No more silent fallbacks to JsonDataConverter

Author: andystaples

CHANGELOG.md

@@ -65,16 +65,23 @@ DEPRECATED
   `JsonDataConverter`) instead. The functions continue to work for backwards
   compatibility.
 
-BREAKING CHANGES (type-level only — no runtime impact for typical users)
+BREAKING CHANGES (no runtime impact for typical users)
 
-These changes do not alter runtime behavior, but because the package ships
-`py.typed`, consumers running strict type checkers (pyright/mypy) — or
-subclassing the public abstract types — may need to update their code:
+Most of these are type-level only: because the package ships `py.typed`,
+consumers running strict type checkers (pyright/mypy) — or subclassing the
+public abstract types — may need to update their code. The constructor change
+below also affects callers who *directly* construct the named classes, which is
+uncommon since they are normally handed to you by the SDK.
 
 - `OrchestrationContext.call_activity`, `call_sub_orchestrator`, `call_entity`,
   and `wait_for_external_event` gained new keyword-only parameters
   (`return_type` / `data_type`). Subclasses overriding these methods should add
   the parameter to match the base signature.
+- `EntityContext` and `EntityMetadata` (and its `from_entity_metadata` /
+  `from_entity_response` factories) now require a `data_converter` argument.
+  These objects are normally constructed by the SDK — you receive an
+  `EntityContext` in an entity function and an `EntityMetadata` from the client —
+  so this only affects code that constructs them directly.
 
 ## v1.6.0
 

durabletask-azuremanaged/durabletask/azuremanaged/client.py

@@ -20,6 +20,7 @@
 )
 import durabletask.internal.shared as shared
 from durabletask.payload.store import PayloadStore
+from durabletask.serialization import DataConverter
 
 
 # Client class used for Durable Task Scheduler (DTS)
@@ -35,6 +36,7 @@ def __init__(self, *,
                  resiliency_options: GrpcClientResiliencyOptions | None = None,
                  default_version: str | None = None,
                  payload_store: PayloadStore | None = None,
+                 data_converter: DataConverter | None = None,
                  log_handler: logging.Handler | None = None,
                  log_formatter: logging.Formatter | None = None):
 
@@ -59,7 +61,8 @@ def __init__(self, *,
             channel_options=channel_options,
             resiliency_options=resiliency_options,
             default_version=default_version,
-            payload_store=payload_store)
+            payload_store=payload_store,
+            data_converter=data_converter)
 
 
 # Async client class used for Durable Task Scheduler (DTS)
@@ -113,6 +116,7 @@ def __init__(self, *,
                  resiliency_options: GrpcClientResiliencyOptions | None = None,
                  default_version: str | None = None,
                  payload_store: PayloadStore | None = None,
+                 data_converter: DataConverter | None = None,
                  log_handler: logging.Handler | None = None,
                  log_formatter: logging.Formatter | None = None):
 
@@ -137,4 +141,5 @@ def __init__(self, *,
             channel_options=channel_options,
             resiliency_options=resiliency_options,
             default_version=default_version,
-            payload_store=payload_store)
+            payload_store=payload_store,
+            data_converter=data_converter)

durabletask-azuremanaged/durabletask/azuremanaged/worker.py

@@ -18,6 +18,7 @@
 )
 import durabletask.internal.shared as shared
 from durabletask.payload.store import PayloadStore
+from durabletask.serialization import DataConverter
 from durabletask.worker import ConcurrencyOptions, TaskHubGrpcWorker
 
 
@@ -81,6 +82,7 @@ def __init__(self, *,
                  resiliency_options: GrpcWorkerResiliencyOptions | None = None,
                  concurrency_options: ConcurrencyOptions | None = None,
                  payload_store: PayloadStore | None = None,
+                 data_converter: DataConverter | None = None,
                  log_handler: logging.Handler | None = None,
                  log_formatter: logging.Formatter | None = None):
 
@@ -110,5 +112,6 @@ def __init__(self, *,
             concurrency_options=concurrency_options,
             # DTS natively supports long timers so chunking is unnecessary
             maximum_timer_interval=None,
-            payload_store=payload_store
+            payload_store=payload_store,
+            data_converter=data_converter
         )

durabletask/entities/entity_context.py

@@ -16,14 +16,11 @@
 
 class EntityContext:
     def __init__(self, orchestration_id: str, operation: str, state: StateShim,
-                 entity_id: EntityInstanceId, data_converter: "DataConverter | None" = None):
+                 entity_id: EntityInstanceId, data_converter: "DataConverter"):
         self._orchestration_id = orchestration_id
         self._operation = operation
         self._state = state
         self._entity_id = entity_id
-        if data_converter is None:
-            from durabletask.serialization import JsonDataConverter
-            data_converter = JsonDataConverter()
         self._data_converter = data_converter
 
     @property

durabletask/entities/entity_metadata.py

@@ -36,7 +36,7 @@ def __init__(self,
                  locked_by: str,
                  includes_state: bool,
                  state: Any | None,
-                 data_converter: "DataConverter | None" = None):
+                 data_converter: "DataConverter"):
         """Initializes a new instance of the EntityMetadata class.
 
