python-mmtk 2.7.9-1 / usr / share / pyshared / MMTK / Trajectory.py

# This module implements trajetories and trajectory generators.
#
# Written by Konrad Hinsen
#

"""
Trajectory files and their contents
"""

__docformat__ = 'restructuredtext'

from MMTK import Collections, Units, Universe, Utility, \
                 ParticleProperties, Visualization
from Scientific.Geometry import Vector
from Scientific import N
import copy, os, sys

# Report error if the netCDF module is not available.
try:
    from Scientific.IO import NetCDF
except ImportError:
    raise Utility.MMTKError("Trajectories are not available " +
                             "because the netCDF module is missing.")
#
# Trajectory class
#
class Trajectory(object):

    #
    # Trajectory cache
    #
    # This cache is maintained for better efficiency in parallel
    # processing. The cache contains all trajectories currently open
    # for reading. When a trajectory object is unpickled, a trajectory
    # from the cache is reused if possible. This means that 


    """
    Trajectory file

    The data in a trajectory file can be accessed by step or by
    variable. If t is a Trajectory object, then:

     * len(t) is the number of steps
     * t[i] is the data for step i, in the form of a dictionary that
       maps variable names to data
     * t[i:j] and t[i:j:n] return a :class:`~MMTK.Trajectory.SubTrajectory` 
       object that refers to a subset of the total number of steps 
       (no data is copied)
     * t.variable returns the value of the named variable at all
       time steps. If the variable is a simple scalar, it is read
       completely and returned as an array. If the variable contains
       data for each atom, a :class:`~MMTK.Trajectory.TrajectoryVariable` 
       object is returned from which data at specific steps can be obtained 
       by further indexing operations.

    The routines that generate trajectories decide what variables
    are used and what they contain. The most frequently used variable
    is "configuration", which stores the positions of all atoms.
    Other common variables are "time", "velocities", "temperature",
    "pressure", and various energy terms whose name end with "_energy".
    """

    def __init__(self, object, filename, mode = 'r', comment = None,
                 double_precision = False, cycle = 0, block_size = 1):
        """
        :param object: the object whose data is stored in the trajectory file.
                       This can be 'None' when opening a file for reading;
                       in that case, a universe object is constructed from the
                       description stored in the trajectory file. This universe
                       object can be accessed via the attribute 'universe'
                       of the trajectory object.
        :type object: :class:`~MMTK.ChemicalObjects.ChemicalObject`
        :param filename: the name of the trajectory file
        :type filename: str
        :param mode: one of "r" (read-only), "w" (create new file for writing),
                     or "a" (append to existing file or create if the file does
                     not exist)
        :type mode: str
        :param comment: optional comment that is stored in the file;
                        allowed only with mode="r"
        :type comment: str
        :param double_precision: if True, data in the file is stored using
                                 double precision; default is single precision.
                                 Note that all I/O via trajectory objects is
                                 double precision; conversion from and to
                                 single precision file variables is handled
                                 automatically.
        :type double_precision: bool
        :param cycle: if non-zero, a trajectory is created for a fixed number
                      of steps equal to the value of cycle, and these steps
                      are used cyclically. This is meant for restart
                      trajectories.
        :type cycle: int
        :param block_size: an optimization parameter that influences the file
                           structure and the I/O performance for very large
                           files. A block size of 1 is optimal for sequential
                           access to configurations etc., whereas a block size
                           equal to the number of steps is optimal for reading
                           coordinates or scalar variables along the time axis.
                           The default value is 1. Note that older MMTK releases
                           always used a block size of 1 and cannot handle
                           trajectories with different block sizes.
        :type block_size: int
        """
        filename = os.path.expanduser(filename)
        self.filename = filename
        self.mode = mode
        if object is None and mode == 'r':
            file = NetCDF.NetCDFFile(filename, 'r')
            description = file.variables['description'][:].tostring()
            try:
                self.block_size = file.dimensions['minor_step_number']
            except KeyError:
                self.block_size = 1
            conf = None
            cell = None
            if self.block_size == 1:
                try:
                    conf_var = file.variables['configuration']
                    conf = conf_var[0, :, :]
                except KeyError: pass
                try:
                    cell = file.variables['box_size'][0, :]
                except KeyError: pass
            else:
                try:
                    conf_var = file.variables['configuration']
                    conf = conf_var[0, :, :, 0]
                except KeyError: pass
                try:
                    cell = file.variables['box_size'][0, :, 0]
                except KeyError: pass
            file.close()
            import Skeleton
            local = {}
            skeleton = eval(description, vars(Skeleton), local)
            universe = skeleton.make({}, conf)
            universe.setCellParameters(cell)
            object = universe
            initialize = 1
        else:
            universe = object.universe()
            if universe is None:
                raise ValueError("objects not in the same universe")
            description = None
            initialize = 0
        universe.configuration()
        if object is universe:
            index_map = None
            inverse_map = None
        else:
            if mode == 'r':
                raise ValueError("can't read trajectory for a non-universe")
            index_map = N.array([a.index for a in  object.atomList()])
            inverse_map = universe.numberOfPoints()*[None]
            for i in range(len(index_map)):
                inverse_map[index_map[i]] = i
            toplevel = set()
            for o in Collections.Collection(object):
                toplevel.add(o.topLevelChemicalObject())
            object = Collections.Collection(list(toplevel))
        if description is None:
            description = universe.description(object, inverse_map)
        import MMTK_trajectory
        self.trajectory = MMTK_trajectory.Trajectory(universe, description,
                                                     index_map, filename,
                                                     mode + 's',
                                                     double_precision, cycle,
                                                     block_size)
        self.universe = universe
        self.index_map = index_map
        try:
            self.block_size = \
                       self.trajectory.file.dimensions['minor_step_number']
        except KeyError:
            self.block_size = 1
        if comment is not None:
            if mode == 'r':
                raise IOError('cannot add comment in read-only mode')
            self.trajectory.file.comment = comment
        if initialize and conf is not None:
            self.universe.setFromTrajectory(self)
        self.particle_trajectory_reader = ParticleTrajectoryReader(self)

