Make mypy run on more CI matrix, bring back "True-True"

.github/workflows/CI.yml

@@ -28,8 +28,6 @@ jobs:
         exclude:
           - rdkit: false
             openeye: false
-          - openeye: true
-            python-version: "3.11"
           - openeye: true
             python-version: "3.12"
           - openeye: true
@@ -138,12 +136,6 @@ jobs:
         run: pytest -r fE --tb=short openff/toolkit/_tests/test_links.py
 
       - name: Run mypy
-        # When possible re-enable this with OE+RDK=True, but for now rdkit builds often have bugs
-        # in stubs and there are other subtleties (in the source code and builds) that reduce the number
-        # of available builds.
-        # See ex https://github.com/rdkit/rdkit/issues/7221
-        # and https://github.com/rdkit/rdkit/issues/7583
-        if: ${{ matrix.rdkit == false && matrix.openeye == true }}
         run: mypy -p "openff.toolkit"
 
       - name: Run unit tests
@@ -156,23 +148,6 @@ jobs:
 
           python -m pytest -x --durations=20 $PYTEST_ARGS $COV openff/toolkit/_tests
 
-      - name: Run code snippets in docs
-        if: ${{ matrix.rdkit == true && matrix.openeye == true }}
-        run: pytest -v --no-cov --doctest-glob="docs/*.rst" --doctest-glob="docs/*.md" docs/
-
-      - name: Run notebooks in docs
-        if: ${{ matrix.rdkit == true && matrix.openeye == true }}
-        run: python -m pytest -v --no-cov --nbval-lax docs/
-
-      - name: Run examples in docstrings
-        if: ${{ matrix.rdkit == true && matrix.openeye == true }}
-        run: |
-          pytest openff \
-            -v -x -n logical --no-cov --doctest-modules \
-            --ignore-glob='openff/toolkit/_tests*' \
-            --ignore=openff/toolkit/data/ \
-            --ignore=openff/toolkit/utils/utils.py
-
       - name: Codecov
         uses: codecov/codecov-action@v5
         with:

.github/workflows/examples.yml

@@ -20,7 +20,6 @@ jobs:
     name: ${{ matrix.os }}, Python ${{ matrix.python-version }}, RDKit=${{ matrix.rdkit }}, OpenEye=${{ matrix.openeye }}
     runs-on: ${{ matrix.os }}
     strategy:
-      fail-fast: false
       matrix:
         os: [ubuntu-latest, macos-latest]
         python-version: ["3.11", "3.12"]
@@ -110,6 +109,23 @@ jobs:
       - name: Shim for `pytest-xdist` + Pint cross-interaction
         run: python -c "from openff.toolkit import *"
 
+      - name: Run code snippets in docs
+        if: ${{ matrix.openeye == true && matrix.rdkit == true }}
+        run: pytest -v --doctest-glob="docs/*.rst" --doctest-glob="docs/*.md" docs/
+
+      - name: Run notebooks in docs
+        if: ${{ matrix.openeye == true && matrix.rdkit == true }}
+        run: python -m pytest -v --nbval-lax docs/
+
+      - name: Run examples in docstrings
+        if: ${{ matrix.openeye == true && matrix.rdkit == true }}
+        run: |
+          pytest openff \
+            -v -x -n logical --doctest-modules \
+            --ignore-glob='openff/toolkit/_tests*' \
+            --ignore=openff/toolkit/data/ \
+            --ignore=openff/toolkit/utils/utils.py
+
       - name: Run example scripts
         run: |
           if [[ ${{ matrix.rdkit }} == false ]]; then

