def write_dcd(in_name, out_name, remarks='testing', header=None):
with DCDFile(in_name) as f_in, DCDFile(out_name, 'w') as f_out:
if header is None:
header = f_in.header
f_out.write_header(**header)
for frame in f_in:
f_out.write(xyz=frame.xyz, box=frame.unitcell)
def test_written_remarks_property(remarks, tmpdir_factory):
# property based testing for writing of a wide range of string
# values to REMARKS field
dcd = DCDFile(DCD)
dirname = str(id(remarks)) + "_"
testfile = str(tmpdir_factory.mktemp(dirname).join('test.dcd'))
header = dcd.header
header['remarks'] = remarks
write_dcd(DCD, testfile, header=header)
expected_remarks = remarks
with DCDFile(testfile) as f:
assert f.header['remarks'] == expected_remarks
def test_write_random_unitcell(tmpdir):
testname = str(tmpdir.join('test.dcd'))
rstate = np.random.RandomState(1178083)
random_unitcells = rstate.uniform(high=80, size=(98, 6)).astype(np.float64)
with DCDFile(DCD) as f_in, DCDFile(testname, 'w') as f_out:
header = f_in.header
header['is_periodic'] = True
f_out.write_header(**header)
for index, frame in enumerate(f_in):
f_out.write(xyz=frame.xyz, box=random_unitcells[index])
with DCDFile(testname) as test:
for index, frame in enumerate(test):
assert_array_almost_equal(frame.unitcell, random_unitcells[index])
def test_read_coordsshape(dcdfile, natoms):
# confirm shape of coordinate data against result from previous
# MDAnalysis implementation of DCD file handling
with DCDFile(dcdfile) as dcd:
dcd_frame = dcd.read()
xyz = dcd_frame[0]
assert xyz.shape == (natoms, 3)
def written_dcd(tmpdir_factory):
with DCDFile(DCD) as dcd:
header = dcd.header
testfile = tmpdir_factory.mktemp('dcd').join('test.dcd')
testfile = str(testfile)
write_dcd(DCD, testfile)
Result = namedtuple("Result", "testfile, header, orgfile")
return Result(testfile, header, DCD)
def test_readframes(dcdfile, legacy_data, frame_idx):
legacy = np.load(legacy_data)
with DCDFile(dcdfile) as dcd:
frames = dcd.readframes()
xyz = frames.xyz
assert_equal(len(xyz), len(dcd))
for index, frame_num in enumerate(frame_idx):
assert_array_almost_equal(xyz[frame_num], legacy[index])
def test_nframessize_int(dcdfile):
# require that the (nframessize / framesize) value used by DCDFile
# is an integer (because nframessize / framesize + 1 = total frames,
# which must also be an int)
filesize = os.path.getsize(dcdfile)
with DCDFile(dcdfile) as dcd:
nframessize = filesize - dcd._header_size - dcd._firstframesize
assert float(nframessize) % float(dcd._framesize) == 0
def test_relative_frame_sizes(dcdfile):
# the first frame of a DCD file should always be >= in size
# to subsequent frames, as the first frame contains the same
# atoms + (optional) fixed atoms
with DCDFile(dcdfile) as dcd:
first_frame_size = dcd._firstframesize
general_frame_size = dcd._framesize
assert first_frame_size >= general_frame_size
def written_dcd(tmpdir_factory):
with DCDFile(DCD) as dcd:
header = dcd.header
testfile = tmpdir_factory.mktemp('dcd').join('test.dcd')
testfile = str(testfile)
write_dcd(DCD, testfile)
Result = namedtuple("Result", "testfile, header, orgfile")
# throw away last char we didn't save due to null termination
header['remarks'] = header['remarks'][:-1]
return Result(testfile, header, DCD)
