def __init__(self, file, header_fields=None):
"""Initialize a new instance.
Parameters
----------
file : file-like or str
Stream or filename from which to read the data. The stream
is allowed to be already opened in ``'rb'`` mode.
header_fields : sequence of dicts, optional
Definition of the fields in the header (per row), each
containing key-value pairs for the following keys:
- ``'name'`` : Label for the field.
- ``'offset'`` : Start of the field in bytes.
- ``'size'`` : Size of the field in bytes.
- ``'dtype'`` : Data type in Numpy- or Numpy-readable format.
- ``'dshape'`` (optional) : The array of values is reshaped to
this shape.
- ``'description'`` (optional) : A human-readable description
of the field.
For the default ``None``, the MRC2014 format is used, see
`print_mrc2014_spec`.
"""
if header_fields is None:
header_fields = header_fields_from_table(
spec_table=MRC_2014_SPEC_TABLE,
keys=MRC_SPEC_KEYS,
dtype_map=MRC_DTYPE_TO_NPY_DTYPE)
super().__init__(file, header_fields)
def __init__(self, file, header_fields=None):
"""Initialize a new instance.
Parameters
----------
file : file-like or str
Stream or filename from which to read the data. The stream
is allowed to be already opened in ``'rb'`` mode.
header_fields : sequence of dicts, optional
Definition of the fields in the header (per row), each
containing key-value pairs for the following keys:
- ``'name'`` : Label for the field.
- ``'offset'`` : Start of the field in bytes.
- ``'size'`` : Size of the field in bytes.
- ``'dtype'`` : Data type in Numpy- or Numpy-readable format.
- ``'dshape'`` (optional) : The array of values is reshaped to
this shape.
- ``'description'`` (optional) : A human-readable description
of the field.
For the default ``None``, the MRC2014 format is used, see
`print_mrc2014_spec`.
"""
if header_fields is None:
header_fields = header_fields_from_table(
spec_table=MRC_2014_SPEC_TABLE,
keys=MRC_SPEC_KEYS,
dtype_map=MRC_DTYPE_TO_NPY_DTYPE)
super().__init__(file, header_fields)
def mrc_header_from_params(shape, dtype, kind, **kwargs):
"""Create a minimal MRC2014 header from the given parameters.
Parameters
----------
shape : 3-sequence of ints
3D shape of the stored data. The values are used as
``'nx', 'ny', 'nz'`` header entries, in this order. Note that
this is different from the actual data storage shape for
non-trivial ``axis_order``.
dtype : {'int8', 'int16', 'float32', 'uint16'}
Data type specifier as understood by `numpy.dtype`. It is
translated to a ``'mode'`` header entry. See `this page
<http://www.ccpem.ac.uk/mrc_format/mrc2014.php>`_ for valid
modes.
kind : {'volume', 'projections'}
Interpretation of the 3D data, either as single 3D volume or as
a stack of 2D projections. The value is used for the ``'ispg'``
header entry.
extent : 3-sequence of floats, optional
Size of the 3D volume in meters. The values are used for
the ``'cella'`` header entry.
Default: ``shape``, resulting in ``(1, 1, 1)`` unit cells
axis_order : permutation of ``(0, 1, 2)`` optional
Order of the data axes as they should appear in the stored file.
The values are used for the ``'mapc', 'mapr', 'maps'`` header
entries.
Default: ``(0, 1, 2)``
dmin, dmax : float, optional
Minimum and maximum values of the data, used for header entries
``'dmin'`` and ``'dmax'``, resp.
Default: 1.0, 0.0. These values indicate according to [Che+2015]_
that the values are considered as undetermined.
dmean, rms : float, optional
Mean and variance of the data, used for header entries ``'dmean'``
and ``'rms'``, resp.
Default: ``min(dmin, dmax) - 1, -1.0``. These values indicate
according to [Che+2015]_ that the values are considered as
undetermined.
mrc_version : 2-tuple of int, optional
Version identifier for the MRC file, used for the ``'nversion'``
header entry.
Default: ``(2014, 0)``
text_labels : sequence of strings, optional
Maximal 10 strings with 80 characters each, used for the
``'nlabl'`` and ``'label'`` header entries.
Default: ``[]``
Returns
-------
header : `OrderedDict`
Header stored in an ordered dictionary, where each entry has the
following form::
'name': {'value': value_as_array,
'offset': offset_in_bytes
'description': description_string}
All ``'value'``'s are `numpy.ndarray`'s with at least one
dimension.
