Fix CI: reorder install steps and clean up ruff/mypy errors

`uv pip install dgl` was running before `uv sync` in both lint and
test workflows. setup-uv v8 no longer auto-creates a venv, so the
pip install failed with "No virtual environment found." Swap the
order: sync first to create the venv, then pip install dgl to
override the version.

Also clean up lint errors that would surface once install succeeds:
- Add PLC0415 to ruff ignore (lazy imports for optional deps like
  tensorflow/megnet/m3gnet are intentional; 31 violations)
- Drop obsolete PD901 ignore (rule was removed from ruff)
- Prefix 11 unused unpacked variables with _ (RUF059)
- Replace `key=lambda x: Element(x)` with `key=Element` (PLW0108, 5)
- Default reduction_params to {} before ** unpack (_composition.py)
- Extend type: ignore codes in gbe/describer.py and data/_mp.py
- Add `equil` (LAMMPS equilibration filename) to codespell allowlist

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Author: shyuep

.github/workflows/lint.yml

@@ -20,8 +20,8 @@ jobs:
         cache-dependency-glob: "uv.lock"
     - name: Install dependencies
       run: |
-        uv pip install dgl==2.4.0 -f https://data.dgl.ai/wheels/torch-2.4/repo.html
         uv sync --group lint
+        uv pip install dgl==2.4.0 -f https://data.dgl.ai/wheels/torch-2.4/repo.html
 
     - name: ruff
       run: |

.github/workflows/test.yml

@@ -38,8 +38,8 @@ jobs:
         cache-dependency-glob: "uv.lock"
     - name: Install dependencies
       run: |
-        uv pip install dgl==2.4.0 -f https://data.dgl.ai/wheels/torch-2.4/repo.html
         uv sync
+        uv pip install dgl==2.4.0 -f https://data.dgl.ai/wheels/torch-2.4/repo.html
     - name: pytest
       run: |
         uv run pytest tests --cov=maml
@@ -66,8 +66,8 @@ jobs:
           cache-dependency-glob: "uv.lock"
       - name: Install dependencies
         run: |
-          uv pip install dgl==2.4.0 -f https://data.dgl.ai/wheels/torch-2.4/repo.html
           uv sync
+          uv pip install dgl==2.4.0 -f https://data.dgl.ai/wheels/torch-2.4/repo.html
       - run: |
           uv build
       - name: Publish to PyPi

pyproject.toml

@@ -135,9 +135,9 @@ lint.ignore = [
     "D105", # Missing docstring in magic method
     "D205", # 1 blank line required between summary line and description
     "D212", # Multi-line docstring summary should start at the first line
-    "PD901", # pandas-df-variable-name
     "PERF203", # try-except-in-loop
     "PERF401", # manual-list-comprehension (TODO fix these or wait for autofix)
+    "PLC0415", # `import` should be at the top-level of a file (lazy imports for optional deps are intentional)
     "PLR", # pylint refactor
     "PLW2901", # Outer for loop variable overwritten by inner assignment target
     "PT013", # pytest-incorrect-pytest-import
@@ -187,7 +187,8 @@ ignore_missing_imports = true
 [tool.codespell]
 ignore-words-list = """
 titel,alls,ans,nd,mater,nwo,te,hart,ontop,ist,ot,fo,nax,coo,coul,ser,leary,thre,
-fase,rute,reson,titels,ges,scalr,strat,struc,hda,nin,ons,pres,kno,loos,lamda,lew
+fase,rute,reson,titels,ges,scalr,strat,struc,hda,nin,ons,pres,kno,loos,lamda,lew,
+equil
 """
 skip = "pymatgen/analysis/aflow_prototypes.json"
 check-filenames = true

src/maml/apps/bowsr/model/dft.py

@@ -58,7 +58,7 @@ def predict_energy(self, structure: Structure):
                 stdout=subprocess.PIPE,
                 stderr=subprocess.PIPE,  # type: ignore
             ) as p_exe:
-                stdout, stderr = p_exe.communicate()
+                _stdout, stderr = p_exe.communicate()
                 rc = p_exe.returncode
             if rc != 0:
                 error_msg = f"vasp exited with return code {rc}"

src/maml/apps/gbe/describer.py

@@ -274,7 +274,7 @@ def get_elemental_feature(
     el = Element(db_entry["pretty_formula"])
     f_dict[preset.ar.str_name] = el.atomic_radius
     f_dict[preset.a0.str_name] = bulk_s.lattice.a
-    f_dict[preset.e_coh.str_name] = rester.get_cohesive_energy(db_entry["material_id"])  # type: ignore[attr-defined]
+    f_dict[preset.e_coh.str_name] = rester.get_cohesive_energy(db_entry["material_id"])  # type: ignore[attr-defined,misc]
     f_dict[preset.hb.str_name] = el.brinell_hardness
     f_dict[preset.CLTE.str_name] = el.coefficient_of_linear_thermal_expansion
 

src/maml/apps/pes/_gap.py

@@ -281,7 +281,7 @@ def train(
                 if str(specie) not in gap_sorted_elements:
                     gap_sorted_elements.append(str(specie))
 
