def test_add_element(self):
# start with empty pipeline without auto-parametrization
p = api.pipeline([], run=False)
# add some reader
reader = api.source(self.traj_files, top=self.pdb_file)
p.add_element(reader)
p.parametrize()
# get the result immediately
out1 = reader.get_output()
# add some kmeans
kmeans = api.cluster_kmeans(k=15)
p.add_element(kmeans)
p.parametrize()
# get the result immediately
kmeans1 = kmeans.get_output()
# get reader output again
out2 = reader.get_output()
p.add_element(api.kmeans(k=2))
p.parametrize()
# get kmeans output again
kmeans2 = kmeans.get_output()
# check if add_element changes the intermediate results
np.testing.assert_array_equal(out1[0], out2[0])
np.testing.assert_array_equal(out1[1], out2[1])
np.testing.assert_array_equal(kmeans1[0], kmeans2[0])
np.testing.assert_array_equal(kmeans1[1], kmeans2[1])
def test_read_multiple_files_topology_file(self):
reader = api.source(self.traj_files, top=self.pdb_file)
self.assertIsNotNone(reader, "The reader should not be none.")
self.assertEqual(reader.topfile, self.pdb_file,
"Reader topology file and input topology file should coincide.")
self.assertListEqual(reader.trajfiles, self.traj_files, "Reader trajectories and input"
" trajectories should coincide.")
self.assertEqual(reader.featurizer.topologyfile, self.pdb_file, "Featurizers topology file and input "
"topology file should coincide.")
def test_no_cluster(self):
reader_xtc = api.source(self.traj_files, top=self.pdb_file)
# only reader
api.pipeline(reader_xtc)
reader_xtc.get_output()
# reader + pca / tica
tica = api.tica()
pca = api.pca()
api.pipeline([reader_xtc, tica])._chain[-1].get_output()
api.pipeline([reader_xtc, pca])._chain[-1].get_output()
def test_read_single_file_featurizer(self):
featurizer = MDFeaturizer(self.pdb_file)
reader = api.source(self.traj_files[0], features=featurizer)
self.assertIsNotNone(reader, "The reader should not be none.")
self.assertEqual(reader.topfile, self.pdb_file,
"Reader topology file and input topology file should coincide.")
self.assertListEqual(reader.trajfiles, [self.traj_files[0]], "Reader trajectories and input"
" trajectories should coincide.")
self.assertEqual(reader.featurizer.topologyfile, self.pdb_file, "Featurizers topology file and input "
"topology file should coincide.")
def testIteratorAccess(self):
reader = api.source(self.trajfile, top=self.topfile)
frames = 0
data = []
for i, X in reader:
frames += X.shape[0]
data.append(X)
# restore shape of input
data = np.array(data).reshape(self.xyz.shape)
self.assertEqual(frames, reader.trajectory_lengths()[0])
self.assertTrue(np.allclose(data, self.xyz))
def test_set_element(self):
reader = api.source(self.traj_files, top=self.pdb_file)
pca = api.pca()
p = api.pipeline([reader, pca])
self.assertTrue(p._is_parametrized())
pca_out = pca.get_output()
tica = api.tica(lag=self.generated_lag)
# replace pca with tica
p.set_element(1, tica)
self.assertFalse(p._is_parametrized(), "After replacing an element, the pipeline should not be parametrized.")
p.parametrize()
tica_out = tica.get_output()
# check if replacement actually happened
self.assertFalse(np.array_equal(pca_out[0], tica_out[0]),
"The output should not be the same when the method got replaced.")
def test_replace_data_source(self):
reader_xtc = api.source(self.traj_files, top=self.pdb_file)
reader_gen = DataInMemory(data=self.generated_data)
kmeans = api.kmeans(k=10)
assert hasattr(kmeans, '_chunks')
p = api.pipeline([reader_xtc, kmeans])
out1 = kmeans.get_output()
# replace source
print reader_gen
p.set_element(0, reader_gen)
assert hasattr(kmeans, '_chunks')
p.parametrize()
out2 = kmeans.get_output()
self.assertFalse(np.array_equal(out1, out2), "Data source changed, so should the resulting clusters.")