def test_file_size_breakdown(dcdfile):
# the size of a DCD file is equivalent to the sum of the header
# size, first frame size, and (N - 1 frames) * size per general
# frame
expected = os.path.getsize(dcdfile)
with DCDFile(dcdfile) as dcd:
actual = dcd._header_size + dcd._firstframesize + (
(dcd.n_frames - 1) * dcd._framesize)
assert actual == expected
def test_written_remarks_property(remarks, tmpdir, dcd):
# property based testing for writing of a wide range of string
# values to REMARKS field
testfile = str(tmpdir.join('test.dcd'))
header = dcd.header
header['remarks'] = remarks
write_dcd(DCD, testfile, header=header)
expected_remarks = remarks
with DCDFile(testfile) as f:
assert f.header['remarks'] == expected_remarks
def test_read_coord_values(dcdfile, legacy_data, frames):
# test the actual values of coordinates read in versus
# stored values read in by the legacy DCD handling framework
# to reduce repo storage burden, we only compare for a few
# randomly selected frames
legacy = np.load(legacy_data)
with DCDFile(dcdfile) as dcd:
for index, frame_num in enumerate(frames):
dcd.seek(frame_num)
actual_coords = dcd.read()[0]
desired_coords = legacy[index]
assert_array_equal(actual_coords, desired_coords)
def test_write_header(tmpdir):
# test that _write_header() can produce a very crude
# header for a new / empty file
testfile = str(tmpdir.join('test.dcd'))
with DCDFile(testfile, 'w') as dcd:
dcd.write_header(
remarks='Crazy!',
natoms=22,
istart=12,
nsavc=10,
delta=0.02,
is_periodic=1)
with DCDFile(testfile) as dcd:
header = dcd.header
assert header['remarks'] == 'Crazy!'
assert header['natoms'] == 22
assert header['istart'] == 12
assert header['is_periodic'] == 1
assert header['nsavc'] == 10
assert np.allclose(header['delta'], .02)
def test_write_all_dtypes(tmpdir, dtype):
fname = str(tmpdir.join('foo.dcd'))
with DCDFile(fname, 'w') as out:
natoms = 10
xyz = np.ones((natoms, 3), dtype=dtype)
box = np.ones(6, dtype=dtype)
out.write_header(remarks='test',
natoms=natoms,
is_periodic=1,
delta=1,
nsavc=1,
istart=1)
out.write(xyz=xyz, box=box)
def test_write_array_like(tmpdir, array_like):
fname = str(tmpdir.join('foo.dcd'))
with DCDFile(fname, 'w') as out:
natoms = 10
xyz = array_like([[1, 1, 1] for i in range(natoms)])
box = array_like([i for i in range(6)])
out.write_header(remarks='test',
natoms=natoms,
is_periodic=1,
delta=1,
nsavc=1,
istart=1)
out.write(xyz=xyz, box=box)
def test_write_wrong_shape_box(tmpdir):
fname = str(tmpdir.join('foo.dcd'))
with DCDFile(fname, 'w') as out:
natoms = 10
xyz = np.ones((natoms, 3))
box = np.ones(7)
out.write_header(remarks='test',
natoms=natoms,
is_periodic=1,
delta=1,
nsavc=1,
istart=1)
with pytest.raises(ValueError):
out.write(xyz=xyz, box=box)
def test_write_header(tmpdir):
# test that _write_header() can produce a very crude
# header for a new / empty file
testfile = str(tmpdir.join('test.dcd'))
with DCDFile(testfile, 'w') as dcd:
dcd.write_header(remarks='Crazy!',
natoms=22,
istart=12,
nsavc=10,
delta=0.02,
is_periodic=1)
with DCDFile(testfile) as dcd:
header = dcd.header
assert header['remarks'] == 'Crazy!'
assert header['natoms'] == 22
assert header['istart'] == 12
assert header['is_periodic'] == 1
assert header['nsavc'] == 10
assert np.allclose(header['delta'], .02)