    def __getstate__(self):
        if self.mode != 'r':
            raise ValueError("Cannot copy or pickle write-mode trajectories")
        return self.filename

    def __setstate__(self, state):
        self.__init__(None, state)

    def flush(self):
        """
        Make sure that all data that has been written to the trajectory
        is also written to the file.
        """
        self.trajectory.flush()

    def close(self):
        """
        Close the trajectory file. Must be called after writing to
        ensure that all buffered data is written to the file. No data
        access is possible after closing a file.
        """
        self.trajectory.close()

    def __len__(self):
        return self.trajectory.nsteps

    def __getitem__(self, item):
        if not isinstance(item, int):
            return SubTrajectory(self, N.arange(len(self)))[item]
        if item < 0:
            item += len(self)
        if item >= len(self):
            raise IndexError
        data = {}
        for name, var in self.trajectory.file.variables.items():
            if 'step_number' not in var.dimensions:
                continue
            if 'atom_number' in var.dimensions:
                if 'xyz' in var.dimensions:
                    array = ParticleProperties.ParticleVector(self.universe,
                                self.trajectory.readParticleVector(name, item))
                else:
                    array = ParticleProperties.ParticleScalar(self.universe,
                                self.trajectory.readParticleScalar(name, item))
            else:
                bs = self.block_size
                if bs == 1:
                    array = var[item]
                else:
                    if len(var.shape) == 2:
                        array = var[item/bs, item%bs]
                    else:
                        array = var[item/bs, ..., item%bs]
            data[name] = 0.+array
        if data.has_key('configuration'):
            box = data.get('box_size', None)
            if box is not None:
                box = box.astype(N.Float)
            conf = data['configuration']
            data['configuration'] = \
               ParticleProperties.Configuration(conf.universe, conf.array, box)
        return data

    def __getslice__(self, first, last):
        return self[(slice(first, last),)]

    def __getattr__(self, name):
        try:
            var = self.trajectory.file.variables[name]
        except KeyError:
            raise AttributeError("no variable named " + name)
        if 'atom_number' in var.dimensions:
            return TrajectoryVariable(self.universe, self, name)
        elif 'box_size_length' in var.dimensions:
            if 'minor_step_number' in var.dimensions:
                bs = N.transpose(var[:], [0, 2, 1])
                bs = N.reshape(bs, (bs.shape[0]*bs.shape[1], bs.shape[2]))
                return bs[:len(self)]
            else:
                return var[:]
        else:
            return N.ravel(N.array(var))[:len(self)]

    def defaultStep(self):
        try:
            step = int(self.trajectory.file.last_step[0])
        except AttributeError:
            step = 0
        return step

    def readParticleTrajectory(self, atom, first=0, last=None, skip=1,
                               variable = "configuration"):
        """
        Read trajectory information for a single atom but for multiple
        time steps.

        :param atom: the atom whose trajectory is requested
        :type atom: :class:`~MMTK.ChemicalObjects.Atom`
        :param first: the number of the first step to be read
        :type first: int
        :param last: the number of the first step not to be read.
                     A value of None indicates that the
                     whole trajectory should be read.
        :type last: int
        :param skip: the number of steps to skip between two steps read
        :type skip: int
        :param variable: the name of the trajectory variable to be read.
                         If the variable is "configuration", the resulting
                         trajectory is made continuous by eliminating all
                         jumps caused by periodic boundary conditions.
                         The pseudo-variable "box_coordinates" can be read
                         to obtain the values of the variable "configuration"
                         scaled to box coordinates. For non-periodic universes
                         there is no difference between box coordinates
                         and real coordinates.
        :type variable: str
        :returns: the trajectory for a single atom
        :rtype: :class:`~MMTK.Trajectory.ParticleTrajectory`
        """
        return ParticleTrajectory(self, atom, first, last, skip, variable)

    def readRigidBodyTrajectory(self, object, first=0, last=None, skip=1,
                                reference = None):
        """
        Read the positions for an object at multiple time steps
        and extract the rigid-body motion (center-of-mass position plus
        orientation as a quaternion) by an optimal-transformation fit.

