def test_duplicate_different_topology(self):
conformer = get_stage1_conformer()
# bond topology is conformer_id // 1000
conformer.duplicated_by = conformer.conformer_id + 1000
self.assertEqual(
dataset_pb2.Conformer.FATE_DUPLICATE_DIFFERENT_TOPOLOGY,
smu_utils_lib.determine_fate(conformer))
def test_duplicate_same_topology(self):
conformer = get_stage1_conformer()
# bond topology is conformer_id // 1000
conformer.duplicated_by = conformer.conformer_id + 1
smu_utils_lib.clean_up_error_codes(conformer)
self.assertEqual(dataset_pb2.Conformer.FATE_DUPLICATE_SAME_TOPOLOGY,
smu_utils_lib.determine_fate(conformer))
def process(self, conformer):
conformer = copy.deepcopy(conformer)
conformer.fate = smu_utils_lib.determine_fate(conformer)
yield from self._compare_smiles(conformer)
yield conformer
def process(self, conformer, bond_length_records, smiles_id_dict):
"""Per conformer updates.
Args:
conformer: dataset_pb2.Conformer
bond_length_records: tuples to go to
bond_length_distribution.AllAtomPairLengthDistributions
smiles_id_dict: dict from SMILES to bond topology id
Yields:
Conformer.
"""
# There is probably a better way to do this.
# We get the side input with each call to process. We'll assume that it's
# always the same input, so we set our cache value and never update it.
# We only do this with bond_length_records because there is a reasonable
# amount of processing in creating AllAtomPairLengthDistributions.
# The smiles_id_dict is used directly.
if not self._cached_bond_lengths:
self._cached_bond_lengths = (
bond_length_distribution.AllAtomPairLengthDistributions())
try:
self._cached_bond_lengths.add_from_sparse_dataframe(
bond_length_distribution.sparse_dataframe_from_records(
bond_length_records), _BOND_LENGTHS_UNBONDED_RIGHT_TAIL_MASS,
_BOND_LENGTHS_SIG_DIGITS)
except ValueError as err:
raise ValueError(
'Invalid sparse dataframe for conformer {0} org. ValueError: {1}'
.format(str(conformer.conformer_id), err))
conformer = copy.deepcopy(conformer)
conformer.fate = smu_utils_lib.determine_fate(conformer)
yield from self._compare_smiles(conformer)
if (conformer.duplicated_by == 0 and
conformer.properties.errors.status < 512):
# The duplicate records do not need topology extraction and anything
# with this high an error is pretty messed so, do we won't bother trying
# to match the topolgy.
self._add_alternative_bond_topologies(conformer, smiles_id_dict)
else:
beam.metrics.Metrics.counter(_METRICS_NAMESPACE,
'skipped_topology_matches').inc()
yield conformer
def process(self, molecule, bond_length_records, smiles_id_dict):
"""Per molecule updates.
Args:
molecule: dataset_pb2.Molecule
bond_length_records: tuples to go to
bond_length_distribution.AllAtomPairLengthDistributions
smiles_id_dict: dict from SMILES to bond topology id
Yields:
Molecule.
"""
# There is probably a better way to do this.
# We get the side input with each call to process. We'll assume that it's
# always the same input, so we set our cache value and never update it.
# We only do this with bond_length_records because there is a reasonable
# amount of processing in creating AllAtomPairLengthDistributions.
# The smiles_id_dict is used directly.
if not self._cached_bond_lengths:
self._cached_bond_lengths = (
bond_length_distribution.AllAtomPairLengthDistributions())
try:
self._cached_bond_lengths.add_from_sparse_dataframe(
bond_length_distribution.sparse_dataframe_from_records(
bond_length_records),
bond_length_distribution.STANDARD_UNBONDED_RIGHT_TAIL_MASS,
bond_length_distribution.STANDARD_SIG_DIGITS)
except ValueError as err:
raise ValueError(
'Invalid sparse dataframe for molecule {0} org. ValueError: {1}'
.format(str(molecule.molecule_id), err)) from err
molecule = copy.deepcopy(molecule)
molecule.properties.errors.fate = smu_utils_lib.determine_fate(molecule)
yield from self._compare_smiles(molecule)
if smu_utils_lib.molecule_eligible_for_topology_detection(molecule):
self._add_alternative_bond_topologies(molecule, smiles_id_dict)
else:
molecule.bond_topologies[
0].source = dataset_pb2.BondTopology.SOURCE_STARTING
beam.metrics.Metrics.counter(_METRICS_NAMESPACE,
'skipped_topology_matches').inc()
yield molecule
def test_success(self):
conformer = get_stage2_conformer()
self.assertEqual(dataset_pb2.Conformer.FATE_SUCCESS,
smu_utils_lib.determine_fate(conformer))
def test_calculation_errors(self):
conformer = get_stage2_conformer()
# This is a random choice of an error to set. I just need some error.
conformer.properties.errors.error_atomic_analysis = 999
self.assertEqual(dataset_pb2.Conformer.FATE_CALCULATION_WITH_ERROR,
smu_utils_lib.determine_fate(conformer))
def test_no_result(self):
conformer = get_stage1_conformer()
self.assertEqual(dataset_pb2.Conformer.FATE_NO_CALCULATION_RESULTS,
smu_utils_lib.determine_fate(conformer))
def test_geometry_failures(self, nstat1, expected_fate): conformer = get_stage1_conformer() conformer.properties.errors.error_nstat1 = nstat1 self.assertEqual(expected_fate, smu_utils_lib.determine_fate(conformer))
def test_calculation_warnings_vibrational(self):
conformer = get_stage2_conformer()
conformer.properties.errors.warn_vib_linearity = 1234
self.assertEqual(
dataset_pb2.Conformer.FATE_CALCULATION_WITH_WARNING_VIBRATIONAL,
smu_utils_lib.determine_fate(conformer))
def test_calculation_warnings_serious(self):
conformer = get_stage2_conformer()
conformer.properties.errors.warn_t1_excess = 1234
self.assertEqual(
dataset_pb2.Conformer.FATE_CALCULATION_WITH_WARNING_SERIOUS,
smu_utils_lib.determine_fate(conformer))
def test_calculation_errors(self, status, expected): conformer = get_stage2_conformer() conformer.properties.errors.status = status self.assertEqual(expected, smu_utils_lib.determine_fate(conformer))
def test_discarded_other(self, status):
conformer = get_stage1_conformer()
conformer.properties.errors.status = status
smu_utils_lib.clean_up_error_codes(conformer)
self.assertEqual(dataset_pb2.Conformer.FATE_DISCARDED_OTHER,
smu_utils_lib.determine_fate(conformer))
