7.37. XDR based trajectory files — MDAnalysis.coordinates.XDR

This module contains helper function and classes to read the XTC and TRR file formats.

See also

MDAnalysis.coordinates.XTC

Read and write GROMACS XTC trajectory files.

MDAnalysis.coordinates.TRR

Read and write GROMACS TRR trajectory files.

MDAnalysis.lib.formats.libmdaxdr

Low level xdr format reader

7.37.1. XDR reader class

The XDRBaseReader contains common functionality for the TRR and XTC reader for GROMACS files, which are implemented in the MDAnalysis.lib.formats.libmdaxdr module.

Both formats have in common that they do not allow native random frame access. Therefore, we first scan the whole trajectory to build an index of frames in the file (“offsets”) as a look-up for seeking to frames. This process is initially slow so we save the offsets to a hidden file next to the trajectory (if possible) and then read the offset file when the trajectory is opened the next time, as described under Offsets.

class MDAnalysis.coordinates.XDR.XDRBaseReader(filename, convert_units=True, sub=None, refresh_offsets=False, dt=None, **kwargs)[source]

Base class for libmdaxdr file formats xtc and trr

This class handles integration of XDR based formats into MDAnalysis. The XTC and TRR classes only implement _write_next_frame() and _frame_to_ts().

Notes

XDR based readers store persistent offsets on disk. The offsets are used to enable access to random frames efficiently. These offsets will be generated automatically the first time the trajectory is opened. Generally offsets are stored in hidden *_offsets.npz files. Afterwards opening the same file again is fast. It sometimes can happen that the stored offsets get out off sync with the trajectory they refer to. For this the offsets also store the number of atoms, size of the file and last modification time. If any of them change the offsets are recalculated. Writing of the offset file can fail when the directory where the trajectory file resides is not writable or if the disk is full. In this case a warning message will be shown but the offsets will nevertheless be used during the lifetime of the trajectory Reader. However, the next time the trajectory is opened, the offsets will have to be rebuilt again.

See _load_offsets() for further details.

Changed in version 1.0.0: XDR offsets read from trajectory if offsets file read-in fails

Changed in version 2.0.0: Add a InterProcessLock when generating offsets

Changed in version 2.4.0: Use a direct read into ts attributes

Changed in version 2.9.0: Changed fasteners.InterProcessLock() to filelock.FileLock

Parameters:

  • filename (str) – trajectory filename

  • convert_units (bool (optional)) – convert units to MDAnalysis units

  • sub (array_like (optional)) – sub is an array of indices to pick out the corresponding coordinates and load only them; this requires that the topology itself is that of the sub system.

  • refresh_offsets (bool (optional)) – force refresh of offsets

  • dt (float (optional)) – timestep in MDAnalysis units to load trajectory with; if dt is None, the time is taken from the xdr file; else, the time is set to dt * frame

  • **kwargs (dict) – General reader arguments.

OtherWriter(filename, **kwargs)

Returns a writer appropriate for filename.

Sets the default keywords start, step and dt (if available). n_atoms is always set from Reader.n_atoms.

See also

Reader.Writer()

Writer(filename, n_atoms=None, **kwargs)[source]

Return writer for trajectory format

_Timestep

alias of Timestep

_apply_transformations(ts)

Applies all the transformations given by the user

_check_for_aux(auxname)

Check for the existance of an auxiliary auxname. If present, return the AuxReader; if not, raise ValueError

_load_offsets()[source]

load frame offsets from file or recalculate if necessary

Frame offsets are cached in an offsets file, which is stored as a hidden file in the same directory as the trajectory. If the file does not exist we generate the offsets and store them.

If the data in the offset file are outdated (older than the trajectory file or different number of frames from the trajectory or different file size) then the offset file is also regenerated.

Note

Generating offsets can take minutes for large trajectories because the whole file must be scanned. During this time, code appears to hang.