        :param object: the object whose rigid-body trajectory is requested
        :type object: :class:`~MMTK.Collections.GroupOfAtoms`
        :param first: the number of the first step to be read
        :type first: int
        :param last: the number of the first step not to be read.
                     A value of None indicates that the
                     whole trajectory should be read.
        :type last: int
        :param skip: the number of steps to skip between two steps read
        :type skip: int
        :param reference: the reference configuration for the fit
        :type reference: :class:`~MMTK.ParticleProperties.Configuration`
        :returns: the trajectory for a single rigid body
        :rtype: :class:`~MMTK.Trajectory.RigidBodyTrajectory`
        """
        return RigidBodyTrajectory(self, object, first, last, skip, reference)

    def variables(self):
        """
        :returns: a list of the names of all variables that are stored
                  in the trajectory
        :rtype: list of str
        """
        vars = copy.copy(self.trajectory.file.variables.keys())
        vars.remove('step')
        try:
            vars.remove('description')
        except ValueError: pass
        return vars

    def view(self, first=0, last=None, skip=1, object = None):
        """
        Show an animation of the trajectory using an external visualization
        program.

        :param first: the number of the first step in the animation
        :type first: int
        :param last: the number of the first step not to include in the
                     animation. A value of None indicates that the
                     whole trajectory should be used.
        :type last: int
        :param skip: the number of steps to skip between two steps read
        :type skip: int
        :param object: the object to be animated, which must be in the
                       universe stored in the trajectory. None
                       stands for the whole universe.
        :type object: :class:`~MMTK.Collections.GroupOfAtoms`
        """
        Visualization.viewTrajectory(self, first, last, skip, object)

    def _boxTransformation(self, pt_in, pt_out, to_box=0):
        from MMTK_trajectory import boxTransformation
        try:
            box_size = self.trajectory.recently_read_box_size
        except AttributeError:
            return
        boxTransformation(self.universe._spec,
                          pt_in, pt_out, box_size, to_box)


class SubTrajectory(object):

    """
    Reference to a subset of a trajectory

    A SubTrajectory object is created by slicing a Trajectory object
    or another SubTrajectory object. It provides all the operations
    defined on Trajectory objects.
    """

    def __init__(self, trajectory, indices):
        self.trajectory = trajectory
        self.indices = indices
        self.universe = trajectory.universe

    def __len__(self):
        return len(self.indices)

    def __getitem__(self, item):
        if isinstance(item, int):
            return self.trajectory[int(self.indices[item])]
        else:
            return SubTrajectory(self.trajectory, self.indices[item])

    def __getslice__(self, first, last):
        return self[(slice(first, last),)]

    def __getattr__(self, name):
        return SubVariable(getattr(self.trajectory, name), self.indices)

    def readParticleTrajectory(self, atom, first=0, last=None, skip=1,
                               variable = "configuration"):
        if last is None:
            last = len(self.indices)
        indices = self.indices[first:last:skip]
        first = indices[0]
        last = indices[-1]+1
        if len(self.indices) > 1:
            skip = self.indices[1]-self.indices[0]
        else:
            skip = 1
        return self.trajectory.readParticleTrajectory(atom, first, last,
                                                      skip, variable)

    def readRigidBodyTrajectory(self, object, first=0, last=None, skip=1,
                                reference = None):
        if last is None:
            last = len(self.indices)
        indices = self.indices[first:last:skip]
        first = indices[0]
        last = indices[-1]+1
        if len(self.indices) > 1:
            skip = self.indices[1]-self.indices[0]
        else:
            skip = 1
        return RigidBodyTrajectory(self.trajectory, object,
                                   first, last, skip, reference)

    def variables(self):
        return self.trajectory.variables()

    def view(self, first=0, last=None, step=1, subset = None):
        Visualization.viewTrajectory(self, first, last, step, subset)

    def close(self):
        del self.trajectory

    def _boxTransformation(self, pt_in, pt_out, to_box=0):
        Trajectory._boxTransformation(self.trajectory, pt_in, pt_out, to_box)

#
# Trajectory variables
#
class TrajectoryVariable(object):

    """
    Variable in a trajectory

    A TrajectoryVariable object is created by extracting a variable from
    a Trajectory object if that variable contains data for each atom and
    is thus potentially large. No data is read from the trajectory file
    when a TrajectoryVariable object is created; the read operation
    takes place when the TrajectoryVariable is indexed with a specific
    step number.

