def topology(self, top):
"""Set the topology in the file
Parameters
----------
top : mdtraj.Topology
A topology object
"""
_check_mode(self.mode, ('w', ))
if self._needs_initialization:
self._initialize_headers(top.n_atoms)
self._needs_initialization = False
top, bonds = top.to_dataframe()
data = {
"AtomID": top.index.values + 1,
"AtomNames": top.name.values,
"ResidueNames": top.resName.values,
"ChainID": top.chainID.values,
"ResidueID": top.resSeq.values + 1,
}
for key, val in iteritems(data):
node = self._get_node(where='/', name=key)[:] = val[:]
node[:] = val[:]
def _typecast(self, namespace):
"""Work around for the argparse bug with respect to defaults and FileType not
playing together nicely -- http://stackoverflow.com/questions/8236954/specifying-default-filenames-with-argparse-but-not-opening-them-on-help"""
for name, type in iteritems(self.name_to_type):
setattr(namespace, name, type(getattr(namespace, name)))
return namespace
def assert_dict_equal(t1, t2, decimal=6):
"""Assert two dicts are equal. This method should actually
work for any dict of numpy arrays/objects
Parameters
----------
t1 : object
t2 : object
decimal : int
Number of decimal places to check, for arrays inside the dicts
"""
# make sure the keys are the same
eq_(list(t1.keys()), list(t2.keys()))
for key, val in iteritems(t1):
# compare numpy arrays using numpy.testing
if isinstance(val, np.ndarray) or ('pandas' in sys.modules and isinstance(t1, pd.DataFrame)):
if val.dtype.kind == 'f':
# compare floats for almost equality
assert_array_almost_equal(val, t2[key], decimal)
else:
# compare everything else (ints, bools) for absolute equality
assert_array_equal(val, t2[key])
else:
eq_(val, t2[key])
def topology(self, top):
"""Set the topology in the file
Parameters
----------
top : mdtraj.Topology
A topology object
"""
_check_mode(self.mode, ('w',))
if self._needs_initialization:
self._initialize_headers(top.n_atoms)
self._needs_initialization = False
top, bonds = top.to_dataframe()
data = {
"AtomID": top.index.values + 1,
"AtomNames": top.name.values,
"ResidueNames": top.resName.values,
"ChainID": top.chainID.values,
"ResidueID": top.resSeq.values + 1,
}
for key, val in iteritems(data):
node = self._get_node(where='/', name=key)[:] = val[:]
node[:] = val[:]
def assert_dict_equal(t1, t2, decimal=6):
"""Assert two dicts are equal. This method should actually
work for any dict of numpy arrays/objects
Parameters
----------
t1 : object
t2 : object
decimal : int
Number of decimal places to check, for arrays inside the dicts
"""
# make sure the keys are the same
eq_(list(t1.keys()), list(t2.keys()))
for key, val in iteritems(t1):
# compare numpy arrays using numpy.testing
if isinstance(val, np.ndarray) or ('pandas' in sys.modules and isinstance(t1, pd.DataFrame)):
if val.dtype.kind == 'f':
# compare floats for almost equality
assert_array_almost_equal(val, t2[key], decimal)
else:
# compare everything else (ints, bools) for absolute equality
assert_array_equal(val, t2[key])
else:
eq_(val, t2[key])
def _typecast(self, namespace):
"""Work around for the argparse bug with respect to defaults and FileType not
playing together nicely -- http://stackoverflow.com/questions/8236954/specifying-default-filenames-with-argparse-but-not-opening-them-on-help"""
for name, type in iteritems(self.name_to_type):
setattr(namespace, name, type(getattr(namespace, name)))
return namespace
def _str_index(self):
idx = self['index']
out = []
out += ['.. index:: %s' % idx.get('default','')]
for section, references in iteritems(idx):
if section == 'default':
continue
out += [' :%s: %s' % (section, ', '.join(references))]
return out
def _str_index(self):
idx = self['index']
out = []
out += ['.. index:: %s' % idx.get('default', '')]
for section, references in iteritems(idx):
if section == 'default':
continue
out += [' :%s: %s' % (section, ', '.join(references))]
return out
def save_legacy_hdf(traj, filename):
"""Saves an MDTraj Trajectory as an MSMB2 lh5 file.
Parameters
----------
traj : MDTraj.Trajectory
Trajectory object to be saved
filename : str
String filename of HDF Trajectory file.