-        self.elements = sorted(gap_sorted_elements, key=lambda x: Element(x))
+        self.elements = sorted(gap_sorted_elements, key=Element)
 
         atoms_filename = "train.xyz"
         xml_filename = "train.xml"
@@ -504,5 +504,5 @@ def get_xml(xml_file):
         tree, element_param, potential_label = get_xml(filename)
         parameters = dict(xml=tree, element_param=element_param, potential_label=potential_label)
         gap = GAPotential(param=parameters)
-        gap.elements = sorted(element_param.keys(), key=lambda x: Element(x))
+        gap.elements = sorted(element_param.keys(), key=Element)
         return gap

src/maml/apps/pes/_lammps.py

@@ -211,7 +211,7 @@ def calculate(self, structures):
 
     @staticmethod
     def _rotate_force_stress(structure, forces, stresses):
-        _, symmop, rot_matrix = get_lammps_lattice_and_rotation(structure)
+        _, _symmop, rot_matrix = get_lammps_lattice_and_rotation(structure)
         inv_rot_matrix = np.linalg.inv(rot_matrix)
         forces = forces.dot(inv_rot_matrix.T)
         stresses = stress_list_to_matrix(stresses, stress_format="LAMMPS")
@@ -323,7 +323,7 @@ def add_args(line):
             return line + compute_args if line.startswith("compute") else line
 
         compute_args = f"1 0.99363 {self.twojmax} "
-        el_in_seq = sorted(self.element_profile.keys(), key=lambda x: Element(x))
+        el_in_seq = sorted(self.element_profile.keys(), key=Element)
         cutoffs = [self.element_profile[e]["r"] * self.rcutfac for e in el_in_seq]
         weights = [self.element_profile[e]["w"] for e in el_in_seq]
         compute_args += " ".join([str(p) for p in cutoffs + weights])

src/maml/apps/pes/_nnp.py

@@ -521,7 +521,7 @@ def str_formatify(string):
         df = pd.DataFrame([line.split() for line in lines if "#" not in line])
         self.elements = sorted(
             (element for element in np.ravel(df[df[0] == "elements"])[1:] if element is not None),
-            key=lambda x: Element(x),
+            key=Element,
         )
 
         atom_energy = {}
@@ -731,7 +731,7 @@ def train(self, train_structures, train_energies, train_forces, train_stresses=N
 
             self.write_input(**kwargs)
             with subprocess.Popen(["nnp-scaling", "100"], stdout=subprocess.PIPE, stderr=subprocess.PIPE) as p_scaling:
-                stdout, stderr = p_scaling.communicate()
+                _stdout, stderr = p_scaling.communicate()
                 rc = p_scaling.returncode
             if rc != 0:
                 error_msg = f"n2p2 exited with return code {rc}"
@@ -745,7 +745,7 @@ def train(self, train_structures, train_energies, train_forces, train_stresses=N
                 raise RuntimeError(error_msg)
 
             with open(output, "w") as f, subprocess.Popen(["nnp-train"], stdout=f, stderr=subprocess.PIPE) as p_train:
-                stdout, stderr = p_train.communicate()
+                _stdout, stderr = p_train.communicate()
                 rc = p_train.returncode
 
             if rc != 0:
@@ -821,7 +821,7 @@ def evaluate(self, test_structures, test_energies, test_forces, test_stresses=No
                 with subprocess.Popen(
                     ["nnp-predict", input_filename], stdout=subprocess.PIPE, stderr=subprocess.PIPE
                 ) as p_evaluation:
-                    stdout, stderr = p_evaluation.communicate()
+                    _stdout, stderr = p_evaluation.communicate()
                     rc = p_evaluation.returncode
                 if rc != 0:
                     error_msg = f"n2p2 exited with return code {rc}"

src/maml/apps/symbolic/_selectors.py

@@ -47,7 +47,7 @@ def select(self, x: np.ndarray, y: np.ndarray, options: dict | None = None) -> n
                 to scipy.optimize.minimize
         Returns: list of int indices.
         """
-        n_data, n_dim = x.shape
+        _n_data, n_dim = x.shape
         options = options or {"maxiter": 1e4, "ftol": 1e-12}
         res = minimize(
             lambda beta: self.construct_loss(x=x, y=y, beta=beta),
@@ -535,7 +535,7 @@ def select(self, x: np.ndarray, y: np.ndarray, options: dict | None = None, n_jo
             n_job (int): number of cpu
         Returns:
         """
-        n, p = x.shape
+        _n, p = x.shape
         index_array = list(range(p))
 
         def _lstsq(c):

src/maml/apps/symbolic/_selectors_cvxpy.py

@@ -44,7 +44,7 @@ def select(self, x: np.ndarray, y: np.ndarray, options: dict | None = None) -> n
         Returns: list of int indices.
         """
         options = options or {}
-        n, p = x.shape
+        _n, p = x.shape
         beta = cp.Variable(p)
         objective = cp.Minimize(self.construct_loss(x, y, beta))
         constraints = self.construct_constraints(x, y, beta)