    If t is a TrajectoryVariable object, then:

     * len(t) is the number of steps
     * t[i] is the data for step i, in the form of a ParticleScalar,
       a ParticleVector, or a Configuration object, depending on the
       variable
     * t[i:j] and t[i:j:n] return a SubVariable object that refers
       to a subset of the total number of steps
    """
    
    def __init__(self, universe, trajectory, name):
        self.universe = universe
        self.trajectory = trajectory
        self.name = name
        self.var = self.trajectory.trajectory.file.variables[self.name]
        if self.name == 'configuration':
            try:
                self.box_size = \
                        self.trajectory.trajectory.file.variables['box_size']
            except KeyError:
                self.box_size = None

    def __len__(self):
        return len(self.trajectory)

    def __getitem__(self, item):
        if not isinstance(item, int):
            return SubVariable(self, N.arange(len(self)))[item]
        item = int(item) # gets rid of numpy.intXX objects
        if item < 0:
            item = item + len(self.trajectory)
        if item >= len(self.trajectory):
            raise IndexError
        if self.name == 'configuration':
            if self.box_size is None:
                box = None
            elif len(self.box_size.shape) == 3:
                bs = self.trajectory.block_size
                box = self.box_size[item/bs, :, item%bs].astype(N.Float)
            else:
                box = self.box_size[item].astype(N.Float)
            array = ParticleProperties.Configuration(self.universe,
                self.trajectory.trajectory.readParticleVector(self.name, item),
                box)
        elif 'xyz' in self.var.dimensions:
            array = ParticleProperties.ParticleVector(self.universe,
                self.trajectory.trajectory.readParticleVector(self.name, item))
        else:
            array = ParticleProperties.ParticleScalar(self.universe,
                self.trajectory.trajectory.readParticleScalar(self.name, item))
        return array

    def __getslice__(self, first, last):
        return self[(slice(first, last),)]

    def average(self):
        sum = self[0]
        for value in self[1:]:
            sum = sum + value
        return sum/len(self)

class SubVariable(TrajectoryVariable):

    """
    Reference to a subset of a :class:`~MMTK.Trajectory.TrajectoryVariable`

    A SubVariable object is created by slicing a TrajectoryVariable
    object or another SubVariable object. It provides all the operations
    defined on TrajectoryVariable objects.
    """

    def __init__(self, variable, indices):
        self.variable = variable
        self.indices = indices

    def __len__(self):
        return len(self.indices)

    def __getitem__(self, item):
        if isinstance(item, int):
            return self.variable[self.indices[item]]
        else:
            return SubVariable(self.variable, self.indices[item])

    def __getslice__(self, first, last):
        return self[(slice(first, last),)]

#
# Trajectory consisting of multiple files
#
class TrajectorySet(object):

    """
    Trajectory file set

    A TrajectorySet permits to treat a sequence of trajectory files
    like a single trajectory for reading data. It behaves exactly like a
    :class:`~MMTK.Trajectory.Trajectory` object. The trajectory files must all contain data
    for the same system. The variables stored in the individual files
    need not be the same, but only variables common to all files
    can be accessed.

    Note: depending on how the sequence of trajectories was constructed,
    the first configuration of each trajectory might be the same as the
    last one in the preceding trajectory. To avoid counting it twice,
    specify (filename, 1, None, 1) for all but the first trajectory in
    the set.
    """

    def __init__(self, object, filenames):
        """
        :param object: the object whose data is stored in the trajectory files.
                       This can be (and usually is) None;
                       in that case, a universe object is constructed from the
                       description stored in the first trajectory file.
                       This universe object can be accessed via the attribute
                       universe of the trajectory set object.
        :param filenames: a list of trajectory file names or
                          (filename, first_step, last_step, increment)
                          tuples.
        """
        first = filenames[0]
        if isinstance(first, tuple):
            first = Trajectory(object, first[0])[first[1]:first[2]:first[3]]
        else:
            first = Trajectory(object, first)
        self.universe = first.universe
        self.trajectories = [first]
        self.nsteps = [0, len(first)]
        self.cell_parameters = []
        for file in filenames[1:]:
            if isinstance(file, tuple):
                t = Trajectory(self.universe, file[0])[file[1]:file[2]:file[3]]
            else:
                t = Trajectory(self.universe, file)
            self.trajectories.append(t)
            self.nsteps.append(self.nsteps[-1]+len(t))
            try:
                self.cell_parameters.append(t[0]['box_size'])
            except KeyError:
                pass
        vars = {}
        for t in self.trajectories:
            for v in t.variables():
                vars[v] = vars.get(v, 0) + 1
        self.vars = []
        for v, count in vars.items():
            if count == len(self.trajectories):
                self.vars.append(v)