You can force regenerating offsets with the refresh_offsets keyword argument for Universe, for example,:

u = mda.Universe(TOPOLOGY, XTC, refresh_offsets=True)

To prevent the competition of generating the same offset file from multiple processes, a filelock.FileLock is used, which is implemented via a lock file (in the same directory as the offset file and ending in “.lock”). This lock file is not automatically deleted because doing so could lead to race conditions.

Once this method completes, the offsets attribute of the underlying reader contains current offsets for the trajectory.

See also

_read_frame(i)[source]

read frame i

_read_frame_with_aux(frame)

Move to frame, updating ts with trajectory, transformations and auxiliary data.

_read_offsets(store=False)[source]

read frame offsets from trajectory

Scan the trajectory for frames and build an index that relates frame number to the position in the file, thus enabling direct seeking to specific frames. The trajectory scan can take minutes for large trajectories.

Parameters:

store (bool) –

Save the frame index (“offsets”) to a file with name generated from the trajectory name (filename) with function offsets_filename(). The offsets file also contains, ctime, file size, number of frames, and number of atoms of the trajectory. The file format is a compressed numpy array (numpy.savez()).

If saving the file fails for any reasons, only a warning is issued.

_reopen()[source]

reopen trajectory

_sliced_iter(start, stop, step)

Generator for slicing a trajectory.

start stop and step are 3 integers describing the slice. Error checking is not done past this point.

A NotImplementedError is raised if random access to frames is not implemented.

add_auxiliary(aux_spec: str | Dict[str, str] = None, auxdata: str | AuxReader = None, format: str = None, **kwargs) → None

Add auxiliary data to be read alongside trajectory.

Auxiliary data may be any data timeseries from the trajectory additional to that read in by the trajectory reader. auxdata can be an AuxReader instance, or the data itself as e.g. a filename; in the latter case an appropriate AuxReader is guessed from the data/file format. An appropriate format may also be directly provided as a key word argument.

On adding, the AuxReader is initially matched to the current timestep of the trajectory, and will be updated when the trajectory timestep changes (through a call to next() or jumping timesteps with trajectory[i]).

The representative value(s) of the auxiliary data for each timestep (as calculated by the AuxReader) are stored in the current timestep in the ts.aux namespace under aux_spec; e.g. to add additional pull force data stored in pull-force.xvg:

u = MDAnalysis.Universe(PDB, XTC)
u.trajectory.add_auxiliary('pull', 'pull-force.xvg')

The representative value for the current timestep may then be accessed as u.trajectory.ts.aux.pull or u.trajectory.ts.aux['pull'].

The following applies to energy readers like the EDRReader.

All data that is present in the (energy) file can be added by omitting aux_spec like so:

u.trajectory.add_auxiliary(auxdata="ener.edr")

aux_spec is expected to be a dictionary that maps the desired attribute name in the ts.aux namespace to the precise data to be added as identified by a data_selector:

term_dict = {"temp": "Temperature", "epot": "Potential"}
u.trajectory.add_auxiliary(term_dict, "ener.edr")

Adding this data can be useful, for example, to filter trajectory frames based on non-coordinate data like the potential energy of each time step. Trajectory slicing allows working on a subset of frames:

selected_frames = np.array([ts.frame for ts in u.trajectory
                            if ts.aux.epot < some_threshold])
subset = u.trajectory[selected_frames]

See also

remove_auxiliary()

Note

Auxiliary data is assumed to be time-ordered, with no duplicates. See the Auxiliary API.

add_transformations(*transformations)

Add all transformations to be applied to the trajectory.

This function take as list of transformations as an argument. These transformations are functions that will be called by the Reader and given a Timestep object as argument, which will be transformed and returned to the Reader. The transformations can be part of the transformations module, or created by the user, and are stored as a list transformations. This list can only be modified once, and further calls of this function will raise an exception.

u = MDAnalysis.Universe(topology, coordinates)
workflow = [some_transform, another_transform, this_transform]
u.trajectory.add_transformations(*workflow)

The transformations are applied in the order given in the list transformations, i.e., the first transformation is the first or innermost one to be applied to the Timestep. The example above would be equivalent to

for ts in u.trajectory:
   ts = this_transform(another_transform(some_transform(ts)))
Parameters:

transform_list (list) – list of all the transformations that will be applied to the coordinates in the order given in the list

See also

MDAnalysis.transformations

Raises:

TypeError – If any transformation is not callable

property aux_list

Lists the names of added auxiliary data.

check_slice_indices(start, stop, step)

Check frame indices are valid and clip to fit trajectory.