.pre-commit-config.yaml

@@ -3,7 +3,7 @@ ci:
 files: ^openff|(^examples/((?!deprecated).)*$)|^docs|(^utilities/((?!deprecated).)*$)
 repos:
 - repo: https://github.com/astral-sh/ruff-pre-commit
-  rev: v0.12.11
+  rev: v0.12.12
   hooks:
     - id: ruff-format
     - id: ruff-check

devtools/conda-envs/openeye.yaml

@@ -40,7 +40,7 @@ dependencies:
   - mdtraj
   - nglview
   - parmed =4
-  - mypy =1.15
+  - mypy =1.17
   - typing_extensions
   - pip:
     - types-setuptools

devtools/conda-envs/rdkit-examples.yaml

@@ -23,9 +23,8 @@ dependencies:
   - typing_extensions
   - nglview
     # Toolkit-specific
-  - ambertools
-    # https://github.com/rdkit/rdkit/issues/7221 and https://github.com/rdkit/rdkit/issues/7583
-  - rdkit =2024
+  - ambertools =24
+  - rdkit =2025
     # Test-only/optional/dev/typing/examples
   - pytest =8
   - pytest-xdist

devtools/conda-envs/rdkit.yaml

@@ -22,10 +22,10 @@ dependencies:
   - openff-nagl-models >=0.3.0
   - typing_extensions
     # Toolkit-specific
-  - ambertools >=22
-    # https://github.com/rdkit/rdkit/issues/7221 and https://github.com/rdkit/rdkit/issues/7583
-  - rdkit !=2024.03.6,!=2024.03.5
+  - ambertools =24
+  - rdkit =2025
     # Test-only/optional/dev/typing
+  - mypy =1.17
   - pytest =8
   - pytest-cov
   - pytest-xdist
@@ -39,3 +39,11 @@ dependencies:
   - qcportal >=0.50
   - qcengine
   - nglview
+  - pip:
+    - types-setuptools
+    - types-toml
+    - types-PyYAML
+    - types-networkx
+    - types-xmltodict
+    - types-cachetools
+    - mongo-types

devtools/conda-envs/test_env.yaml

@@ -23,13 +23,12 @@ dependencies:
   - openff-nagl-models >=0.3.0
   - typing_extensions
     # Toolkit-specific
-  - ambertools >=22
-    # rdkit 2024.03.6 and 2024.09.1 packages fail when run natively on osx-arm64 macs
-    # https://github.com/rdkit/rdkit/issues/7583
-  - rdkit !=2024.03.6,!=2024.09.1
+  - ambertools =24
+  - rdkit =2025
 
   - openeye-toolkits
     # Test-only/optional/dev/typing
+  - mypy =1.17
   - pytest =8
   - pytest-cov
   - pytest-xdist
@@ -46,5 +45,11 @@ dependencies:
   - nglview
   - mdtraj
   - nbval
-  # No idea why this is necessary, see https://github.com/openforcefield/openff-toolkit/pull/1821
-  - nomkl
+  - pip:
+    - types-setuptools
+    - types-toml
+    - types-PyYAML
+    - types-networkx
+    - types-xmltodict
+    - types-cachetools
+    - mongo-types

openff/toolkit/topology/_mm_molecule.py

@@ -505,7 +505,7 @@ def node_match_func(node1, node2):
         else:
             return False, None
 
-    def generate_unique_atom_names(self):
+    def generate_unique_atom_names(self) -> None:
         """Generate unique atom names. See `Molecule.generate_unique_atom_names`."""
         from collections import defaultdict
 

openff/toolkit/topology/molecule.py

@@ -1250,8 +1250,8 @@ def ordered_connection_table_hash(self) -> int:
         for bond in self.bonds:
             id += f"{bond.bond_order}_{bond.stereochemistry}_{bond.atom1_index}_{bond.atom2_index}__"
 
-        self._ordered_connection_table_hash = hashlib.sha3_224(id.encode("utf-8")).hexdigest()
-        return self._ordered_connection_table_hash
+        self._ordered_connection_table_hash = hashlib.sha3_224(id.encode("utf-8")).hexdigest()  # type: ignore[assignment]
+        return self._ordered_connection_table_hash  # type: ignore[return-value]
 
     @classmethod
     def from_dict(cls: type[FM], molecule_dict: dict) -> FM:
@@ -1350,7 +1350,7 @@ def __repr__(self) -> str:
             hill = self.to_hill_formula()
             return description + f" with bad SMILES and Hill formula '{hill}'"
 
