def symmetry_filter(x, config, symmetry):
"""
Return True if the candidate sublattice configuration has any symmetry
which matches the phase model symmetry.
Parameters
----------
x : dict
the candidate dataset 'solver' dict. Must contain the "sublattice_configurations" key
config : list
the configuration of interest: e.g. ['AL', ['AL', 'NI'], 'VA']
symmetry : list
tuple of tuples where each inner tuple is a group of equivalent
sublattices. A value of ((0, 1), (2, 3, 4)) means that sublattices
at indices 0 and 1 are symmetrically equivalent to each other and
sublattices at indices 2, 3, and 4 are symetrically equivalent to
each other.
Returns
-------
bool
"""
if x['mode'] == 'manual':
if len(config) != len(x['sublattice_configurations'][0]):
return False
# If even one matches, it's a match
# We do more filtering downstream
for data_config in x['sublattice_configurations']:
if canonicalize(config, symmetry) == canonicalize(data_config, symmetry):
return True
return False
def get_data(comps, phase_name, configuration, symmetry, datasets, prop):
"""
Return list of cleaned single phase datasets matching the passed arguments.
Parameters
----------
comps : list
List of string component names
phase_name : str
Name of phase
configuration : tuple
Sublattice configuration as a tuple, e.g. ("CU", ("CU", "MG"))
symmetry : list of lists
List of sublattice indices with symmetry
datasets : espei.utils.PickleableTinyDB
Database of datasets to search for data
prop : list
String name of the property of interest.
Returns
-------
list
List of datasets matching the arguments.
"""
desired_data = datasets.search((tinydb.where('output').test(lambda x: x in prop)) &
(tinydb.where('components').test(lambda x: set(x).issubset(comps))) &
(tinydb.where('solver').test(symmetry_filter, configuration, recursive_tuplify(symmetry) if symmetry else symmetry)) &
(tinydb.where('phases') == [phase_name]))
# This seems to be necessary because the 'values' member does not modify 'datasets'
# But everything else does!
desired_data = copy.deepcopy(desired_data)
def recursive_zip(a, b):
if isinstance(a, (list, tuple)) and isinstance(b, (list, tuple)):
return list(recursive_zip(x, y) for x, y in zip(a, b))
else:
return list(zip(a, b))
for idx, data in enumerate(desired_data):
# Filter output values to only contain data for matching sublattice configurations
matching_configs = np.array([(canonicalize(sblconf, symmetry) == canonicalize(configuration, symmetry))
for sblconf in data['solver']['sublattice_configurations']])
matching_configs = np.arange(len(data['solver']['sublattice_configurations']))[matching_configs]
# Rewrite output values with filtered data
desired_data[idx]['values'] = np.array(data['values'], dtype=np.float)[..., matching_configs]
desired_data[idx]['solver']['sublattice_configurations'] = recursive_tuplify(np.array(data['solver']['sublattice_configurations'],
dtype=np.object)[matching_configs].tolist())
try:
desired_data[idx]['solver']['sublattice_occupancies'] = np.array(data['solver']['sublattice_occupancies'],
dtype=np.object)[matching_configs].tolist()
except KeyError:
pass
# Filter out temperatures below 298.15 K (for now, until better refstates exist)
temp_filter = np.atleast_1d(data['conditions']['T']) >= 298.15
desired_data[idx]['conditions']['T'] = np.atleast_1d(data['conditions']['T'])[temp_filter]
# Don't use data['values'] because we rewrote it above; not sure what 'data' references now
desired_data[idx]['values'] = desired_data[idx]['values'][..., temp_filter, :]
return desired_data
def filter_sublattice_configurations(
desired_data: List[Dataset],
subl_model) -> List[Dataset]: # TODO: symmetry support
"""Modify the desired_data to remove any configurations that cannot be represented by the sublattice model."""
subl_model_sets = [set(subl) for subl in subl_model]
for data in desired_data:
matching_configs = [
] # binary mask of whether a configuration is represented by the sublattice model
for config in data['solver']['sublattice_configurations']:
config = recursive_tuplify(canonicalize(config, None))
if (len(config) == len(subl_model) and all(
subl.issuperset(tuplify(config_subl))
for subl, config_subl in zip(subl_model_sets, config))):
matching_configs.append(True)
else:
matching_configs.append(False)
matching_configs = np.asarray(matching_configs, dtype=np.bool_)
# Rewrite output values with filtered data
data['values'] = np.array(data['values'],
dtype=np.float_)[..., matching_configs]
data['solver']['sublattice_configurations'] = np.array(
data['solver']['sublattice_configurations'],
dtype=np.object_)[matching_configs].tolist()
if 'sublattice_occupancies' in data['solver']:
data['solver']['sublattice_occupancies'] = np.array(
data['solver']['sublattice_occupancies'],
dtype=np.object_)[matching_configs].tolist()
return desired_data
def filter_configurations(desired_data: List[Dataset], configuration,
symmetry) -> List[Dataset]:
"""
Return non-equilibrium thermochemical datasets with invalid configurations removed.
Parameters
----------
desired_data : List[Dataset]
List of non-equilibrium thermochemical datasets
configuration : tuple
Sublattice configuration as a tuple, e.g. ("CU", ("CU", "MG"))
symmetry : list of lists
List of sublattice indices with symmetry
Returns
-------
List[Dataset]
"""
for data in desired_data:
# Filter output values to only contain data for matching sublattice configurations
matching_configs = np.array([
(canonicalize(sblconf,
symmetry) == canonicalize(configuration, symmetry))
for sblconf in data['solver']['sublattice_configurations']
])
matching_configs = np.arange(
len(data['solver']['sublattice_configurations']))[matching_configs]
# Rewrite output values with filtered data
data['values'] = np.array(data['values'],
dtype=np.float_)[..., matching_configs]
data['solver']['sublattice_configurations'] = recursive_tuplify(
np.array(data['solver']['sublattice_configurations'],
dtype=np.object_)[matching_configs].tolist())
if 'sublattice_occupancies' in data['solver']:
data['solver']['sublattice_occupancies'] = np.array(
data['solver']['sublattice_occupancies'],
dtype=np.object_)[matching_configs].tolist()
return desired_data