         Args:
@@ -48,20 +48,17 @@ def __init__(self,
         self._locked_by = locked_by
         self.includes_state = includes_state
         self._state = state
-        if data_converter is None:
-            from durabletask.serialization import JsonDataConverter
-            data_converter = JsonDataConverter()
         self._data_converter = data_converter
 
     @staticmethod
     def from_entity_response(entity_response: pb.GetEntityResponse, includes_state: bool,
-                             data_converter: "DataConverter | None" = None):
+                             data_converter: "DataConverter"):
         return EntityMetadata.from_entity_metadata(
             entity_response.entity, includes_state, data_converter)
 
     @staticmethod
     def from_entity_metadata(entity: pb.EntityMetadata, includes_state: bool,
-                             data_converter: "DataConverter | None" = None):
+                             data_converter: "DataConverter"):
         try:
             entity_id = EntityInstanceId.parse(entity.instanceId)
         except ValueError:

durabletask/internal/entity_state_shim.py

@@ -36,11 +36,8 @@ class StateShim:
     is written back with :meth:`set_state`.
     """
 
-    def __init__(self, start_state: Any, data_converter: "DataConverter | None" = None,
+    def __init__(self, start_state: Any, data_converter: "DataConverter",
                  *, is_serialized: bool = False):
-        if data_converter is None:
-            from durabletask.serialization import JsonDataConverter
-            data_converter = JsonDataConverter()
         self._data_converter = data_converter
         # The state is normalized to its serialized string form. ``is_serialized``
         # marks ``start_state`` as a raw payload already off the wire (stored

examples/custom_data_converter/src/app.py

@@ -75,6 +75,7 @@ def main() -> None:
         if state and state.runtime_status == client.OrchestrationStatus.COMPLETED:
             # ``get_output(Receipt)`` reconstructs the typed, validated result.
             receipt = state.get_output(Receipt)
+            assert receipt is not None
             print("Orchestration completed. Typed receipt:")
             print(f"  customer        = {receipt.customer}")
             print(f"  total           = {receipt.total}")

tests/durabletask/test_entity_executor.py

@@ -22,7 +22,7 @@ def _make_executor(*entity_args) -> _EntityExecutor:
 def _execute(executor, entity_name, operation, encoded_input=None):
     """Helper to execute an entity operation."""
     entity_id = entities.EntityInstanceId(entity_name, "test-key")
-    state = StateShim(None)
+    state = StateShim(None, JsonDataConverter())
     return executor.execute("test-orchestration", entity_id, operation, state, encoded_input)
 
 
@@ -75,7 +75,7 @@ def get(self):
         entity_id = entities.EntityInstanceId("Counter", "test-key")
 
         # set requires input
-        state = StateShim(None)
+        state = StateShim(None, JsonDataConverter())
         executor.execute("test-orch", entity_id, "set", state, "10")
         state.commit()
 
@@ -127,7 +127,7 @@ def counter(ctx: entities.EntityContext, input):
 
         executor = _make_executor(counter)
         entity_id = entities.EntityInstanceId("counter", "test-key")
-        state = StateShim(None)
+        state = StateShim(None, JsonDataConverter())
 
         executor.execute("test-orch", entity_id, "set", state, "42")
         state.commit()
@@ -140,19 +140,19 @@ class TestStateShimCoercion:
     """Tests for StateShim.get_state type coercion via the data converter."""
 
     def test_get_state_none_returns_default(self):
-        state = StateShim(None)
+        state = StateShim(None, JsonDataConverter())
         assert state.get_state(int, 0) == 0
 
     def test_get_state_none_without_default_returns_none(self):
-        state = StateShim(None)
+        state = StateShim(None, JsonDataConverter())
         assert state.get_state(int) is None
 
     def test_get_state_passes_through_matching_type(self):
-        state = StateShim(5)
+        state = StateShim(5, JsonDataConverter())
         assert state.get_state(int) == 5
 
     def test_get_state_constructor_coercion(self):
-        state = StateShim("5")
+        state = StateShim("5", JsonDataConverter())
         assert state.get_state(int) == 5
 
     def test_get_state_coerces_dataclass(self):
@@ -163,7 +163,7 @@ class Counter:
             value: int
 