The usage follows standard Python conventions for range() but see the warning below.

Parameters:

  • start (int or None) – Starting frame index (inclusive). None corresponds to the default of 0, i.e., the initial frame.

  • stop (int or None) – Last frame index (exclusive). None corresponds to the default of n_frames, i.e., it includes the last frame of the trajectory.

  • step (int or None) – step size of the slice, None corresponds to the default of 1, i.e, include every frame in the range start, stop.

Returns:

start, stop, step – Integers representing the slice

Return type:

tuple (int, int, int)

Warning

The returned values start, stop and step give the expected result when passed in range() but gives unexpected behavior when passed in a slice when stop=None and step=-1

This can be a problem for downstream processing of the output from this method. For example, slicing of trajectories is implemented by passing the values returned by check_slice_indices() to range()

range(start, stop, step)

and using them as the indices to randomly seek to. On the other hand, in MDAnalysis.analysis.base.AnalysisBase the values returned by check_slice_indices() are used to splice the trajectory by creating a slice instance

slice(start, stop, step)

This creates a discrepancy because these two lines are not equivalent:

range(10, -1, -1)             # [10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0]
range(10)[slice(10, -1, -1)]  # []
close()[source]

close reader

convert_forces_from_native(force, inplace=True)

Conversion of forces array force from native to base units

Parameters:

  • force (array_like) – Forces to transform

  • inplace (bool (optional)) – Whether to modify the array inplace, overwriting previous data

Note

By default, the input force is modified in place and also returned. In-place operations improve performance because allocating new arrays is avoided.

New in version 0.7.7.

convert_forces_to_native(force, inplace=True)

Conversion of force array force from base to native units.

Parameters:

  • force (array_like) – Forces to transform

  • inplace (bool (optional)) – Whether to modify the array inplace, overwriting previous data

Note

By default, the input force is modified in place and also returned. In-place operations improve performance because allocating new arrays is avoided.

New in version 0.7.7.

convert_pos_from_native(x, inplace=True)

Conversion of coordinate array x from native units to base units.

Parameters:

  • x (array_like) – Positions to transform

  • inplace (bool (optional)) – Whether to modify the array inplace, overwriting previous data

Note

By default, the input x is modified in place and also returned. In-place operations improve performance because allocating new arrays is avoided.

Changed in version 0.7.5: Keyword inplace can be set to False so that a modified copy is returned unless no conversion takes place, in which case the reference to the unmodified x is returned.

convert_pos_to_native(x, inplace=True)

Conversion of coordinate array x from base units to native units.

Parameters:

  • x (array_like) – Positions to transform

  • inplace (bool (optional)) – Whether to modify the array inplace, overwriting previous data

Note

By default, the input x is modified in place and also returned. In-place operations improve performance because allocating new arrays is avoided.

Changed in version 0.7.5: Keyword inplace can be set to False so that a modified copy is returned unless no conversion takes place, in which case the reference to the unmodified x is returned.

convert_time_from_native(t, inplace=True)

Convert time t from native units to base units.

Parameters:

  • t (array_like) – Time values to transform

  • inplace (bool (optional)) – Whether to modify the array inplace, overwriting previous data

Note

By default, the input t is modified in place and also returned (although note that scalar values t are passed by value in Python and hence an in-place modification has no effect on the caller.) In-place operations improve performance because allocating new arrays is avoided.

Changed in version 0.7.5: Keyword inplace can be set to False so that a modified copy is returned unless no conversion takes place, in which case the reference to the unmodified x is returned.

convert_time_to_native(t, inplace=True)

Convert time t from base units to native units.

