Skip to content
This repository has been archived by the owner on Nov 28, 2023. It is now read-only.

Commit

Permalink
Revert "Started performance upgrade."
Browse files Browse the repository at this point in the history 
This reverts commit b679995.
  • Loading branch information
@lauri-codes
lauri-codes committed Jan 24, 2023
1 parent a77f3ae commit 015f698
Showing 1 changed file with 12 additions and 114 deletions.
126 changes: 12 additions & 114 deletions matid/classification/structureclusterer.py
View file
Original file line number Diff line number Diff line change
@@ -1,82 +1,20 @@
from collections import defaultdict

import numpy as np
from ase.data import covalent_radii

import matid.geometry
from matid.data import constants
from matid.classifications import Class0D, Class1D, Class2D, Class3D, Surface, Material2D, Atom, Unknown
import ase.io
from ase.io import write
from ase.visualize import view
from matid.classification.periodicfinder import PeriodicFinder


class Cluster():
"""
Represents a part of a bigger system.
"""
def __init__(self, system, indices, regions, dist_matrix_radii_mic, species=None):
self.system = system
def __init__(self, indices, regions, species=None):
self.indices = indices
self.regions = regions
self.species = species
self.merged = False
self.dist_matrix_radii_pbc = dist_matrix_radii_mic[indices, indices]

def classify(self, cluster_threshold=constants.CLUSTER_THRESHOLD):
"""Used to classify this cluster.
"""
# Check that the region was connected cyclically in two
# directions. This ensures that finite systems or systems
# with a dislocation at the cell boundary are filtered.
best_region = self.regions[0]
region_conn = best_region.get_connected_directions()
n_region_conn = np.sum(region_conn)
region_is_periodic = n_region_conn == 2

# This might be unnecessary because the connectivity of the
# unit cell is already checked.
clusters = best_region.get_clusters()
basis_indices = set(list(best_region.get_basis_indices()))
split = True
for cluster in clusters:
if basis_indices.issubset(cluster):
split = False

# Get the system dimensionality
classification = Unknown()
dimensionality = matid.geometry.get_dimensionality(
self.system,
cluster_threshold,
self.dist_matrix_radii_pbc
)

# 0D structures
if dimensionality == 0:
classification = Class0D()

# Systems with one atom have their own classification.
n_atoms = len(self.indices)
if n_atoms == 1:
classification = Atom()

# 1D structures
elif dimensionality == 1:
classification = Class1D()

# 2D structures
elif dimensionality == 2:
if not split and region_is_periodic:
if best_region.is_2d:
classification = Material2D()
else:
classification = Surface()
else:
classification = Class2D()

# Bulk structures
elif dimensionality == 3:
classification = Class3D()

return classification


class StructureClusterer():
Expand Down Expand Up @@ -161,14 +99,19 @@ def merge(system, a, b):
remainder_species = set(atomic_numbers[list(remainder_indices)])
final_indices = target.indices.union(common)

return Cluster(system, final_indices, a.regions + b.regions, self.dist_matrix_radii_mic, target.species), Cluster(system, remainder_indices, source.regions, self.dist_matrix_radii_mic, remainder_species)
return Cluster(final_indices, a.regions + b.regions, target.species), Cluster(remainder_indices, source.regions, remainder_species)

covered_grains = set()
new_grain_map = {}
atomic_numbers = system.get_atomic_numbers()
remainders = []
isolated_clusters = []
while (True):
if len(clusters) == 0 or clusters[0].merged:
break
i_cluster = clusters.pop(0)
i_indices = i_cluster.indices
i_region = i_cluster.regions[0]

# Check overlap with all other non-isolated clusters
isolated = True
Expand Down Expand Up @@ -218,12 +161,9 @@ def get_clusters(self, system, angle_tol=20, max_cell_size=5, pos_tol=0.25, merg
Returns:
A list of Clusters.
"""
# Calculate the displacements here once.
disp_tensor_mic, disp_factors, disp_tensor_finite, dist_matrix_radii_mic = self.get_displacements(system)
self.dist_matrix_radii_mic = dist_matrix_radii_mic

# Iteratively search for new clusters until whole system is covered
periodic_finder = PeriodicFinder(angle_tol=angle_tol)
covered = False
indices = set(list(range(len(system))))
clusters = []
atomic_numbers = system.get_atomic_numbers()
Expand All @@ -234,10 +174,6 @@ def get_clusters(self, system, angle_tol=20, max_cell_size=5, pos_tol=0.25, merg
seed_index=i_seed,
max_cell_size=max_cell_size,
pos_tol=pos_tol,
disp_tensor_mic=disp_tensor_mic,
disp_factors=disp_factors,
disp_tensor_finite=disp_tensor_finite,
dist_matrix_radii_mic=dist_matrix_radii_mic,
)
i_indices = [i_seed]
if i_grain is not None:
Expand All @@ -247,7 +183,7 @@ def get_clusters(self, system, angle_tol=20, max_cell_size=5, pos_tol=0.25, merg
i_species = set(atomic_numbers[i_indices])
i_indices = set(i_indices)

clusters.append(Cluster(system, i_indices, [i_grain], self.dist_matrix_radii_mic, i_species))
clusters.append(Cluster(i_indices, [i_grain], i_species))

# Remove found grain from set of indices
if len(i_indices) == 0:
Expand All @@ -265,41 +201,3 @@ def get_clusters(self, system, angle_tol=20, max_cell_size=5, pos_tol=0.25, merg

return clusters

def get_displacements(self, system):
"""Return the necessary displacement information.
Args:
system (ase.Atoms): The system to investigate.
Returns:
A tuple containing all necessary displacement infomration.
"""
pos = system.get_positions()
cell = system.get_cell()
pbc = system.get_pbc()
disp_tensor_finite = matid.geometry.get_displacement_tensor(pos, pos)
if pbc.any():
disp_tensor_mic, disp_factors = matid.geometry.get_displacement_tensor(
pos,
pos,
cell,
pbc,
mic=True,
return_factors=True
)
else:
disp_tensor_mic = disp_tensor_finite
disp_factors = np.zeros(disp_tensor_finite.shape)
dist_matrix_mic = np.linalg.norm(disp_tensor_mic, axis=2)

# Calculate the distance matrix where the periodicity and the covalent
# radii have been taken into account
dist_matrix_radii_mic = np.array(dist_matrix_mic)
num = system.get_atomic_numbers()
radii = covalent_radii[num]
radii_matrix = radii[:, None] + radii[None, :]
dist_matrix_radii_mic -= radii_matrix

return (disp_tensor_mic, disp_factors, disp_tensor_finite, dist_matrix_radii_mic)

0 comments on commit 015f698

Please sign in to comment.