    def close(self):
        for t in self.trajectories:
            t.close()

    def __len__(self):
        return self.nsteps[-1]

    def __getitem__(self, item):
        if not isinstance(item, int):
            return SubTrajectory(self, N.arange(len(self)))[item]
        if item >= len(self):
            raise IndexError
        tindex = N.add.reduce(N.greater_equal(item, self.nsteps))-1
        return self.trajectories[tindex][item-self.nsteps[tindex]]

    def __getslice__(self, first, last):
        return self[(slice(first, last),)]

    def __getattr__(self, name):
        if name not in self.vars+['step']:
            raise AttributeError("no variable named " + name)
        var = self.trajectories[0].trajectory.file.variables[name]
        if 'atom_number' in var.dimensions:
            return TrajectorySetVariable(self.universe, self, name)
        else:
            data = []
            for t in self.trajectories:
                var = t.trajectory.file.variables[name]
                data.append(N.ravel(N.array(var))[:len(t)])
            return N.concatenate(data)

    def readParticleTrajectory(self, atom, first=0, last=None, skip=1,
                               variable = "configuration"):
        total = None
        self.steps_read = []
        for i in range(len(self.trajectories)):
            if self.nsteps[i+1] <= first:
                self.steps_read.append(0)
                continue
            if last is not None and self.nsteps[i] >= last:
                break
            n = max(0, (self.nsteps[i]-first+skip-1)/skip)
            start = first+skip*n-self.nsteps[i]
            n = (self.nsteps[i+1]-first+skip-1)/skip
            stop = first+skip*n
            if last is not None:
                stop = min(stop, last)
            stop = stop-self.nsteps[i]
            if start >= 0 and start < self.nsteps[i+1]-self.nsteps[i]:
                t = self.trajectories[i]
                pt = t.readParticleTrajectory(atom, start, stop, skip,
                                              variable)
                self.steps_read.append((stop-start)/skip)
                if total is None:
                    total = pt
                else:
                    if variable == "configuration" \
                       and self.cell_parameters[0] is not None:
                        jump = pt.array[0]-total.array[-1]
                        mult = -(jump/self.cell_parameters[i-1]).astype('i')
                        if len(N.nonzero(mult)) > 0:
                            t._boxTransformation(pt.array, pt.array, 1)
                            N.add(pt.array, mult[N.NewAxis, : ],
                                        pt.array)
                            t._boxTransformation(pt.array, pt.array, 0)
                            jump = pt.array[0] - total.array[-1]
                        mask = N.less(jump,
                                            -0.5*self.cell_parameters[i-1])- \
                               N.greater(jump,
                                               0.5*self.cell_parameters[i-1])
                        if len(N.nonzero(mask)) > 0:
                            t._boxTransformation(pt.array, pt.array, 1)
                            N.add(pt.array, mask[N.NewAxis, :],
                                        pt.array)
                            t._boxTransformation(pt.array, pt.array, 0)
                    elif variable == "box_coordinates" \
                       and self.cell_parameters[0] is not None:
                        jump = pt.array[0]-total.array[-1]
                        mult = -jump.astype('i')
                        if len(N.nonzero(mult)) > 0:
                            N.add(pt.array, mult[N.NewAxis, : ],
                                        pt.array)
                            jump = pt.array[0] - total.array[-1]
                        mask = N.less(jump, -0.5)- \
                               N.greater(jump, 0.5)
                        if len(N.nonzero(mask)) > 0:
                            N.add(pt.array, mask[N.NewAxis, :],
                                        pt.array)
                    total.array = N.concatenate((total.array, pt.array))
            else:
                self.steps_read.append(0)
        return total

    def readRigidBodyTrajectory(self, object, first=0, last=None, skip=1,
                                reference = None):
        return RigidBodyTrajectory(self, object, first, last, skip, reference)

    def _boxTransformation(self, pt_in, pt_out, to_box=0):
        n = 0
        for i in range(len(self.steps_read)):
            t = self.trajectories[i]
            steps = self.steps_read[i]
            if steps > 0:
                t._boxTransformation(pt_in[n:n+steps], pt_out[n:n+steps],
                                     to_box)
            n = n + steps

    def variables(self):
        return self.vars

    def view(self, first=0, last=None, step=1, object = None):
        Visualization.viewTrajectory(self, first, last, step, object)


class TrajectorySetVariable(TrajectoryVariable):