def test_is_parametrized(self):
# construct pipeline with all possible transformers
p = api.pipeline(
[
api.source(self.traj_files, top=self.pdb_file),
api.tica(),
api.pca(),
api.cluster_kmeans(k=50),
api.cluster_regspace(dmin=50),
api.cluster_uniform_time(k=20)
], run=False
)
self.assertFalse(p._is_parametrized(), "If run=false, the pipeline should not be parametrized.")
p.parametrize()
self.assertTrue(p._is_parametrized(), "If parametrized was called, the pipeline should be parametrized.")
def testIteratorAccess(self):
reader = api.source(self.trajfile, top=self.topfile)
assert isinstance(reader, FeatureReader)
frames = 0
data = []
for i, X in reader:
assert isinstance(X, np.ndarray)
frames += X.shape[0]
data.append(X)
self.assertEqual(frames, reader.trajectory_lengths()[0])
data = np.vstack(data)
# restore shape of input
data.reshape(self.xyz.shape)
self.assertTrue(np.allclose(data, self.xyz.reshape(-1, 9)))
def test_lagged_access_small_files(self):
""" itraj 0 should be skipped, since it is too short."""
top = self.topfile
trajs = [
create_traj(top=top, length=10, format='.xtc', dir=self.tmpdir)[0],
create_traj(top=top, length=20, format='.xtc', dir=self.tmpdir)[0]
]
reader = source(trajs, top=top)
it = reader.iterator(lag=11, chunk=0)
res = {}
with it:
for itraj, x, y in it:
res[itraj] = (x.shape, y.shape)
self.assertNotIn(0, res)
self.assertIn(1, res)
def test_is_parametrized(self):
# construct pipeline with all possible transformers
p = api.pipeline([
api.source(self.traj_files, top=self.pdb_file),
api.tica(),
api.pca(),
api.cluster_kmeans(k=50),
api.cluster_regspace(dmin=50),
api.cluster_uniform_time(k=20)
],
run=False)
self.assertFalse(
p._is_estimated(),
"If run=false, the pipeline should not be parametrized.")
p.parametrize()
self.assertTrue(
p._is_estimated(),
"If parametrized was called, the pipeline should be parametrized.")
def test_old_db_conversion(self):
# prior 2.1, database only contained lengths (int as string) entries
# check conversion is happening
with NamedTemporaryFile(suffix='.npy', delete=False) as f:
db = TrajectoryInfoCache(None)
fn = f.name
np.save(fn, [1, 2, 3])
f.close() # windows sucks
reader = api.source(fn)
hash = db._get_file_hash(fn)
from pyemma.coordinates.data.util.traj_info_backends import DictDB
db._database = DictDB()
db._database.db_version = 0
info = db[fn, reader]
assert info.length == 3
assert info.ndim == 1
assert info.offsets == []
def test_lagged_stridden_access(self):
reader = api.source([self.trajfile, self.trajfile2], top=self.topfile)
reader.chunksize = 210
strides = [2, 3, 5, 7, 15]
lags = [1, 3, 7, 10, 30]
err_msg = "not equal for stride=%i, lag=%i"
for stride in strides:
for lag in lags:
chunks = {itraj: [] for itraj in range(reader.number_of_trajectories())}
for itraj, _, Y in reader.iterator(stride=stride, lag=lag):
chunks[itraj].append(Y)
chunks[0] = np.vstack(chunks[0])
np.testing.assert_almost_equal(
chunks[0], self.xyz.reshape(-1, 9)[lag::stride], err_msg=err_msg % (stride, lag))
chunks[1] = np.vstack(chunks[1])
np.testing.assert_almost_equal(