References
----------
[Che+2015] Cheng, A et al. *MRC2014: Extensions to the MRC format header
for electron cryo-microscopy and tomography*. Journal of Structural
Biology, 129 (2015), pp 146--150.
"""
# Positional args
shape = [int(n) for n in shape]
kind, kind_in = str(kind).lower(), kind
if kind not in ('volume', 'projections'):
raise ValueError("`kind '{}' not understood".format(kind_in))
# Keyword args
extent = kwargs.pop('extent', shape)
axis_order = kwargs.pop('axis_order', (0, 1, 2))
if tuple(axis_order) not in permutations((0, 1, 2)):
raise ValueError('`axis_order` must be a permutation of (0, 1, 2), '
'got {}'.format(axis_order))
dmin = kwargs.pop('dmin', 1.0)
dmax = kwargs.pop('dmax', 0.0)
dmean = kwargs.pop('dmean', min(dmin, dmax) - 1.0)
rms = kwargs.pop('rms', -1.0)
mrc_version = kwargs.pop('mrc_version', (2014, 0))
if len(mrc_version) != 2:
raise ValueError('`mrc_version` must be a sequence of length 2, got '
'{}'.format(mrc_version))
# Text labels: fill each label up with whitespace to 80 characters.
# Create the remaining labels as 80 * '\x00'
text_labels_in = kwargs.pop('text_labels', [])
nlabl = len(text_labels_in)
if nlabl > 10:
raise ValueError('expexted maximum of 10 labels, got {} labels'
''.format(nlabl))
text_labels = [str(label).ljust(80) for label in text_labels_in]
if any(len(label) > 80 for label in text_labels):
raise ValueError('labels cannot have more than 80 characters each')
# Convert to header-friendly form. Names are required to match
# exactly the header field names, and all of them must exist,
# so that `eval` below succeeds for all fields.
nx, ny, nz = [np.array(n, dtype='int32').reshape([1]) for n in shape]
mode = np.array(NPY_DTYPE_TO_MRC_MODE[np.dtype(dtype)],
dtype='int32').reshape([1])
mx, my, mz = nx, ny, nz
cella = np.array(extent).reshape([3]).astype('float32')
mapc, mapr, maps = [np.array(m, dtype='int32').reshape([1]) + 1
for m in axis_order]
dmin, dmax, dmean, rms = [np.array(x, dtype='float32').reshape([1])
for x in (dmin, dmax, dmean, rms)]
ispg = 1 if kind == 'volume' else 0
ispg = np.array(ispg, dtype='int32', ndmin=1)
nsymbt = np.array([0], dtype='int32')
exttype = np.fromstring(' ', dtype='S1')
nversion = np.array(10 * mrc_version[0] + mrc_version[1],
dtype='int32').reshape([1])
origin = np.zeros(3, dtype='int32')
map = np.fromstring('MAP ', dtype='S1')
# TODO: no idea how to properly choose the machine stamp
machst = np.fromiter(b'DD ', dtype='S1')
nlabl = np.array(nlabl, dtype='int32').reshape([1])
label = np.zeros((10, 80), dtype='S1') # ensure correct size
for i, label_i in enumerate(text_labels):
label[i] = np.fromstring(label_i, dtype='S1')
# Make the header
# We use again the specification to set the values
header_fields = header_fields_from_table(
MRC_2014_SPEC_TABLE, MRC_SPEC_KEYS, MRC_DTYPE_TO_NPY_DTYPE)
header = OrderedDict()
for field in header_fields:
header[field['name']] = {'offset': field['offset'],
'value': eval(field['name'])}
return header
def read_extended_header(self, groupby='field', force_type=''):
"""Read the extended header according to `extended_header_type`.
Currently, only the FEI extended header format is supported.
See `print_fei_ext_header_spec` or `this homepage`_ for the format
specification.
The extended header usually has one header section per
image (slice), in case of the FEI header 128 bytes each, with
a maximum of 1024 sections.
Parameters
----------
groupby : {'field', 'section'}, optional
How to group the values in the extended header sections.
``'field'`` : make an array per section field, e.g.::
'defocus': [dval1, dval2, ..., dvalN],
'exp_time': [tval1, tval2, ..., tvalN],
...
``'section'`` : make a dictionary for each section, e.g.::
{'defocus': dval1, 'exp_time': tval1},
{'defocus': dval2, 'exp_time': tval2},
...
force_type : string, optional
If given, this value overrides the `extended_header_type`
from `header`.
