feat(tidy3d): Mutable attrs bypass _json_string cache, causing stale hashes and exports

- hash also sensitive to attrs changes

Jira: FXC-3374

Author: marcorudolphflex

CHANGELOG.md

@@ -10,7 +10,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Changed
 - Improved performance of antenna metrics calculation by utilizing cached wave amplitude calculations instead of recomputing wave amplitudes for each port excitation in the `TerminalComponentModelerData`.
-- Fixed bug for `Tidy3dBaseModel`: Hashes are independent from `attrs`. Still `.json_string` does include them and is cached, but cache is updated if `attrs` are updated.
+- Fixed: In `Tidy3dBaseModel` the hash (and cached `.json_string`) are now sensitive to changes in `.attrs`.
 
 ### Fixed
 - More robust `Sellmeier` and `Debye` material model, and prevent very large pole parameters in `PoleResidue` material model.

tests/test_components/test_base.py

@@ -262,17 +262,16 @@ def test_scientific_notation(min_val, max_val, min_digits, expected):
     assert result == expected
 
 
-def test_stable_hash_with_changed_attr():
+def test_updated_hash_and_json_with_changed_attr():
     obj = td.Medium(attrs={"foo": "attr"})
+
     old_hash = obj._hash_self()
-    obj.attrs["foo"] = "changed"
-    new_hash = obj._hash_self()
-    assert new_hash == old_hash
+    json_old = obj._json_string
 
+    obj.attrs["foo"] = "changed"
 
-def test_updated_json_for_mutable_attr():
-    obj = td.Medium(attrs={"foo": "attr"})
-    json_old = obj._json_string
-    obj.attrs["foo"] = np.array([0])
+    new_hash = obj._hash_self()
     json_new = obj._json_string
+
+    assert new_hash != old_hash
     assert json_old != json_new

tidy3d/components/base.py

@@ -173,7 +173,7 @@ def __hash__(self) -> int:
     def _hash_self(self) -> str:
         """Hash this component with ``hashlib`` in a way that is the same every session."""
         bf = io.BytesIO()
-        self.to_hdf5(bf, for_hashing=True)
+        self.to_hdf5(bf)
         return hashlib.sha256(bf.getvalue()).hexdigest()
 
     def __init__(self, **kwargs):
@@ -776,7 +776,6 @@ def to_hdf5(
         self,
         fname: str,
         custom_encoders: Optional[list[Callable]] = None,
-        for_hashing: bool = False,
     ) -> None:
         """Exports :class:`Tidy3dBaseModel` instance to .hdf5 file.
 
@@ -787,16 +786,14 @@ def to_hdf5(
         custom_encoders : List[Callable]
             List of functions accepting (fname: str, group_path: str, value: Any) that take
             the ``value`` supplied and write it to the hdf5 ``fname`` at ``group_path``.
-        for_hashing : bool = False
-            Whether to use hdf5 for hashing.
 
         Example
         -------
         >>> simulation.to_hdf5(fname='folder/sim.hdf5') # doctest: +SKIP
         """
 
         with h5py.File(fname, "w") as f_handle:
-            json_str = self._json_string if not for_hashing else self._json_string_for_hashing
+            json_str = self._json_string
             for ind in range(ceil(len(json_str) / MAX_STRING_LENGTH)):
                 ind_start = int(ind * MAX_STRING_LENGTH)
                 ind_stop = min(int(ind + 1) * MAX_STRING_LENGTH, len(json_str))
@@ -991,18 +988,6 @@ def check_equal(dict1: dict, dict2: dict) -> bool:
 
         return check_equal(self.dict(), other.dict())
 
-    @cached_property
-    def _json_string_for_hashing(self) -> str:
-        """Returns string representation of a :class:`Tidy3dBaseModel`.
-
-        Returns
-        -------
-        str
-            Json-formatted string holding :class:`Tidy3dBaseModel` data.
-        """
-        obj_no_attrs = self._updated_copy(attrs={}, deep=True, validate=False)
-        return obj_no_attrs._json()
-
     @cached_property_guarded(lambda self: self._attrs_digest())
     def _json_string(self) -> str:
         """Returns string representation of a :class:`Tidy3dBaseModel`.