Parameters:

  • t (array_like) – Time values to transform

  • inplace (bool, optional) – Whether to modify the array inplace, overwriting previous data

Note

By default, the input t is modified in place and also returned. (Also note that scalar values t are passed by value in Python and hence an in-place modification has no effect on the caller.)

Changed in version 0.7.5: Keyword inplace can be set to False so that a modified copy is returned unless no conversion takes place, in which case the reference to the unmodified x is returned.

convert_velocities_from_native(v, inplace=True)

Conversion of velocities array v from native to base units

Parameters:

  • v (array_like) – Velocities to transform

  • inplace (bool (optional)) – Whether to modify the array inplace, overwriting previous data

Note

By default, the input v is modified in place and also returned. In-place operations improve performance because allocating new arrays is avoided.

New in version 0.7.5.

convert_velocities_to_native(v, inplace=True)

Conversion of coordinate array v from base to native units

Parameters:

  • v (array_like) – Velocities to transform

  • inplace (bool (optional)) – Whether to modify the array inplace, overwriting previous data

Note

By default, the input v is modified in place and also returned. In-place operations improve performance because allocating new arrays is avoided.

New in version 0.7.5.

copy()

Return independent copy of this Reader.

New Reader will have its own file handle and can seek/iterate independently of the original.

Will also copy the current state of the Timestep held in the original Reader.

Changed in version 2.2.0: Arguments used to construct the reader are correctly captured and passed to the creation of the new class. Previously the only n_atoms was passed to class copies, leading to a class created with default parameters which may differ from the original class.

property dt: float

Time between two trajectory frames in picoseconds.

property frame: int

Frame number of the current time step.

This is a simple short cut to Timestep.frame.

get_aux_attribute(auxname, attrname)

Get the value of attrname from the auxiliary auxname

Parameters:

  • auxname (str) – Name of the auxiliary to get value for

  • attrname (str) – Name of gettable attribute in the auxiliary reader

See also

set_aux_attribute()

get_aux_descriptions(auxnames=None)

Get descriptions to allow reloading the specified auxiliaries.

If no auxnames are provided, defaults to the full list of added auxiliaries.

Passing the resultant description to add_auxiliary() will allow recreation of the auxiliary. e.g., to duplicate all auxiliaries into a second trajectory:

descriptions = trajectory_1.get_aux_descriptions()
for aux in descriptions:
    trajectory_2.add_auxiliary(**aux)
Returns:

List of dictionaries of the args/kwargs describing each auxiliary.

Return type:

list

See also

MDAnalysis.auxiliary.base.AuxReader.get_description()

iter_as_aux(auxname)

Iterate through timesteps for which there is at least one assigned step from the auxiliary auxname within the cutoff specified in auxname.

See also

next_as_aux(), iter_auxiliary()

iter_auxiliary(auxname, start=None, stop=None, step=None, selected=None)

Iterate through the auxiliary auxname independently of the trajectory.

Will iterate over the specified steps of the auxiliary (defaults to all steps). Allows to access all values in an auxiliary, including those out of the time range of the trajectory, without having to also iterate through the trajectory.

After interation, the auxiliary will be repositioned at the current step.

Parameters:

  • auxname (str) – Name of the auxiliary to iterate over.

  • (start (optional) – Options for iterating over a slice of the auxiliary.

  • stop (optional) – Options for iterating over a slice of the auxiliary.

  • step) (optional) – Options for iterating over a slice of the auxiliary.

  • selected (lst | ndarray, optional) – List of steps to iterate over.

Yields:

AuxStep object

See also

iter_as_aux()

property n_frames

number of frames in trajectory

next() → Timestep

Forward one step to next frame.

next_as_aux(auxname)

Move to the next timestep for which there is at least one step from the auxiliary auxname within the cutoff specified in auxname.

This allows progression through the trajectory without encountering NaN representative values (unless these are specifically part of the auxiliary data).