Currently supported: ``'FEI1'``
Returns
-------
ext_header: `OrderedDict` or tuple
For ``groupby == 'field'``, a dictionary with the field names
as keys, like in the example.
For ``groupby == 'section'``, a tuple of dictionaries as
shown above.
The returned data structures store no offsets, in contrast
to the regular header.
See Also
--------
References
----------
.. _this homepage:
http://www.2dx.unibas.ch/documentation/mrc-software/fei-\
extended-mrc-format-not-used-by-2dx
"""
ext_header_type = str(force_type).upper() or self.extended_header_type
if ext_header_type != 'FEI1':
raise ValueError("extended header type '{}' not supported"
"".format(self.extended_header_type))
groupby, groupby_in = str(groupby).lower(), groupby
ext_header_len = int(self.header['nsymbt']['value'])
if ext_header_len % MRC_FEI_SECTION_SIZE:
raise ValueError('extended header length {} from header is '
'not divisible by extended header section size '
'{}'.format(ext_header_len, MRC_FEI_SECTION_SIZE))
num_sections = ext_header_len // MRC_FEI_SECTION_SIZE
if num_sections > MRC_FEI_MAX_SECTIONS:
raise ValueError('calculated number of sections ({}) exceeds '
'maximum number of sections ({})'
''.format(num_sections, MRC_FEI_MAX_SECTIONS))
section_fields = header_fields_from_table(
MRC_FEI_EXT_HEADER_SECTION, keys=MRC_SPEC_KEYS,
dtype_map=MRC_DTYPE_TO_NPY_DTYPE)
if groupby == 'field':
# Make a list for each field and append the values for that
# field. Then create an array from that list and store it
# under the field name.
ext_header = OrderedDict()
for field in section_fields:
value_list = []
field_offset = field['offset']
field_dtype = field['dtype']
field_dshape = field['dshape']
for section in range(num_sections):
# Get the bytestring from the right position in the file,
# unpack it and append the value to the list.
section_start = section * MRC_FEI_SECTION_SIZE
self.file.seek(section_start + field_offset)
num_items = int(np.prod(field_dshape))
size_bytes = num_items * field_dtype.itemsize
packed_value = self.file.read(size_bytes)
fmt = '{}{}'.format(num_items, field_dtype.char)
value_list.append(struct.unpack(fmt, packed_value))
ext_header[field['name']] = np.array(value_list,
dtype=field_dtype)
return ext_header
elif groupby == 'section':
# Loop though the sections and append all values from that
# section to a list. Return it as a tuple.
ext_header = []
for section in range(num_sections):
entry = {}
section_start = section * MRC_FEI_SECTION_SIZE
for field in section_fields:
# Get the bytestring from the right position in the file,
# unpack it and store the value as array in the dict.
self.file.seek(section_start + field['offset'])
num_items = int(np.prod(field['dshape']))
size_bytes = num_items * field['dtype'].itemsize
packed_value = self.file.read(size_bytes)
fmt = '{}{}'.format(num_items, field['dtype'].char)
value = struct.unpack(fmt, packed_value)
# Make each entry a 1-element 1D array as usual
entry[field['name']] = np.array(
value, dtype=field['dtype']).reshape(field['dshape'])
ext_header.append(entry)
return tuple(ext_header)
else:
raise ValueError("`groupby` '{}' not understood"
"".format(groupby_in))
def mrc_header_from_params(shape, dtype, kind, **kwargs):
"""Create a minimal MRC2014 header from the given parameters.
Parameters
----------
shape : 3-sequence of ints
3D shape of the stored data. The values are used as
``'nx', 'ny', 'nz'`` header entries, in this order. Note that
this is different from the actual data storage shape for
non-trivial ``axis_order``.
dtype : {'int8', 'int16', 'float32', 'uint16'}
Data type specifier as understood by `numpy.dtype`. It is
translated to a ``'mode'`` header entry. See `this page
<http://www.ccpem.ac.uk/mrc_format/mrc2014.php>`_ for valid
modes.
kind : {'volume', 'projections'}
Interpretation of the 3D data, either as single 3D volume or as
a stack of 2D projections. The value is used for the ``'ispg'``
header entry.
extent : 3-sequence of floats, optional
Size of the 3D volume in meters. The values are used for
the ``'cella'`` header entry.
Default: ``shape``, resulting in ``(1, 1, 1)`` unit cells
axis_order : permutation of ``(0, 1, 2)`` optional
Order of the data axes as they should appear in the stored file.
The values are used for the ``'mapc', 'mapr', 'maps'`` header
entries.