# we also check the bytes written directly.
with open(testfile, 'rb') as fh:
header_bytes = fh.read()
# check for magic number
assert struct.unpack('i', header_bytes[:4])[0] == 84
# magic number should be written again before remark section
assert struct.unpack('i', header_bytes[88:92])[0] == 84
# length of remark section. We hard code this to 244 right now
assert struct.unpack('i', header_bytes[92:96])[0] == 244
# say we have 3 block of length 80
assert struct.unpack('i', header_bytes[96:100])[0] == 3
# after the remark section the length should be reported again
assert struct.unpack('i', header_bytes[340:344])[0] == 244
# this is a magic number as far as I see
assert struct.unpack('i', header_bytes[344:348])[0] == 4
def test_write_header_twice(tmpdir):
# an IOError should be raised if a duplicate
# header writing is attempted
header = {
"remarks": 'Crazy!',
"natoms": 22,
"istart": 12,
"nsavc": 10,
"delta": 0.02,
"is_periodic": 1
}
fname = str(tmpdir.join('test.dcd'))
with DCDFile(fname, 'w') as dcd:
dcd.write_header(**header)
with pytest.raises(IOError):
dcd.write_header(**header)
def aa_fetch():
raw_filepath = "aa0.dcd"
record_file = "aa.tfrecords"
if os.path.isfile(record_file):
print("Found existing record file, delete if you want to re-fetch")
return record_file
if not os.path.isfile(raw_filepath):
print("Downloading AA data...", end="")
urllib.request.urlretrieve(
"https://ndownloader.figshare.com/files/1497002", raw_filepath)
print("File downloaded")
else:
print(
f"Found downloaded file {raw_filepath}, delete if you want to redownload"
)
print("Converting...")
try:
import MDAnalysis
from MDAnalysis.lib.formats.libdcd import DCDFile
except ImportError:
raise ImportError("Please install MDanalysis with pip first")
with tf.io.TFRecordWriter(
record_file,
options=tf.io.TFRecordOptions(compression_type="GZIP")) as writer:
with DCDFile(raw_filepath) as dcd:
for frame in dcd:
feature = {
"coords":
tf.train.Feature(float_list=tf.train.FloatList(
value=frame.xyz.flatten()))
}
example = tf.train.Example(features=tf.train.Features(
feature=feature))
writer.write(example.SerializeToString())
return record_file
def dcd():
with DCDFile(DCD) as dcd:
yield dcd
def test_written_coord_match(written_dcd):
with DCDFile(written_dcd.testfile) as test, DCDFile(
written_dcd.orgfile) as ref:
for frame, o_frame in zip(test, ref):
assert_array_almost_equal(frame.xyz, o_frame.xyz)
def test_read_unit_cell(dcdfile, unit_cell):
# confirm unit cell read against result from previous
# MDAnalysis implementation of DCD file handling
with DCDFile(dcdfile) as dcd:
dcd_frame = dcd.read()
assert_array_almost_equal(dcd_frame.unitcell, unit_cell)
def test_seek_over_max():
with DCDFile(DCD) as dcd:
with pytest.raises(EOFError):
dcd.seek(102)
def test_written_header(written_dcd):
header = written_dcd.header
with DCDFile(written_dcd.testfile) as dcd:
dcdheader = dcd.header
assert dcdheader == header
def test_write_no_header(tmpdir):
fname = str(tmpdir.join('test.dcd'))
with DCDFile(fname, 'w') as dcd:
with pytest.raises(IOError):
dcd.write(np.ones(3), np.ones(6))
def test_written_num_frames(written_dcd):
with DCDFile(written_dcd.testfile) as dcd, DCDFile(
written_dcd.orgfile) as other:
assert len(dcd) == len(other)
def test_written_dcd_coordinate_data_shape(written_dcd):
with DCDFile(written_dcd.testfile) as dcd, DCDFile(
written_dcd.orgfile) as other:
for frame, other_frame in zip(dcd, other):
assert frame.xyz.shape == other_frame.xyz.shape
def test_written_seek(written_dcd):
# ensure that we can seek properly on written DCD file
with DCDFile(written_dcd.testfile) as f:
f.seek(40)
assert_equal(f.tell(), 40)
def test_is_periodic(dcdfile, is_periodic):
with DCDFile(dcdfile) as f:
assert f.is_periodic == is_periodic
def test_written_unit_cell(written_dcd):
with DCDFile(written_dcd.testfile) as test, DCDFile(
written_dcd.orgfile) as ref:
for frame, o_frame in zip(test, ref):
assert_array_almost_equal(frame.unitcell, o_frame.unitcell)