    """
    Variable in a trajectory set

    A TrajectorySetVariable object is created by extracting a variable from
    a TrajectorySet object if that variable contains data for each atom and
    is thus potentially large. It behaves exactly like a TrajectoryVariable
    object.
    """
    
    def __init__(self, universe, trajectory_set, name):
        self.universe = universe
        self.trajectory_set = trajectory_set
        self.name = name

    def __len__(self):
        return len(self.trajectory_set)

    def __getitem__(self, item):
        if not isinstance(item, int):
            return SubVariable(self, N.arange(len(self)))[item]
        if item >= len(self.trajectory_set):
            raise IndexError
        tindex = N.add.reduce(N.greater_equal(item,
                                              self.trajectory_set.nsteps))-1
        step = item-self.trajectory_set.nsteps[tindex]
        t = self.trajectory_set.trajectories[tindex]
        return getattr(t, self.name)[step]

#
# Cache for atom trajectories
#
class ParticleTrajectoryReader(object):

    def __init__(self, trajectory):
        self.trajectory = trajectory
        self.natoms = self.trajectory.universe.numberOfAtoms()
        self._trajectory = trajectory.trajectory
        self.cache = {}
        self.cache_lifetime = 2

    def __call__(self, atom, variable, first, last, skip, correct, box):
        if isinstance(atom, int):
            index = atom
        else:
            index = atom.index
            if atom.universe() is not self.trajectory.universe:
                raise ValueError("objects not in the same universe")
        key = (index, variable, first, last, skip, correct, box)
        data, count = self.cache.get(key, (None, 0))
        if data is not None:
            self.cache[key] = (data, self.cache_lifetime)
            return data
        delete = []
        for k, value in self.cache.items():
            data, count = value
            count -= 1
            if count == 0:
                delete.append(k)
            else:
                self.cache[k] = (data, count)
        for k in delete:
            del self.cache[k]
        cache_size = min(10, max(1, 100000/max(1, len(self.trajectory))))
        natoms = min(cache_size, self.natoms-index)
        data = self._trajectory.readParticleTrajectories(index, natoms,
                                                         variable,
                                                         first, last, skip,
                                                         correct, box)
        for i in range(natoms):
            key = (index+i, variable, first, last, skip, correct, box)
            self.cache[key] = (data[i], self.cache_lifetime)
        return data[0]

#
# Single-atom trajectory
#
class ParticleTrajectory(object):

    """
    Trajectory data for a single particle

    A ParticleTrajectory object is created by calling the method
    :func:`~MMTK.Trajectory.Trajectory.readParticleTrajectory`
    on a :class:`~MMTK.Trajectory.Trajectory` object.

    If pt is a ParticleTrajectory object, then

     * len(pt) is the number of steps stored in it
     * pt[i] is the value at step i (a vector)
    """
    
    def __init__(self, trajectory, atom, first=0, last=None, skip=1,
                 variable = "configuration"):
        if last is None:
            last = len(trajectory)
        if variable == "box_coordinates":
            variable = "configuration"
            box = 1
        else:
            box = 0
        reader = trajectory.particle_trajectory_reader
        self.array = reader(atom, variable, first, last, skip,
                            variable == "configuration", box)

    def __len__(self):
        return self.array.shape[0]

    def __getitem__(self, index):
        return Vector(self.array[index])

    def translateBy(self, vector):
        """
        Adds a vector to the values at all steps. This does B{not}
        change the data in the trajectory file.

        :param vector: the vector to be added
        :type vector: Scientific.Geometry.Vector
        """
        N.add(self.array, vector.array[N.NewAxis, :], self.array)

#
# Rigid-body trajectory
#
class RigidBodyTrajectory(object):

    """
    Rigid-body trajectory data

    A RigidBodyTrajectory object is created by calling the method
    :func:`~MMTK.Trajectory.Trajectory.readRigidBodyTrajectory`
    on a :class:`~MMTK.Trajectory.Trajectory` object.

    If rbt is a RigidBodyTrajectory object, then

     * len(rbt) is the number of steps stored in it
     * rbt[i] is the value at step i (a vector for the center of mass
       and a quaternion for the orientation)
    """
    
    def __init__(self, trajectory, object, first=0, last=None, skip=1,
                 reference = None):
        self.trajectory = trajectory
        universe = trajectory.universe
        if last is None: last = len(trajectory)
        first_conf = trajectory.configuration[first]
        offset = universe.contiguousObjectOffset([object], first_conf, True)
        if reference is None:
            reference = first_conf
        reference = universe.contiguousObjectConfiguration([object], reference)
        steps = (last-first+skip-1)/skip
        mass = object.mass()
        ref_cms = object.centerOfMass(reference)
        atoms = object.atomList()

        possq = N.zeros((steps,), N.Float)
        cross = N.zeros((steps, 3, 3), N.Float)
        rcms = N.zeros((steps, 3), N.Float)

        # cms of the CONTIGUOUS object made of CONTINUOUS atom trajectories 
        for a in atoms:
            r = trajectory.readParticleTrajectory(a, first, last, skip,
                                                  "box_coordinates").array
            w = a._mass/mass
            N.add(rcms, w*r, rcms)
            if offset is not None:
                N.add(rcms, w*offset[a].array, rcms)
        