"""
MAXINT16 = np.iinfo(np.int16).max
MAXINT32 = np.iinfo(np.int32).max
DEFAULT_PRECISION = 1000
def _convert_to_lossy_integers(X, precision):
"""Implementation of the lossy compression used in Gromacs XTC using the pytables library. Convert 32 bit floats into 16 bit integers. These conversion functions have been optimized for memory use. Further memory reduction would require an in-place astype() operation, which one could create using ctypes."""
if np.max(X) * float(precision) < MAXINT16 and np.min(X) * float(precision) > -MAXINT16:
X *= float(precision)
Rounded = X.astype("int16")
X /= float(precision)
else:
X *= float(precision)
Rounded = X.astype("int32")
X /= float(precision)
logger.error("Data range too large for int16: try removing center of mass motion, check for 'blowing up, or just use .h5 or .xtc format.'")
return(Rounded)
top, bonds = traj.top.to_dataframe()
data_dict = {}
data_dict["AtomID"] = top.index.values + 1
data_dict["AtomNames"] = top.name.values
data_dict["ResidueNames"] = top.resName.values
data_dict["ChainID"] = top.chainID.values
data_dict["ResidueID"] = top.resSeq.values + 1
data_dict["XYZList"] = _convert_to_lossy_integers(traj.xyz, DEFAULT_PRECISION)
atom_dict = {}
atom_dict["AtomID"] = tables.Int64Atom()
atom_dict["AtomNames"] = tables.StringAtom(itemsize=4)
atom_dict["ResidueNames"] = tables.StringAtom(itemsize=4)
atom_dict["ChainID"] = tables.StringAtom(itemsize=1)
atom_dict["ResidueID"] = tables.Int64Atom()
atom_dict["XYZList"] = tables.Int16Atom()
file_handle = tables.File(filename, 'w')
for key, val in iteritems(data_dict):
node = file_handle.createCArray(where='/', name=key, atom=atom_dict[key], shape=val.shape, filters=COMPRESSION)
node[:] = val[:]
file_handle.close()
def _delayed_init(self):
import cffi
ffi = cffi.FFI()
# some platforms dont have all of these, especially wierd compilers or 32 bit machines
nptypes = [getattr(np, name) for name in self.nptype_names if hasattr(np, name)]
nptype_descr = dict([('%s%d' % (dtype.kind, dtype.itemsize), dtype) for dtype in map(np.dtype, nptypes)])
casts = {}
for code, names in iteritems(self.ctypes):
for name in names:
casts[nptype_descr['%s%d' % (code, ffi.sizeof(name))]] = name + ' *'
# casts is a dict that helps us cast numpy arrays, like
# {np.float32 : 'float *', np.int32: 'int *'}
self._ffi = ffi
self._casts = casts
def __init__(self, traj, path, run_clone_gen, traj_len):
try:
os.mkdir(path + "/")
except OSError:
pass
for (run, clone), num_gens in iteritems(run_clone_gen):
try:
os.mkdir(path + "/RUN%d/" % run)
except OSError:
pass
os.mkdir(path + "/RUN%d/CLONE%d" % (run, clone))
for gen in range(num_gens):
randomize_coordinates(traj, traj_len)
traj.save(path + "/RUN%d/CLONE%d/frame%d.xtc" % (run, clone, gen))
def __init__(self, traj, path, run_clone_gen, traj_len):
try:
os.mkdir(path + "/")
except OSError:
pass
for (run, clone), num_gens in iteritems(run_clone_gen):
try:
os.mkdir(path + "/RUN%d/" % run)
except OSError:
pass
os.mkdir(path + "/RUN%d/CLONE%d" % (run, clone))
for gen in range(num_gens):
randomize_coordinates(traj, traj_len)
traj.save(path + "/RUN%d/CLONE%d/frame%d.xtc" %
(run, clone, gen))
def _delayed_init(self):
import cffi
ffi = cffi.FFI()
# some platforms dont have all of these, especially wierd compilers or 32 bit machines
nptypes = [
getattr(np, name) for name in self.nptype_names
if hasattr(np, name)
]
nptype_descr = dict([('%s%d' % (dtype.kind, dtype.itemsize), dtype)
for dtype in map(np.dtype, nptypes)])
casts = {}
for code, names in iteritems(self.ctypes):
for name in names:
casts[nptype_descr['%s%d' %
(code, ffi.sizeof(name))]] = name + ' *'
# casts is a dict that helps us cast numpy arrays, like
# {np.float32 : 'float *', np.int32: 'int *'}
self._ffi = ffi
self._casts = casts
def invert_assignments(assignments):
"""Invert an assignments array -- that is, produce a mapping
from state -> traj/frame
Parameters
----------
assignments : np.ndarray
2D array of MSMBuilder assignments
Returns
-------
inverse_mapping : collections.defaultdict
Mapping from state -> traj,frame, such that inverse_mapping[s]
gives the conformations assigned to state s.