chunks[1], self.xyz2.reshape(-1, 9)[lag::stride], err_msg=err_msg % (stride, lag))
def test_fragmented_reader_random_access1(self):
with TemporaryDirectory() as td:
trajfiles = []
for i in range(3):
trajfiles.append(
create_traj(start=i * 10, dir=td, length=20)[0])
topfile = get_top()
trajfiles = [(trajfiles[0], trajfiles[1]), trajfiles[0],
trajfiles[2]]
source = coor.source(trajfiles, top=topfile)
assert isinstance(source, FragmentedTrajectoryReader)
for r in source._readers:
if not isinstance(r, (list, tuple)):
r = r[0]
for _r in r:
_r._return_traj_obj = True
from collections import defaultdict
for chunksize in [0, 2, 3, 100000]:
frames = defaultdict(list)
with source.iterator(chunk=chunksize,
return_trajindex=True,
stride=self.stride) as it:
for itraj, t in it:
frames[itraj].append(t)
dest = []
for itraj in frames.keys():
dest.append(frames[itraj][0])
for t in frames[itraj][1:]:
dest[-1] = dest[-1].join(t)
keys = np.unique(self.stride[:, 0])
for i, coords in enumerate(dest):
if i in keys:
traj = mdtraj.load(trajfiles[i], top=topfile)
np.testing.assert_equal(
coords.xyz,
traj.xyz[np.array(
self.stride[self.stride[:, 0] == i][:, 1])],
err_msg="not equal for chunksize=%s" % chunksize)
def test_set_element(self):
reader = api.source(self.traj_files, top=self.pdb_file)
pca = api.pca()
p = api.pipeline([reader, pca])
self.assertTrue(p._is_estimated())
pca_out = pca.get_output()
tica = api.tica(lag=self.generated_lag)
# replace pca with tica
p.set_element(1, tica)
self.assertFalse(
p._is_estimated(),
"After replacing an element, the pipeline should not be parametrized."
)
p.parametrize()
tica_out = tica.get_output()
# check if replacement actually happened
self.assertFalse(
np.array_equal(pca_out[0], tica_out[0]),
"The output should not be the same when the method got replaced.")
def test_cols_with_features(self):
trajs = glob.glob(
pkg_resources.resource_filename('pyemma.coordinates.tests',
'data/bpti_mini.xtc'))
top = pkg_resources.resource_filename('pyemma.coordinates.tests',
'data/bpti_ca.pdb')
reader = api.source(trajs, top=top)
feat = reader.featurizer
inds = feat.pairs(feat.select('name CA'))
reader.featurizer.add_distances(inds)
# select first and second atom?
cols = np.array((0, 2))
ref = mdtraj.load(trajs, top=top)
ref = mdtraj.compute_distances(ref, inds)
ref = ref[:, cols]
it = reader.iterator(chunk=0, return_trajindex=False, cols=cols)
with it:
for x in it:
np.testing.assert_equal(x, ref)
def setUpClass(cls):
with numpy_random_seed(123):
import msmtools.generation as msmgen
# generate HMM with two Gaussians
cls.P = np.array([[0.99, 0.01],
[0.01, 0.99]])
cls.T = 40000
means = [np.array([-1, 1]), np.array([1, -1])]
widths = [np.array([0.3, 2]), np.array([0.3, 2])]
# continuous trajectory
cls.X = np.zeros((cls.T, 2))
# hidden trajectory
dtraj = msmgen.generate_traj(cls.P, cls.T)
for t in range(cls.T):
s = dtraj[t]
cls.X[t, 0] = widths[s][0] * np.random.randn() + means[s][0]
cls.X[t, 1] = widths[s][1] * np.random.randn() + means[s][1]
# Set the lag time:
cls.lag = 10
# Compute mean free data:
mref = (np.sum(cls.X[:-cls.lag, :], axis=0) +
np.sum(cls.X[cls.lag:, :], axis=0)) / float(2*(cls.T-cls.lag))