Default: ``(0, 1, 2)``
dmin, dmax : float, optional
Minimum and maximum values of the data, used for header entries
``'dmin'`` and ``'dmax'``, resp.
Default: 1.0, 0.0. These values indicate according to [Che+2015]_
that the values are considered as undetermined.
dmean, rms : float, optional
Mean and variance of the data, used for header entries ``'dmean'``
and ``'rms'``, resp.
Default: ``min(dmin, dmax) - 1, -1.0``. These values indicate
according to [Che+2015]_ that the values are considered as
undetermined.
mrc_version : 2-tuple of int, optional
Version identifier for the MRC file, used for the ``'nversion'``
header entry.
Default: ``(2014, 0)``
text_labels : sequence of strings, optional
Maximal 10 strings with 80 characters each, used for the
``'nlabl'`` and ``'label'`` header entries.
Default: ``[]``
Returns
-------
header : `OrderedDict`
Header stored in an ordered dictionary, where each entry has the
following form::
'name': {'value': value_as_array,
'offset': offset_in_bytes
'description': description_string}
All ``'value'``'s are `numpy.ndarray`'s with at least one
dimension.
References
----------
[Che+2015] Cheng, A et al. *MRC2014: Extensions to the MRC format header
for electron cryo-microscopy and tomography*. Journal of Structural
Biology, 129 (2015), pp 146--150.
"""
# Positional args
shape = [int(n) for n in shape]
kind, kind_in = str(kind).lower(), kind
if kind not in ('volume', 'projections'):
raise ValueError("`kind '{}' not understood".format(kind_in))
# Keyword args
extent = kwargs.pop('extent', shape)
axis_order = kwargs.pop('axis_order', (0, 1, 2))
if tuple(axis_order) not in permutations((0, 1, 2)):
raise ValueError('`axis_order` must be a permutation of (0, 1, 2), '
'got {}'.format(axis_order))
dmin = kwargs.pop('dmin', 1.0)
dmax = kwargs.pop('dmax', 0.0)
dmean = kwargs.pop('dmean', min(dmin, dmax) - 1.0)
rms = kwargs.pop('rms', -1.0)
mrc_version = kwargs.pop('mrc_version', (2014, 0))
if len(mrc_version) != 2:
raise ValueError('`mrc_version` must be a sequence of length 2, got '
'{}'.format(mrc_version))
# Text labels: fill each label up with whitespace to 80 characters.
# Create the remaining labels as 80 * '\x00'
text_labels_in = kwargs.pop('text_labels', [])
nlabl = len(text_labels_in)
if nlabl > 10:
raise ValueError('expexted maximum of 10 labels, got {} labels'
''.format(nlabl))
text_labels = [str(label).ljust(80) for label in text_labels_in]
if any(len(label) > 80 for label in text_labels):
raise ValueError('labels cannot have more than 80 characters each')
# Convert to header-friendly form. Names are required to match
# exactly the header field names, and all of them must exist,
# so that `eval` below succeeds for all fields.
nx, ny, nz = [np.array(n, dtype='int32').reshape([1]) for n in shape]
mode = np.array(NPY_DTYPE_TO_MRC_MODE[np.dtype(dtype)],
dtype='int32').reshape([1])
mx, my, mz = nx, ny, nz
cella = np.array(extent).reshape([3]).astype('float32')
mapc, mapr, maps = [
np.array(m, dtype='int32').reshape([1]) + 1 for m in axis_order
]
dmin, dmax, dmean, rms = [
np.array(x, dtype='float32').reshape([1])
for x in (dmin, dmax, dmean, rms)
]
ispg = 1 if kind == 'volume' else 0
ispg = np.array(ispg, dtype='int32', ndmin=1)
nsymbt = np.array([0], dtype='int32')
exttype = np.fromstring(' ', dtype='S1')
nversion = np.array(10 * mrc_version[0] + mrc_version[1],
dtype='int32').reshape([1])
origin = np.zeros(3, dtype='int32')
map = np.fromstring('MAP ', dtype='S1')
# TODO: no idea how to properly choose the machine stamp
machst = np.fromiter(b'DD ', dtype='S1')
nlabl = np.array(nlabl, dtype='int32').reshape([1])
label = np.zeros((10, 80), dtype='S1') # ensure correct size
for i, label_i in enumerate(text_labels):
label[i] = np.fromstring(label_i, dtype='S1')
# Make the header
# We use again the specification to set the values
header_fields = header_fields_from_table(MRC_2014_SPEC_TABLE,
MRC_SPEC_KEYS,
MRC_DTYPE_TO_NPY_DTYPE)
header = OrderedDict()
for field in header_fields:
header[field['name']] = {
'offset': field['offset'],
'value': eval(field['name'])
}
return header
def read_extended_header(self, groupby='field', force_type=''):
"""Read the extended header according to `extended_header_type`.