Notes
-----
The assignments array may have -1's, which are simply placeholders
we do not add these to the inverted assignments. Therefore, doing
the following will raise a KeyError:
>>> inv_assignments = MSMLib.invert_assignments(assignments)
>>> print inv_assignments[-1]
KeyError: -1
"""
check_assignment_array_input(assignments)
inverse_mapping = defaultdict(lambda: ([], []))
non_neg_inds = np.array(np.where(assignments != -1)).T
# we do not care about -1's
for (i, j) in non_neg_inds:
inverse_mapping[assignments[i, j]][0].append(i)
inverse_mapping[assignments[i, j]][1].append(j)
# convert from lists to numpy arrays
for key, (trajs, frames) in iteritems(inverse_mapping):
inverse_mapping[key] = (np.array(trajs), np.array(frames))
return inverse_mapping
def assert_dict_equal(t1, t2, decimal=6):
"""
Assert two dicts are equal.
This method should actually
work for any dict of numpy arrays/objects
"""
# make sure the keys are the same
eq_(list(t1.keys()), list(t2.keys()))
for key, val in iteritems(t1):
# compare numpy arrays using numpy.testing
if isinstance(val, np.ndarray):
if val.dtype.kind == 'f':
# compare floats for almost equality
assert_array_almost_equal(val, t2[key], decimal)
else:
# compare everything else (ints, bools) for absolute equality
assert_array_equal(val, t2[key])
else:
eq_(val, t2[key])
def assert_dict_equal(t1, t2, decimal=6):
"""
Assert two dicts are equal.
This method should actually
work for any dict of numpy arrays/objects
"""
# make sure the keys are the same
eq_(list(t1.keys()), list(t2.keys()))
for key, val in iteritems(t1):
# compare numpy arrays using numpy.testing
if isinstance(val, np.ndarray):
if val.dtype.kind == 'f':
# compare floats for almost equality
assert_array_almost_equal(val, t2[key], decimal)
else:
# compare everything else (ints, bools) for absolute equality
assert_array_equal(val, t2[key])
else:
eq_(val, t2[key])
def main(args, verbose=True):
"""Run the main script.
Parameters
----------
args : argparse.Namespace
The collected command line arguments
"""
if args.atom_indices is not None:
atom_indices = np.loadtxt(args.atom_indices, int)
else:
atom_indices = None
out_x = ext(args.output)
out_units = units[out_x]
out_fields = fields[out_x]
OutFileFormat = formats[out_x]
in_x = ext(args.input[0])
InFileFormat = formats[in_x]
if args.topology is not None:
topology = _parse_topology(args.topology)
else:
topology = None
if topology is not None and atom_indices is not None:
topology = topology.subset(atom_indices)
n_total = 0
if args.index is not None:
assert len(args.input) == 1
# when chunk is None, we load up ALL of the frames. this isn't
# strictly necessary, and it costs more memory, but it's ALOT
# harder to get the code correct when we need to use data[start:end]
# notation when all of the data isn't loaded up at once. it's easy
# for hdf5 and netcdf, but for the others...
assert args.chunk is None
# this is the normal invocation pattern, but for PDBTrajectoryFile it's
# different
outfile_factory = functools.partial(OutFileFormat, args.output, 'w',
force_overwrite=args.force)
with outfile_factory() as outfile:
for fn in args.input:
assert in_x == ext(fn)
with InFileFormat(fn, 'r') as infile:
while True:
data, in_units, n_frames = read(infile, args.chunk, stride=args.stride,
atom_indices=atom_indices)
if n_frames == 0:
break
if topology is not None:
# if the user supplied a topology, we should probably
# do some simple checks
if data['xyz'].shape[1] != topology._numAtoms:
warnings.warn('sdsfsd!!!!')
data['topology'] = topology
# if they want a specific set of frames, get those
# with slice notation
if args.index is not None:
_data = {}
for k, v in iteritems(data):
if isinstance(v, np.ndarray):
# we don't want the dimensionality to go deficient
if isinstance(args.index, int):
_data[k] = v[np.newaxis, args.index]
else:
_data[k] = v[args.index]
elif isinstance(v, md.Topology):
_data[k] = v
else:
raise RuntineError()
data = _data
print(list(data.keys()))
n_frames = len(data['xyz'])
convert(data, in_units, out_units, out_fields)
write(outfile, data)
n_total += n_frames
if verbose:
sys.stdout.write('\rconverted %d frames, %d atoms' % (n_total, data['xyz'].shape[1]))
sys.stdout.flush()
if verbose:
print(' ')
def saveh(file, *args, **kwargs):
"""Save several numpy arrays into a single file in compressed ``.hdf`` format.