mref_nr = np.sum(cls.X[:-cls.lag, :], axis=0) / float(cls.T-cls.lag)
cls.X_mf = cls.X - mref[None, :]
cls.X_mf_nr = cls.X - mref_nr[None, :]
# Compute correlation matrices:
cls.cov_ref = (np.dot(cls.X_mf[:-cls.lag, :].T, cls.X_mf[:-cls.lag, :]) +\
np.dot(cls.X_mf[cls.lag:, :].T, cls.X_mf[cls.lag:, :])) / float(2*(cls.T-cls.lag))
cls.cov_ref_nr = np.dot(cls.X_mf_nr[:-cls.lag, :].T, cls.X_mf_nr[:-cls.lag, :]) / float(cls.T - cls.lag)
cls.cov_tau_ref = (np.dot(cls.X_mf[:-cls.lag, :].T, cls.X_mf[cls.lag:, :]) +\
np.dot(cls.X_mf[cls.lag:, :].T, cls.X_mf[:-cls.lag, :])) / float(2*(cls.T-cls.lag))
cls.cov_tau_ref_nr = np.dot(cls.X_mf_nr[:-cls.lag, :].T, cls.X_mf_nr[cls.lag:, :]) / float(cls.T - cls.lag)
# do unscaled TICA
reader=api.source(cls.X, chunk_size=0)
cls.tica_obj = api.tica(data=reader, lag=cls.lag, dim=1, kinetic_map=False)
# non-reversible TICA
cls.tica_obj_nr = api.tica(data=reader, lag=cls.lag, dim=1, kinetic_map=False, reversible=False)
def setUpClass(cls):
with numpy_random_seed(123):
import msmtools.generation as msmgen
# generate HMM with two Gaussians
cls.P = np.array([[0.99, 0.01],
[0.01, 0.99]])
cls.T = 40000
means = [np.array([-1, 1]), np.array([1, -1])]
widths = [np.array([0.3, 2]), np.array([0.3, 2])]
# continuous trajectory
cls.X = np.zeros((cls.T, 2))
# hidden trajectory
dtraj = msmgen.generate_traj(cls.P, cls.T)
for t in range(cls.T):
s = dtraj[t]
cls.X[t, 0] = widths[s][0] * np.random.randn() + means[s][0]
cls.X[t, 1] = widths[s][1] * np.random.randn() + means[s][1]
cls.lag = 10
# do unscaled TICA
reader=api.source(cls.X, chunk_size=0)
cls.tica_obj = api.tica(data=reader, lag=cls.lag, dim=1, kinetic_map=False)
def test_RA_high_stride(self):
""" ensure we use a random access pattern for high strides chunksize combinations to avoid memory issues."""
from pyemma.coordinates.util.patches import iterload
n = int(1e5)
n_bytes = 3 * 3 * 8 * n # ~8Mb
savable_formats_mdtra_18 = ('.xtc', '.trr', '.dcd', '.h5', '.binpos',
'.nc', '.netcdf', '.ncdf', '.tng')
for ext in savable_formats_mdtra_18:
traj = create_traj(length=n, dir=self.tmpdir, format=ext)[0]
from unittest.mock import patch
# temporarily overwrite the memory cutoff with a smaller value, to trigger the switch to RA stride.
with patch(
'pyemma.coordinates.util.patches.iterload.MEMORY_CUTOFF',
n_bytes - 1):
r = coor.source(traj, top=get_top())
it = r.iterator(stride=1000, chunk=100000)
next(it)
assert iterload._DEACTIVATE_RANDOM_ACCESS_OPTIMIZATION or it._mditer.is_ra_iter
out_ra = r.get_output(stride=1000, chunk=10000)
it = r.iterator(stride=1)
next(it)
assert iterload._DEACTIVATE_RANDOM_ACCESS_OPTIMIZATION or not it._mditer.is_ra_iter
out = r.get_output(stride=1000)
np.testing.assert_equal(out_ra, out)
# check max stride exceeding
it = r.iterator(stride=iterload.MAX_STRIDE_SWITCH_TO_RA + 1)
next(it)
assert iterload._DEACTIVATE_RANDOM_ACCESS_OPTIMIZATION or it._mditer.is_ra_iter
it = r.iterator(stride=iterload.MAX_STRIDE_SWITCH_TO_RA)
next(it)
assert iterload._DEACTIVATE_RANDOM_ACCESS_OPTIMIZATION or not it._mditer.is_ra_iter
def testTimeLaggedAccess(self):