If the auxiliary cutoff is not set, where auxiliary steps are less frequent (auxiliary.dt > trajectory.dt), this allows progression at the auxiliary pace (rounded to nearest timestep); while if the auxiliary steps are more frequent, this will work the same as calling next().

See the Auxiliary API.

See also

iter_as_aux()

classmethod parse_n_atoms(filename, **kwargs)[source]

Read the coordinate file and deduce the number of atoms

Returns:

n_atoms – the number of atoms in the coordinate file

Return type:

int

Raises:

NotImplementedError – when the number of atoms can’t be deduced

remove_auxiliary(auxname)

Clear data and close the AuxReader for the auxiliary auxname.

See also

add_auxiliary()

rename_aux(auxname, new)

Change the name of the auxiliary auxname to new.

Provided there is not already an auxiliary named new, the auxiliary name will be changed in ts.aux namespace, the trajectory’s list of added auxiliaries, and in the auxiliary reader itself.

Parameters:

  • auxname (str) – Name of the auxiliary to rename

  • new (str) – New name to try set

Raises:

ValueError – If the name new is already in use by an existing auxiliary.

rewind() → Timestep

Position at beginning of trajectory

set_aux_attribute(auxname, attrname, new)

Set the value of attrname in the auxiliary auxname.

Parameters:

  • auxname (str) – Name of the auxiliary to alter

  • attrname (str) – Name of settable attribute in the auxiliary reader

  • new – New value to try set attrname to

See also

get_aux_attribute(), rename_aux()

property time

Time of the current frame in MDAnalysis time units (typically ps).

This is either read straight from the Timestep, or calculated as time = Timestep.frame * Timestep.dt

timeseries(asel: AtomGroup | None = None, atomgroup: Atomgroup | None = None, start: int | None = None, stop: int | None = None, step: int | None = None, order: str | None = 'fac') → ndarray

Return a subset of coordinate data for an AtomGroup

Parameters:

  • asel (AtomGroup (optional)) –

    The AtomGroup to read the coordinates from. Defaults to None, in which case the full set of coordinate data is returned.

    Deprecated since version 2.7.0: asel argument will be renamed to atomgroup in 3.0.0

  • atomgroup (AtomGroup (optional)) – Same as asel, will replace asel in 3.0.0

  • start (int (optional)) – Begin reading the trajectory at frame index start (where 0 is the index of the first frame in the trajectory); the default None starts at the beginning.

  • stop (int (optional)) – End reading the trajectory at frame index stop-1, i.e, stop is excluded. The trajectory is read to the end with the default None.

  • step (int (optional)) – Step size for reading; the default None is equivalent to 1 and means to read every frame.

  • order (str (optional)) – the order/shape of the return data array, corresponding to (a)tom, (f)rame, (c)oordinates all six combinations of ‘a’, ‘f’, ‘c’ are allowed ie “fac” - return array where the shape is (frame, number of atoms, coordinates)

See also

MDAnalysis.coordinates.memory

New in version 2.4.0.

property totaltime: float

Total length of the trajectory

The time is calculated as (n_frames - 1) * dt, i.e., we assume that the first frame no time as elapsed. Thus, a trajectory with two frames will be considered to have a length of a single time step dt and a “trajectory” with a single frame will be reported as length 0.

property transformations

Returns the list of transformations

units = {'length': None, 'time': None, 'velocity': None}

dict with units of of time and length (and velocity, force, … for formats that support it)

7.37.2. Functions

MDAnalysis.coordinates.XDR.offsets_filename(filename, ending='npz')[source]

Return offset or its lock filename for XDR files. For this the filename is appended with _offsets.{ending}.

Parameters:

  • filename (str) – filename of trajectory

  • ending (str (optional)) – fileending of offsets file

Returns:

offset_filename

Return type:

str

MDAnalysis.coordinates.XDR.read_numpy_offsets(filename)[source]

read offsets into dictionary.

This assume offsets have been saved using numpy

Parameters:

filename (str) – filename of offsets

Returns:

offsets – dictionary of offsets information

Return type:

dict