Currently, only the FEI extended header format is supported.
See `print_fei_ext_header_spec` or `this homepage`_ for the format
specification.
The extended header usually has one header section per
image (slice), in case of the FEI header 128 bytes each, with
a maximum of 1024 sections.
Parameters
----------
groupby : {'field', 'section'}, optional
How to group the values in the extended header sections.
``'field'`` : make an array per section field, e.g.::
'defocus': [dval1, dval2, ..., dvalN],
'exp_time': [tval1, tval2, ..., tvalN],
...
``'section'`` : make a dictionary for each section, e.g.::
{'defocus': dval1, 'exp_time': tval1},
{'defocus': dval2, 'exp_time': tval2},
...
force_type : string, optional
If given, this value overrides the `extended_header_type`
from `header`.
Currently supported: ``'FEI1'``
Returns
-------
ext_header: `OrderedDict` or tuple
For ``groupby == 'field'``, a dictionary with the field names
as keys, like in the example.
For ``groupby == 'section'``, a tuple of dictionaries as
shown above.
The returned data structures store no offsets, in contrast
to the regular header.
See Also
--------
References
----------
.. _this homepage:
http://www.2dx.unibas.ch/documentation/mrc-software/fei-\
extended-mrc-format-not-used-by-2dx
"""
ext_header_type = str(force_type).upper() or self.extended_header_type
if ext_header_type != 'FEI1':
raise ValueError("extended header type '{}' not supported"
"".format(self.extended_header_type))
groupby, groupby_in = str(groupby).lower(), groupby
ext_header_len = int(self.header['nsymbt']['value'])
if ext_header_len % MRC_FEI_SECTION_SIZE:
raise ValueError('extended header length {} from header is '
'not divisible by extended header section size '
'{}'.format(ext_header_len, MRC_FEI_SECTION_SIZE))
num_sections = ext_header_len // MRC_FEI_SECTION_SIZE
if num_sections > MRC_FEI_MAX_SECTIONS:
raise ValueError('calculated number of sections ({}) exceeds '
'maximum number of sections ({})'
''.format(num_sections, MRC_FEI_MAX_SECTIONS))
section_fields = header_fields_from_table(
MRC_FEI_EXT_HEADER_SECTION,
keys=MRC_SPEC_KEYS,
dtype_map=MRC_DTYPE_TO_NPY_DTYPE)
if groupby == 'field':
# Make a list for each field and append the values for that
# field. Then create an array from that list and store it
# under the field name.
ext_header = OrderedDict()
for field in section_fields:
value_list = []
field_offset = field['offset']
field_dtype = field['dtype']
field_dshape = field['dshape']
for section in range(num_sections):
# Get the bytestring from the right position in the file,
# unpack it and append the value to the list.
section_start = section * MRC_FEI_SECTION_SIZE
self.file.seek(section_start + field_offset)
num_items = int(np.prod(field_dshape))
size_bytes = num_items * field_dtype.itemsize
packed_value = self.file.read(size_bytes)
fmt = '{}{}'.format(num_items, field_dtype.char)
value_list.append(struct.unpack(fmt, packed_value))
ext_header[field['name']] = np.array(value_list,
dtype=field_dtype)
return ext_header
elif groupby == 'section':
# Loop though the sections and append all values from that
# section to a list. Return it as a tuple.
ext_header = []
for section in range(num_sections):
entry = {}
section_start = section * MRC_FEI_SECTION_SIZE
for field in section_fields:
# Get the bytestring from the right position in the file,
# unpack it and store the value as array in the dict.
self.file.seek(section_start + field['offset'])
num_items = int(np.prod(field['dshape']))
size_bytes = num_items * field['dtype'].itemsize
packed_value = self.file.read(size_bytes)
fmt = '{}{}'.format(num_items, field['dtype'].char)
value = struct.unpack(fmt, packed_value)
# Make each entry a 1-element 1D array as usual
entry[field['name']] = np.array(
value, dtype=field['dtype']).reshape(field['dshape'])
ext_header.append(entry)
return tuple(ext_header)
else:
raise ValueError("`groupby` '{}' not understood"
"".format(groupby_in))