# each frame has 2 atoms with 3 coords = 6 coords per frame.
# coords are sequential through all frames and start with 0.
lags = [2, 200]
chunksizes = [1, 100]
for lag in lags:
for chunksize in chunksizes:
log.info("chunksize=%i\tlag=%i" % (chunksize, lag))
lagged_chunks = []
reader = api.source(self.trajfile, top=self.topfile)
reader.chunksize = chunksize
for _, _, y in reader.iterator(lag=lag):
lagged_chunks.append(y)
coords = self.xyz.reshape((self.xyz.shape[0], -1))
for ii, c in enumerate(lagged_chunks[:-1]):
# all despite last chunk shall have chunksize
self.assertTrue(c.shape[0] <= chunksize)
# first lagged chunk should start at lag and stop at chunksize +
# lag
ind1 = ii * chunksize + lag
ind2 = ind1 + chunksize
#log.debug("coor slice[%i: %i]" % (ind1, ind2))
np.testing.assert_allclose(c, coords[ind1:ind2])
# TODO: check last lagged frame
# last lagged chunk should miss "lag" frames of input! e.g
# padded to maintain chunksize
last_chunk = lagged_chunks[-1]
def testTimeLaggedAccess(self):
# each frame has 2 atoms with 3 coords = 6 coords per frame.
# coords are sequential through all frames and start with 0.
lags = [2, 200]
chunksizes = [1, 100]
for lag in lags:
for chunksize in chunksizes:
log.info("chunksize=%i\tlag=%i" % (chunksize, lag))
lagged_chunks = []
reader = api.source(self.trajfile, top=self.topfile)
reader.chunksize = chunksize
for _, _, y in reader.iterator(lag=lag):
lagged_chunks.append(y)
coords = self.xyz.reshape((self.xyz.shape[0], -1))
for ii, c in enumerate(lagged_chunks[:-1]):
# all despite last chunk shall have chunksize
self.assertTrue(c.shape[0] <= chunksize)
# first lagged chunk should start at lag and stop at chunksize +
# lag
ind1 = ii * chunksize + lag
ind2 = ind1 + chunksize
#log.debug("coor slice[%i: %i]" % (ind1, ind2))
np.testing.assert_allclose(c, coords[ind1:ind2])
# TODO: check last lagged frame
# last lagged chunk should miss "lag" frames of input! e.g
# padded to maintain chunksize
last_chunk = lagged_chunks[-1]
def test_exceptions(self):
# in accessible files
not_existant = ''.join(
chr(i) for i in np.random.randint(65, 90, size=10)) + '.npy'
bad = [not_existant] # should be unaccessible or non existent
with self.assertRaises(ValueError) as cm:
api.source(bad)
assert bad[0] in cm.exception.message
# empty files
with NamedTemporaryFile(delete=False) as f:
f.close()
with self.assertRaises(ValueError) as cm:
api.source(f.name)
assert f.name in cm.exception.message
# bogus files
with NamedTemporaryFile(suffix='.npy', delete=False) as f:
x = np.array([1, 2, 3])
np.save(f, x)
with open(f.name, 'wb') as f2:
f2.write(b'asdf')
with self.assertRaises(IOError) as cm:
api.source(f.name)
def test_obtain_csv_file_reader_csv(self):
reader = api.source(self.csv)
self.assertIsNotNone(reader, "Reader object should not be none.")
self.assertTrue(isinstance(reader, CSVReader), "Should be a CSVReader.")
def setUpClass(cls):
path = pkg_resources.resource_filename(__name__, 'data') + os.path.sep
cls.pdb_file = os.path.join(path, 'bpti_ca.pdb')
cls.xtc_file = os.path.join(path, 'bpti_mini.xtc')
cls.inp = api.source(cls.xtc_file, top=cls.pdb_file)
def test_format_loading_via_feature_reader(self):
reader = source(traj_file, top=top, dir=self.tmpdir)
reader.get_output()
def setUpClass(cls):
path = os.path.join(os.path.split(__file__)[0], 'data')
cls.pdb_file = os.path.join(path, 'bpti_ca.pdb')
cls.xtc_file = os.path.join(path, 'bpti_mini.xtc')
cls.inp = api.source(cls.xtc_file, top=cls.pdb_file)
def test_pdb_traj_unsupported(self):
with self.assertRaises(ValueError) as c, tempfile.NamedTemporaryFile(
suffix='.pdb') as ntf:
api.source([ntf.name], top=self.bpti_pdbfile)
assert 'PDB' in c.exception.args[0]
def test_no_transform(self):
reader_xtc = api.source(self.traj_files, top=self.pdb_file)
api.pipeline([reader_xtc, api.cluster_kmeans(k=10)])._chain[-1].get_output()
api.pipeline([reader_xtc, api.cluster_regspace(dmin=10)])._chain[-1].get_output()
api.pipeline([reader_xtc, api.cluster_uniform_time()])._chain[-1].get_output()
def test_obtain_csv_file_reader_csv(self):
reader = api.source(self.csv)
self.assertIsNotNone(reader, "Reader object should not be none.")