        # relative coords of the CONTIGUOUS reference
        r_ref = N.zeros((len(atoms), 3), N.Float)
        for a in range(len(atoms)):
            r_ref[a] = atoms[a].position(reference).array - ref_cms.array

        # main loop: storing data needed to fill M matrix 
        for a in range(len(atoms)):
            r = trajectory.readParticleTrajectory(atoms[a],
                                                  first, last, skip,
                                                  "box_coordinates").array
            r = r - rcms # (a-b)**2 != a**2 - b**2
            if offset is not None:
                N.add(r, offset[atoms[a]].array,r)
            trajectory._boxTransformation(r, r)
            w = atoms[a]._mass/mass
            N.add(possq, w*N.add.reduce(r*r, -1), possq)
            N.add(possq, w*N.add.reduce(r_ref[a]*r_ref[a],-1),
                        possq)
            N.add(cross, w*r[:,:,N.NewAxis]*r_ref[N.NewAxis,
                                                              a,:],cross)
        self.trajectory._boxTransformation(rcms, rcms)

        # filling matrix M (formula no 40)
        k = N.zeros((steps, 4, 4), N.Float)
        k[:, 0, 0] = -cross[:, 0, 0]-cross[:, 1, 1]-cross[:, 2, 2]
        k[:, 0, 1] = cross[:, 1, 2]-cross[:, 2, 1]
        k[:, 0, 2] = cross[:, 2, 0]-cross[:, 0, 2]
        k[:, 0, 3] = cross[:, 0, 1]-cross[:, 1, 0]
        k[:, 1, 1] = -cross[:, 0, 0]+cross[:, 1, 1]+cross[:, 2, 2]
        k[:, 1, 2] = -cross[:, 0, 1]-cross[:, 1, 0]
        k[:, 1, 3] = -cross[:, 0, 2]-cross[:, 2, 0]
        k[:, 2, 2] = cross[:, 0, 0]-cross[:, 1, 1]+cross[:, 2, 2]
        k[:, 2, 3] = -cross[:, 1, 2]-cross[:, 2, 1]
        k[:, 3, 3] = cross[:, 0, 0]+cross[:, 1, 1]-cross[:, 2, 2]
        del cross
        for i in range(1, 4):
            for j in range(i):
                k[:, i, j] = k[:, j, i]
        N.multiply(k, 2., k)
        for i in range(4):
            N.add(k[:,i,i], possq, k[:,i,i])
        del possq

        quaternions = N.zeros((steps, 4), N.Float)
        fit = N.zeros((steps,), N.Float)
        from Scientific.LA import eigenvectors
        for i in range(steps):
            e, v = eigenvectors(k[i])
            j = N.argmin(e)
            if e[j] < 0.:
                fit[i] = 0.
            else:
                fit[i] = N.sqrt(e[j])
            if v[j,0] < 0.: quaternions[i] = -v[j] # eliminate jumps
            else: quaternions[i] = v[j]
        self.fit = fit
        self.cms = rcms
        self.quaternions = quaternions

    def __len__(self):
        return self.cms.shape[0]

    def __getitem__(self, index):
        from Scientific.Geometry.Quaternion import Quaternion
        return Vector(self.cms[index]), Quaternion(self.quaternions[index])

#
# Type check for trajectory objects
#
def isTrajectory(object):
    """
    :param object: any Python object
    :returns: True if object is a trajectory
    """
    import MMTK_trajectory
    return isinstance(object, (Trajectory, MMTK_trajectory.trajectory_type))

#
# Base class for all objects that generate trajectories
#
class TrajectoryGenerator(object):

    """
    Trajectory generator base class

    This base class implements the common aspects of everything that
    generates trajectories: integrators, minimizers, etc.
    """

    def __init__(self, universe, options):
        self.universe = universe
        self.options = options

    def setCallOptions(self, options):
        self.call_options = options

    def getActions(self):
        try:
            self.actions = self.getOption('actions')
        except ValueError:
            self.actions = []
        try:
            if self.getOption('background'):
                import MMTK_state_accessor
                self.state_accessor = MMTK_state_accessor.StateAccessor()
                self.actions.append(self.state_accessor)
        except ValueError:
            pass
        try:
            steps = self.getOption('steps')
        except ValueError:
            steps = None
        return map(lambda a, t=self, s=steps: a.getSpecificationList(t, s),
                   self.actions)

    def cleanupActions(self):
        for a in self.actions:
            a.cleanup()

    def getOption(self, option):
        try:
            value = self.call_options[option]
        except KeyError:
            try:
                value = self.options[option]
            except KeyError:
                try:
                    value = self.default_options[option]
                except KeyError:
                    raise ValueError('undefined option: ' + option)
        return value

    def optionString(self, options):
        s = ''
        for o in options:
            s = s + o + '=' + `self.getOption(o)` + ', '
        return s[:-2]

    def run(self, function, args):
        if self.getOption('background'):
            import ThreadManager
            return ThreadManager.TrajectoryGeneratorThread(self.universe,
                                      function, args, self.state_accessor)
        else:
            apply(function, args)
        