         # State is stored as a plain dict (as it would be after from_json).
-        state = StateShim({"value": 7})
+        state = StateShim({"value": 7}, JsonDataConverter())
         result = state.get_state(Counter)
         assert isinstance(result, Counter)
         assert result.value == 7
@@ -177,7 +177,7 @@ def __init__(self, n: int):
             def from_json(cls, data):
                 return cls(data["n"])
 
-        state = StateShim({"n": 3})
+        state = StateShim({"n": 3}, JsonDataConverter())
         result = state.get_state(Wrapped)
         assert isinstance(result, Wrapped)
         assert result.n == 3
@@ -187,7 +187,7 @@ def test_get_state_invalid_coercion_raises(self):
         # restoring the pre-existing strict contract for entity state access.
         import pytest
 
-        state = StateShim("not-an-int")
+        state = StateShim("not-an-int", JsonDataConverter())
         with pytest.raises(TypeError):
             state.get_state(int)
 
@@ -197,7 +197,7 @@ class TestStateShimDeferredDeserialization:
 
     def test_constructor_does_not_deserialize_serialized_state(self):
         # A serialized payload is held verbatim until read, not eagerly parsed.
-        state = StateShim('{"value": 7}', is_serialized=True)
+        state = StateShim('{"value": 7}', JsonDataConverter(), is_serialized=True)
         assert state._current_state == '{"value": 7}'
 
     def test_get_state_defers_deserialization_with_type(self):
@@ -207,13 +207,13 @@ def test_get_state_defers_deserialization_with_type(self):
         class Counter:
             value: int
 
-        state = StateShim('{"value": 7}', is_serialized=True)
+        state = StateShim('{"value": 7}', JsonDataConverter(), is_serialized=True)
         result = state.get_state(Counter)
         assert isinstance(result, Counter)
         assert result.value == 7
 
     def test_get_state_no_type_returns_parsed_value(self):
-        state = StateShim('{"value": 7}', is_serialized=True)
+        state = StateShim('{"value": 7}', JsonDataConverter(), is_serialized=True)
         assert state.get_state() == {"value": 7}
 
     def test_deferred_deserialization_passes_raw_string_to_converter(self):
@@ -247,11 +247,11 @@ def coerce(self, value, target_type=None):
     def test_encode_state_passes_through_unmodified_payload(self):
         # An unread/unmodified serialized payload is returned verbatim, never
         # re-serialized (which would double-encode the JSON string).
-        state = StateShim('{"value": 7}', is_serialized=True)
+        state = StateShim('{"value": 7}', JsonDataConverter(), is_serialized=True)
         assert state.encode_state() == '{"value": 7}'
 
     def test_reading_does_not_trigger_double_encoding(self):
-        state = StateShim('{"value": 7}', is_serialized=True)
+        state = StateShim('{"value": 7}', JsonDataConverter(), is_serialized=True)
         # Reading (even with a type) must not turn the payload into a live value
         # that would be re-serialized into a JSON-encoded string.
         state.get_state()
@@ -261,31 +261,31 @@ def test_reading_does_not_trigger_double_encoding(self):
         assert json.loads(encoded) == {"value": 7}
 
     def test_encode_state_serializes_live_value_after_set_state(self):
-        state = StateShim('{"value": 7}', is_serialized=True)
+        state = StateShim('{"value": 7}', JsonDataConverter(), is_serialized=True)
         state.set_state({"value": 8})
         encoded = state.encode_state()
         assert json.loads(encoded) == {"value": 8}
 
     def test_encode_state_none_when_state_is_none(self):
-        state = StateShim(None, is_serialized=True)
+        state = StateShim(None, JsonDataConverter(), is_serialized=True)
         assert state.encode_state() is None
 
     def test_commit_preserves_unmodified_payload(self):
-        state = StateShim('{"value": 7}', is_serialized=True)
+        state = StateShim('{"value": 7}', JsonDataConverter(), is_serialized=True)
         state.commit()
         # After commit, the (unmodified) state still round-trips without
         # double-encoding.
         assert state.encode_state() == '{"value": 7}'
 
     def test_rollback_restores_unmodified_payload(self):
-        state = StateShim('{"value": 7}', is_serialized=True)
+        state = StateShim('{"value": 7}', JsonDataConverter(), is_serialized=True)
         state.commit()
         state.set_state({"value": 99})
         state.rollback()
         assert state.encode_state() == '{"value": 7}'
 
     def test_falsy_serialized_state_is_not_dropped(self):
         # A serialized falsy value (e.g. 0) is preserved, not treated as cleared.
-        state = StateShim("0", is_serialized=True)
+        state = StateShim("0", JsonDataConverter(), is_serialized=True)
         assert state.get_state(int) == 0
         assert state.encode_state() == "0"