self.assertTrue(isinstance(reader, CSVReader), "Should be a CSVReader.")
def test_obtain_numpy_file_reader_npz(self):
reader = api.source(self.npz)
self.assertIsNotNone(reader, "Reader object should not be none.")
self.assertTrue(
isinstance(reader, NumPyFileReader), "Should be a NumPyFileReader.")
def test_obtain_numpy_file_reader_npz(self):
reader = api.source(self.npz)
self.assertIsNotNone(reader, "Reader object should not be none.")
self.assertTrue(
isinstance(reader, NumPyFileReader), "Should be a NumPyFileReader.")
def test_data_in_mem(self):
# make sure cache is not used for data in memory!
data = [np.empty((3, 3))] * 3
api.source(data)
self.assertEqual(self.db.num_entries, 0)
def test_data_in_memory_without_first_two_trajs(self):
data_in_memory = coor.source(self.data, chunksize=10)
out = data_in_memory.get_output(stride=self.stride2)
np.testing.assert_array_almost_equal(out[2], [self.data[2][0]])
def test_data_in_mem(self):
# make sure cache is not used for data in memory!
data = [np.empty((3, 3))] * 3
api.source(data)
assert len(self.db._database) == 1
def __init__(self,
trajectories,
topologyfile=None,
chunksize=1000,
featurizer=None):
self._args = (trajectories, topologyfile, chunksize, featurizer)
# sanity checks
assert isinstance(
trajectories,
(list,
tuple)), "input trajectories should be of list or tuple type"
# if it contains no further list: treat as single trajectory
if not any([isinstance(traj, (list, tuple)) for traj in trajectories]):
trajectories = [trajectories]
# if not list of lists, treat as single-element-fragment-trajectory
trajectories = [
traj if isinstance(traj, (list, tuple)) else [traj]
for traj in trajectories
]
# some trajectory should be provided
assert len(trajectories) > 0, "no input trajectories provided"
# call super
super(FragmentedTrajectoryReader, self).__init__(chunksize=chunksize)
self._is_reader = True
# number of trajectories
self._ntraj = len(trajectories)
# store readers
from pyemma.coordinates.api import source
self._readers = [[
source(input_item,
features=featurizer,
top=topologyfile,
chunksize=chunksize) for input_item in trajectories[itraj]
] for itraj in range(0, self._ntraj)]
# check all readers have same dimension
if not len(set(itraj_r.ndim for r in self._readers
for itraj_r in r)) == 1:
# lookup the evil reader:
last_dim = -1
for r in self._readers:
for itraj_r in r:
if last_dim == -1:
last_dim = itraj_r.ndim
if itraj_r.ndim != last_dim:
raise ValueError(
"%s has different dimension (%i) than expected (%i)"
% (itraj_r.describe(), itraj_r.ndim, last_dim))
from collections import defaultdict
self._reader_by_filename = defaultdict(list)
for r in self._readers:
for itraj_r in r:
for filename in itraj_r.filenames:
self._reader_by_filename[filename].append(itraj_r)
# lengths array per reader
self._reader_lengths = [[
reader.trajectory_length(0, 1) for reader in self._readers[itraj]
] for itraj in range(0, self._ntraj)]
# composite trajectory length
self._lengths = [
sum(self._reader_lengths[itraj])
for itraj in range(0, self._ntraj)
]
# mapping reader_index -> cumulative length
self._cumulative_lengths = [
np.cumsum(self._reader_lengths[itraj])
for itraj in range(0, self._ntraj)
]
# store trajectory files
self._trajectories = trajectories
self._filenames = trajectories