-    def _initialize(self):
+    def _initialize(self) -> None:
         """
         Clear the contents of the current molecule.
         """
@@ -1359,9 +1359,9 @@ def _initialize(self):
         self._bonds: list[Bond] = list()  # list of bonds between Atom objects
         self._properties = {}  # Attached properties to be preserved
         # self._cached_properties = None # Cached properties (such as partial charges) can be recomputed as needed
-        self._partial_charges = None
-        self._conformers = None  # Optional conformers
-        self._hill_formula = None  # Cached Hill formula
+        self._partial_charges: Quantity | None = None
+        self._conformers: list[Quantity] | None = None  # Optional conformers
+        self._hill_formula: str | None = None  # Cached Hill formula
         self._hierarchy_schemes = dict()
         self._ordered_connection_table_hash = None
         self._invalidate_cached_properties()
@@ -2771,7 +2771,7 @@ def assign_fractional_bond_orders(
                 f"Expected ToolkitRegistry or ToolkitWrapper. Got {type(toolkit_registry)}."
             )
 
-    def _invalidate_cached_properties(self):
+    def _invalidate_cached_properties(self) -> None:
         """
         Indicate that the chemical entity has been altered.
 
@@ -4095,7 +4095,12 @@ def title(frame):
         # now close the file
         xyz_data.close()
 
-    def to_file(self, file_path, file_format, toolkit_registry=GLOBAL_TOOLKIT_REGISTRY):
+    def to_file(
+        self,
+        file_path,
+        file_format: str,
+        toolkit_registry: TKR = GLOBAL_TOOLKIT_REGISTRY,
+    ):
         """Write the current molecule to a file or file-like object
 
         Parameters
@@ -4123,14 +4128,11 @@ def to_file(self, file_path, file_format, toolkit_registry=GLOBAL_TOOLKIT_REGIST
         >>> molecule.to_file('imatinib.pdb', file_format='pdb')  # doctest: +SKIP
 
         """
-        toolkit: ToolkitRegistry | None
-
         if isinstance(toolkit_registry, ToolkitRegistry):
             pass
         elif isinstance(toolkit_registry, ToolkitWrapper):
-            toolkit = toolkit_registry
-            toolkit_registry = ToolkitRegistry(toolkit_precedence=[])
-            toolkit_registry.add_toolkit(toolkit)
+            # make the toolkit_registry: ToolkitWrapper actually a ToolkitRegistry
+            toolkit_registry = ToolkitRegistry(toolkit_precedence=[toolkit_registry])  # type: ignore[list-item]
         else:
             raise InvalidToolkitRegistryError(
                 "'toolkit_registry' must be either a ToolkitRegistry or a ToolkitWrapper"
@@ -4142,10 +4144,10 @@ def to_file(self, file_path, file_format, toolkit_registry=GLOBAL_TOOLKIT_REGIST
             return self._to_xyz_file(file_path=file_path)
 
         # Take the first toolkit that can write the desired output format
-        toolkit = None
+        toolkit: ToolkitRegistry | None = None
         for query_toolkit in toolkit_registry.registered_toolkits:
             if file_format in query_toolkit.toolkit_file_write_formats:
-                toolkit = query_toolkit
+                toolkit = query_toolkit  # type: ignore[assignment]
                 break
 
         # Raise an exception if no toolkit was found to provide the requested file_format
@@ -4159,9 +4161,9 @@ def to_file(self, file_path, file_format, toolkit_registry=GLOBAL_TOOLKIT_REGIST
             )
 
         if isinstance(file_path, str | pathlib.Path):
-            toolkit.to_file(self, file_path, file_format)
+            toolkit.to_file(self, file_path, file_format)  # type: ignore[attr-defined]
         else:
-            toolkit.to_file_obj(self, file_path, file_format)
+            toolkit.to_file_obj(self, file_path, file_format)  # type: ignore[attr-defined]
 
     def enumerate_tautomers(self, max_states=20, toolkit_registry=GLOBAL_TOOLKIT_REGISTRY):
         """
@@ -5815,7 +5817,7 @@ def to_dict(self) -> dict:
         return_dict["hierarchy_elements"] = [e.to_dict() for e in self.hierarchy_elements]
         return return_dict
 