If arguments are passed in with no keywords, the corresponding variable
names, in the ``.hdf`` file, are 'arr_0', 'arr_1', etc. If keyword arguments
are given, the corresponding variable names, in the ``.hdf`` file will
match the keyword names.
Parameters
----------
file : str or tables.File
Either the file name (string) or an open pytables file
(file-like object opened with tables.openFile(...))
where the data will be saved.
args : Arguments, optional
Arrays to save to the file. Since it is not possible for Python to
know the names of the arrays outside `savez`, the arrays will be saved
with names "arr_0", "arr_1", and so on. These arguments can be any
expression.
kwds : Keyword arguments, optional
Arrays to save to the file. Arrays will be saved in the file with the
keyword names.
Notes
-----
`saveh` will overwrite files by default. If you have an hdf5 that contains the
arrays `arr_0` and `arr_1` and you attempt to save a new array `x`, it will
go in side by side. But if you save a new `arr_0`, it will overwrite your
previous array.
Returns
-------
None
Raises
------
TypeError
When arrays are of an unsupported type
See Also
--------
numpy.savez : Save several arrays into a single file in uncompressed ``.npz`` format.
"""
if isinstance(file, basestring):
if TABLES2:
handle = tables.openFile(file, 'a')
else:
handle = tables.open_file(file, 'a')
own_fid = True
else:
if not isinstance(file, tables.File):
raise TypeError('file must be either a string '
'or an open tables.File: %s' % file)
handle = file
own_fid = False
# name all the arrays
namedict = kwargs
for i, val in enumerate(args):
key = 'arr_%d' % i
if key in namedict.keys():
if own_fid:
handle.close()
raise ValueError('Cannot use un-named variables '
' and keyword %s' % key)
namedict[key] = val
# ensure that they don't already exist
if TABLES2:
current_nodes = [e.name for e in handle.listNodes(where='/')]
else:
current_nodes = [e.name for e in handle.list_nodes(where='/')]
for key in namedict.keys():
if key in current_nodes:
if TABLES2:
handle.removeNode('/', name=key)
else:
handle.remove_node('/', name=key)
# per discussion on github, https://github.com/rmcgibbo/mdtraj/issues/5
# silent overwriting appears to be the desired functionality
# raise IOError('Array already exists in file: %s' % key)
# save all the arrays
try:
for key, val in iteritems(namedict):
if not isinstance(val, np.ndarray):
raise TypeError('Only numpy arrays can '
'be saved: type(%s) is %s' % (key, type(val)))
try:
atom = tables.Atom.from_dtype(val.dtype)
except ValueError:
raise TypeError('Arrays of this dtype '
'cannot be saved: %s' % val.dtype)
if TABLES2:
node = handle.createCArray(where='/',
name=key,
atom=atom,
shape=val.shape,
filters=COMPRESSION)
else:
node = handle.create_carray(where='/',
name=key,
atom=atom,
shape=val.shape,
filters=COMPRESSION)
node[:] = val
except Exception:
handle.close()
if own_fid:
os.unlink(file)
raise
handle.flush()
if own_fid:
handle.close()
def main(args, verbose=True):
"""Run the main script.
Parameters
----------
args : argparse.Namespace
The collected commandline arguments
"""
if args.atom_indices is not None:
atom_indices = np.loadtxt(args.atom_indices, int)
else:
atom_indices = None
out_x = ext(args.output)
out_units = units[out_x]
out_fields = fields[out_x]
OutFileFormat = formats[out_x]
in_x = ext(args.input[0])
InFileFormat = formats[in_x]
if args.topology is not None:
topology = md.PDBTrajectoryFile(args.topology).topology
else:
topology = None
if topology is not None and atom_indices is not None:
topology = topology.subset(atom_indices)
n_total = 0
if args.index is not None:
assert len(args.input) == 1
# when chunk is None, we load up ALL of the frames. this isn't
# strictly necessary, and it costs more memory, but it's ALOT
# harder to get the code correct when we need to use data[start:end]
# notation when all of the data isn't loaded up at once. it's easy
# for hdf5 and netcdf, but for the others...
assert args.chunk is None
# this is the normal invocation pattern, but for PDBTrajectoryFile it's
# different
outfile_factory = functools.partial(OutFileFormat,
args.output,
'w',
force_overwrite=args.force)
with outfile_factory() as outfile:
for fn in args.input:
assert in_x == ext(fn)
with InFileFormat(fn, 'r') as infile:
while True:
data, in_units, n_frames = read(infile,
args.chunk,
stride=args.stride,
atom_indices=atom_indices)
if n_frames == 0:
break
if topology is not None:
# if the user supplied a topology, we should probably
# do some simple checks
if data['xyz'].shape[1] != topology._numAtoms:
warnings.warn('sdsfsd!!!!')
data['topology'] = topology
# if they want a specific set of frames, get those
# with slice notation
if args.index is not None:
_data = {}
for k, v in iteritems(data):
if isinstance(v, np.ndarray):
# we don't want the dimensionality to go deficient
if isinstance(args.index, int):
_data[k] = v[np.newaxis, args.index]
else:
_data[k] = v[args.index]
elif isinstance(v, md.Topology):
_data[k] = v
else:
raise RuntineError()
data = _data
print(list(data.keys()))
n_frames = len(data['xyz'])
convert(data, in_units, out_units, out_fields)
write(outfile, data)
n_total += n_frames
if verbose:
sys.stdout.write('\rconverted %d frames, %d atoms' %
(n_total, data['xyz'].shape[1]))
sys.stdout.flush()
if verbose:
print(' ')
def saveh(file, *args, **kwargs):
"""Save several numpy arrays into a single file in compressed ``.hdf`` format.
If arguments are passed in with no keywords, the corresponding variable
names, in the ``.hdf`` file, are 'arr_0', 'arr_1', etc. If keyword arguments
are given, the corresponding variable names, in the ``.hdf`` file will
match the keyword names.
Parameters
----------
file : str or tables.File
Either the file name (string) or an open pytables file
(file-like object opened with tables.openFile(...))
where the data will be saved.
args : Arguments, optional
Arrays to save to the file. Since it is not possible for Python to
know the names of the arrays outside `savez`, the arrays will be saved
with names "arr_0", "arr_1", and so on. These arguments can be any
expression.
kwds : Keyword arguments, optional
Arrays to save to the file. Arrays will be saved in the file with the
keyword names.
Notes
-----
`saveh` will overwrite files by default. If you have an hdf5 that contains the
arrays `arr_0` and `arr_1` and you attempt to save a new array `x`, it will
go in side by side. But if you save a new `arr_0`, it will overwrite your
previous array.
Returns
-------
None
Raises
------
TypeError
When arrays are of an unsupported type
See Also
--------
numpy.savez : Save several arrays into a single file in uncompressed ``.npz`` format.
"""
if isinstance(file, basestring):
handle = tables.openFile(file, 'a')
own_fid = True
else:
if not isinstance(file, tables.File):
raise TypeError('file must be either a string '
'or an open tables.File: %s' % file)
handle = file
own_fid = False
# name all the arrays
namedict = kwargs
for i, val in enumerate(args):
key = 'arr_%d' % i
if key in namedict.keys():
if own_fid:
handle.close()
raise ValueError('Cannot use un-named variables '
' and keyword %s' % key)
namedict[key] = val
# ensure that they don't already exist
current_nodes = [e.name for e in handle.listNodes(where='/')]
for key in namedict.keys():
if key in current_nodes:
handle.removeNode('/', name=key)
# per discussion on github, https://github.com/rmcgibbo/mdtraj/issues/5
# silent overwriting appears to be the desired functionality
# raise IOError('Array already exists in file: %s' % key)
# save all the arrays
try:
for key, val in iteritems(namedict):
if not isinstance(val, np.ndarray):
raise TypeError('Only numpy arrays can '
'be saved: type(%s) is %s' % (key, type(val)))
try:
atom = tables.Atom.from_dtype(val.dtype)
except ValueError:
raise TypeError('Arrays of this dtype '
'cannot be saved: %s' % val.dtype)
node = handle.createCArray(where='/', name=key,
atom=atom, shape=val.shape, filters=COMPRESSION)
node[:] = val
except Exception:
handle.close()
if own_fid:
os.unlink(file)
raise
handle.flush()
if own_fid:
handle.close()
