],
             show_deprecation_warning(_pre_dict, reason)
 
             _locals['__pre_dict__'] = _pre_dict
+
+            # Call the optional hook that runs before we process the dataclass
             fn_gen.add_line('__pre_dict__(o)')
 
         # Initialize result list to hold field mappings

dataclass_wizard/loaders.py

@@ -643,7 +643,11 @@ def load_func_for_dataclass(
     # Fix for using `auto_assign_tags` and `raise_on_unknown_json_key` together
     # See https://github.com/rnag/dataclass-wizard/issues/137
     has_tag_assigned = meta.tag is not None
-    if has_tag_assigned:
+    if (has_tag_assigned and
+            # Ensure `tag_key` isn't a dataclass field before assigning an
+            # `ExplicitNull`, as assigning it directly can cause issues.
+            # See https://github.com/rnag/dataclass-wizard/issues/148
+            meta.tag_key not in field_to_parser):
         json_to_field[meta.tag_key] = ExplicitNull
 
     _locals = {
@@ -743,6 +747,7 @@ def load_func_for_dataclass(
                                 fn_gen.add_line("field = json_to_field[json_key] = ExplicitNull")
                                 fn_gen.add_line("LOG.warning('JSON field %r missing from dataclass schema, "
                                                 "class=%r, parsed field=%r',json_key,cls,py_field)")
+
                                 # Raise an error here (if needed)
                                 if meta.raise_on_unknown_json_key:
                                     _globals['UnknownJSONKey'] = UnknownJSONKey
@@ -759,6 +764,11 @@ def load_func_for_dataclass(
                         with fn_gen.except_(ParseError, 'e'):
                             # We run into a parsing error while loading the field value;
                             # Add additional info on the Exception object before re-raising it.
+                            #
+                            # First confirm these values are not already set by an
+                            # inner dataclass. If so, it likely makes it easier to
+                            # debug the cause. Note that this should already be
+                            # handled by the `setter` methods.
                             fn_gen.add_line("e.class_name, e.field_name, e.json_object = cls, field, o")
                             fn_gen.add_line("raise")
 
@@ -772,9 +782,11 @@ def load_func_for_dataclass(
                                 fn_gen.add_line(line)
 
             with fn_gen.except_(TypeError):
-                # If the object `o` is None, then raise an error with the relevant info included.
+                # If the object `o` is None, then raise an error with
+                # the relevant info included.
                 with fn_gen.if_('o is None'):
                     fn_gen.add_line("raise MissingData(cls) from None")
+
                 # Check if the object `o` is some other type than what we expect -
                 # for example, we could be passed in a `list` type instead.
                 with fn_gen.if_('not isinstance(o, dict)'):
@@ -784,17 +796,20 @@ def load_func_for_dataclass(
                 # Else, just re-raise the error.
                 fn_gen.add_line("raise")
 
-        # Now pass the arguments to the constructor method, and return the new dataclass instance.
-        # If there are any missing fields, we raise them here.
         if has_catch_all:
             if catch_all_field.endswith('?'):  # Default value
                 with fn_gen.if_('catch_all'):
                     fn_gen.add_line(f'init_kwargs[{catch_all_field.rstrip("?")!r}] = catch_all')
             else:
                 fn_gen.add_line(f'init_kwargs[{catch_all_field!r}] = catch_all')
 
+        # Now pass the arguments to the constructor method, and return
+        # the new dataclass instance. If there are any missing fields,
+        # we raise them here.
+
         with fn_gen.try_():
             fn_gen.add_line("return cls(**init_kwargs)")
+
         with fn_gen.except_(TypeError, 'e'):
             fn_gen.add_line("raise MissingFields(e, o, cls, init_kwargs, cls_fields) from None")
 

tests/unit/test_load.py

@@ -2441,3 +2441,37 @@ class Example(JSONWizard):
 
         # Attempt to serialize an instance, which should raise the error.
         Example(my_field=3).to_dict()
+
+
+def test_dataclass_in_union_when_tag_key_is_field():
+    """
+    Test case for dataclasses in `Union` when the `Meta.tag_key` is a dataclass field.
+    """
+    @dataclass
+    class DataType(JSONWizard):
+        id: int
+        type: str
+
+    @dataclass
+    class XML(DataType):
+        class _(JSONWizard.Meta):
+            tag = "xml"
+
+        field_type_1: str
+
+    @dataclass
+    class HTML(DataType):
+        class _(JSONWizard.Meta):
+            tag = "html"
+
+        field_type_2: str
+
+    @dataclass
+    class Result(JSONWizard):
+        class _(JSONWizard.Meta):
+            tag_key = "type"
+
+        data: Union[XML, HTML]
+
+    t1 = Result.from_dict({"data": {"id": 1, "type": "xml", "field_type_1": "value"}})
+    assert t1 == Result(data=XML(id=1, type='xml', field_type_1='value'))