-    def perceive_hierarchy(self):
+    def perceive_hierarchy(self) -> None:
         """
         Prepare the parent ``Molecule`` for iteration according to this scheme.
 

openff/toolkit/topology/topology.py

@@ -1299,7 +1299,7 @@ def _initialize_from_dict(self, topology_dict):
 
     @staticmethod
     @requires_package("openmm")
-    def _openmm_topology_to_networkx(openmm_topology):
+    def _openmm_topology_to_networkx(openmm_topology: "openmm.app.Topology"):
         import networkx as nx
 
         # Convert all openMM mols to graphs
@@ -1763,8 +1763,8 @@ def from_pdb(
             off_atom.metadata["chain_id"] = atom.residue.chain.id
             off_atom.name = atom.name
 
-        for offmol in topology.molecules:
-            offmol.add_default_hierarchy_schemes()
+        for offmol in topology.molecules:  # type: ignore[operator]
+            offmol.add_default_hierarchy_schemes()  # type: ignore[operator]
 
         return topology
 
@@ -2104,9 +2104,11 @@ def set_positions(self, array: Quantity) -> None:
             raise IncompatibleUnitError("array should be an OpenFF Quantity with dimensions of length")
 
         # Copy the array in nanometers and make it an OpenFF Quantity
-        array = Quantity(np.asarray(array.to(unit.nanometer).magnitude), unit.nanometer)
-        if array.shape != (self.n_atoms, 3):
-            raise WrongShapeError(f"Array has shape {array.shape} but should have shape {self.n_atoms, 3}")
+        array = Quantity(np.asarray(array.to(unit.nanometer).magnitude), unit.nanometer)  # type: ignore[attr-defined]
+        if array.shape != (self.n_atoms, 3):  # type: ignore[attr-defined]
+            raise WrongShapeError(
+                f"Array has shape {array.shape} but should have shape {self.n_atoms, 3}"  # type: ignore[attr-defined]
+            )
 
         start = 0
         for molecule in self.molecules:

openff/toolkit/typing/engines/smirnoff/forcefield.py

@@ -335,7 +335,7 @@ def __init__(
         # Parse all sources containing SMIRNOFF parameter definitions
         self.parse_sources(sources, allow_cosmetic_attributes=allow_cosmetic_attributes)
 
-    def _initialize(self):
+    def _initialize(self) -> None:
         """
         Initialize all object fields.
         """

openff/toolkit/utils/utils.py

@@ -166,7 +166,7 @@ def quantity_to_string(input_quantity: Quantity) -> str:
     if isinstance(unitless_value, np.ndarray):
         unitless_value = list(unitless_value)
 
-    unit_string = unit_to_string(input_quantity.units)
+    unit_string = unit_to_string(input_quantity.units)  # type: ignore
 
     return f"{unitless_value} * {unit_string}"
 

pyproject.toml

@@ -61,7 +61,6 @@ exclude_lines = [
 
 [tool.mypy]
 python_version = 3.12
-plugins = "numpy.typing.mypy_plugin"
 warn_unused_configs = true
 # would be nice to flip back to true
 warn_unused_ignores = false
@@ -79,7 +78,7 @@ module = [
     "openmm",
     "openmm.app",
     "openmm.unit",
-    "rdkit",
+    "rdkit",  # maybe revisit if/when stubs are fixed
     "rdkit.Chem",
     "rdkit.Chem.Draw",
     "rdkit.DataStructs.cDataStructs",
@@ -93,3 +92,11 @@ module = [
     "constraint",
 ]
 ignore_missing_imports = true
+
+[[tool.mypy.overrides]]
+# bit of a hack to prevent crash
+module=[
+    "rdkit"
+]
+follow_imports = "skip"
+follow_imports_for_stubs = true