Performance improvements

.prospector.yaml

@@ -2,6 +2,8 @@ max-line-length: 120
 test-warnings: false
 doc-warnings: false
 strictness: medium
+with: []
+uses: []
 ignore-paths:
   - docs
   - tests
@@ -21,6 +23,7 @@ pycodestyle:
 
 pylint:
   disable:
+    - django-not-available
     - unsubscriptable-object
     - invalid-name
     - arguments-differ # to account for jobflow

pyproject.toml

@@ -113,6 +113,8 @@ enabled = true
 
 [dependency-groups]
 dev = [
+    "black>=24.8.0",
+    "prospector>=1.10.3",
     "pytest>=7.1.3",
     "pytest-cov>=4.0.0",
-]
+]

src/pylattica/core/basic_controller.py

@@ -15,8 +15,18 @@ class has a single responsibility, which is to implement the update
     SimulationState will be passed to this method, along with the ID of
     the site at which the update rule should be applied. It is up to the
     user to decide what updates should be produced using this information.
+
+    Attributes
+    ----------
+    max_history : int, optional
+        Maximum number of step diffs to keep in memory during simulation.
+        Set this to limit memory usage for long simulations. When exceeded,
+        older steps are checkpointed and dropped. Default is None (unlimited).
     """
 
+    # Override this in subclasses to limit memory usage
+    max_history: int = None
+
     @abstractmethod
     def get_state_update(self, site_id: int, prev_state: SimulationState):
         pass  # pragma: no cover
@@ -25,7 +35,10 @@ def pre_run(self, initial_state: SimulationState) -> None:
         pass
 
     def get_random_site(self, state: SimulationState):
-        return random.randint(0, len(state.site_ids()) - 1)
+        # Use state.size (O(1)) instead of len(state.site_ids()) which is O(n)
+        return random.randint(0, state.size - 1)
 
     def instantiate_result(self, starting_state: SimulationState):
-        return SimulationResult(starting_state=starting_state)
+        return SimulationResult(
+            starting_state=starting_state, max_history=self.max_history
+        )

src/pylattica/core/neighborhood_builders.py

@@ -139,16 +139,14 @@ def get(self, struct: PeriodicStructure, site_class: str = None) -> Neighborhood
         else:
             sites_to_process = struct.sites(site_class=site_class)
 
-        n_sites = len(all_sites)
-
         # Extract locations and IDs as arrays for vectorized operations
         locations = np.array([s[LOCATION] for s in all_sites])
         site_ids = np.array([s[SITE_ID] for s in all_sites])
 
         # Convert to fractional coordinates for periodic KD-tree
-        frac_coords = np.array([
-            struct.lattice.get_fractional_coords(loc) for loc in locations
-        ])
+        frac_coords = np.array(
+            [struct.lattice.get_fractional_coords(loc) for loc in locations]
+        )
 
         # Compute the maximum fractional radius that could correspond to
         # the Cartesian cutoff. For non-orthogonal lattices, we need to use
@@ -163,9 +161,7 @@ def get(self, struct: PeriodicStructure, site_class: str = None) -> Neighborhood
         dim = struct.lattice.dim
 
         # Build boxsize array: 1.0 for periodic dimensions, large value for non-periodic
-        boxsize = np.array([
-            1.0 if periodic[i] else 1e10 for i in range(dim)
-        ])
+        boxsize = np.array([1.0 if periodic[i] else 1e10 for i in range(dim)])
 
         # Wrap fractional coordinates to [0, 1) for periodic dimensions
         frac_coords_wrapped = frac_coords.copy()
@@ -176,9 +172,6 @@ def get(self, struct: PeriodicStructure, site_class: str = None) -> Neighborhood
         # Build KD-tree with periodic boundary conditions
         tree = cKDTree(frac_coords_wrapped, boxsize=boxsize)
 
-        # Create index mapping from site_id to array index
-        id_to_idx = {sid: idx for idx, sid in enumerate(site_ids)}
-
         # Process each site
         sites_to_process_ids = set(s[SITE_ID] for s in sites_to_process)
 
@@ -296,9 +289,9 @@ def get(self, struct: PeriodicStructure, site_class: str = None) -> Neighborhood
         site_ids = np.array([s[SITE_ID] for s in all_sites])
 
         # Convert to fractional coordinates for periodic KD-tree
-        frac_coords = np.array([
-            struct.lattice.get_fractional_coords(loc) for loc in locations
-        ])
+        frac_coords = np.array(
+            [struct.lattice.get_fractional_coords(loc) for loc in locations]
+        )
 
         # Compute the maximum fractional radius for the outer cutoff.
         # Use the maximum stretch factor of the inverse matrix for non-orthogonal lattices.
@@ -311,9 +304,7 @@ def get(self, struct: PeriodicStructure, site_class: str = None) -> Neighborhood
         dim = struct.lattice.dim
 
         # Build boxsize array
-        boxsize = np.array([
-            1.0 if periodic[i] else 1e10 for i in range(dim)
-        ])
+        boxsize = np.array([1.0 if periodic[i] else 1e10 for i in range(dim)])
 
         # Wrap fractional coordinates to [0, 1) for periodic dimensions
         frac_coords_wrapped = frac_coords.copy()

src/pylattica/core/runner/asynchronous_runner.py

@@ -27,7 +27,7 @@ class AsynchronousRunner(Runner):
     that this mode should be used with the is_async initialization parameter.
     """
 
-    def _run(
+    def _run(  # pylint: disable=too-many-positional-arguments
         self,
         _: SimulationState,
         result: SimulationResult,

src/pylattica/core/runner/synchronous_runner.py

@@ -38,7 +38,7 @@ def __init__(self, parallel: bool = False, workers: int = None) -> None:
         self.parallel = parallel
         self.workers = workers
 
-    def _run(
+    def _run(  # pylint: disable=too-many-positional-arguments
         self,
         initial_state: SimulationState,
         result: SimulationResult,