#
# Trajectory action base class
#
class TrajectoryAction(object):

    """
    Trajectory action base class

    Subclasses of this base class implement the actions that can be
    inserted into trajectory generation at regular intervals.
    """

    def __init__(self, first, last, skip):
        self.first = first
        self.last = last
        self.skip = skip

    spec_type = 'function'

    def _getSpecificationList(self, trajectory_generator, steps):
        first = self.first
        last = self.last
        if first < 0:
            first = first + steps
        if last is None:
            import MMTK_trajectory
            last = MMTK_trajectory.maxint
        elif last < 0:
            last = last + steps+1
        return (self.spec_type, first, last, self.skip)

    def getSpecificationList(self, trajectory_generator, steps):
        return self._getSpecificationList(trajectory_generator, steps) \
               + (self.Cfunction, self.parameters)

    def cleanup(self):
        pass

class TrajectoryOutput(TrajectoryAction):

    """
    Trajectory output action

    A TrajectoryOutput object can be used in the action list of any
    trajectory-generating operation. It writes any of the available
    data to a trajectory file. It is possible to use several
    TrajectoryOutput objects at the same time in order to produce
    multiple trajectories from a single run.
    """

    def __init__(self, trajectory, data = None,
                 first=0, last=None, skip=1):
        """
        :param trajectory: a trajectory object or a string, which is
                           interpreted as the name of a file that is opened
                           as a trajectory in append mode
        :param data: a list of data categories. All variables provided by the
                     trajectory generator that fall in any of the listed
                     categories are written to the trajectory file. See the
                     descriptions of the trajectory generators for a list
                     of variables and categories. By default (data = None)
                     the categories "configuration", "energy",
                     "thermodynamic", and "time" are written.
        :param first: the number of the first step at which the action is run
        :type first: int
        :param last: the number of the step at which the action is suspended.
                     A value of None indicates that the action should
                     be applied indefinitely.
        :type last: int
        :param skip: the number of steps to skip between two action runs
        :type skip: int
        """
        TrajectoryAction.__init__(self, first, last, skip)
        self.destination = trajectory
        self.categories = data
        self.must_be_closed = None

    spec_type = 'trajectory'

    def getSpecificationList(self, trajectory_generator, steps):
        if type(self.destination) == type(''):
            destination = self._setupDestination(self.destination,
                                                 trajectory_generator.universe)
        else:
            destination = self.destination
        if self.categories is None:
            categories = self._defaultCategories(trajectory_generator)
        else:
            if self.categories == 'all' or self.categories == ['all']:
                categories = trajectory_generator.available_data
            else:
                categories = self.categories
                for item in categories:
                    if item not in trajectory_generator.available_data:
                        raise ValueError('data item %s is not available' % item)
        return self._getSpecificationList(trajectory_generator, steps) \
               + (destination, categories)

    def _setupDestination(self, destination, universe):
        self.must_be_closed = Trajectory(universe, destination, 'a')
        return self.must_be_closed
        
    def cleanup(self):
        if self.must_be_closed is not None:
            self.must_be_closed.close()

    def _defaultCategories(self, trajectory_generator):
        available = trajectory_generator.available_data
        return tuple(filter(lambda x, a=available: x in a, self.default_data))

    default_data = ['configuration', 'energy', 'thermodynamic', 'time']

class RestartTrajectoryOutput(TrajectoryOutput):

    """
    Restart trajectory output action

    A RestartTrajectoryOutput object is used in the action list of any
    trajectory-generating operation. It writes those variables to a
    trajectory that the trajectory generator declares as necessary
    for restarting.
    """

    def __init__(self, trajectory, skip=100, length=3):
        """
        :param trajectory: a trajectory object or a string, which is interpreted
                           as the name of a file that is opened as a trajectory
                           in append mode with a cycle length of length and
                           double-precision variables
        :param skip: the number of steps between two write operations to the
                     restart trajectory
        :type skip: int
        :param length: the number of steps stored in the restart trajectory;
                       used only if trajectory is a string
        """
        TrajectoryAction.__init__(self, 0, None, skip)
        self.destination = trajectory
        self.categories = None
        self.length = length

    def _setupDestination(self, destination, universe):
        self.must_be_closed = Trajectory(universe, destination, 'a',
                                         'Restart trajectory', 1, self.length)
        return self.must_be_closed
        
    def _defaultCategories(self, trajectory_generator):
        if trajectory_generator.restart_data is None:
            raise ValueError("Trajectory generator does not permit restart")
        return trajectory_generator.restart_data

class LogOutput(TrajectoryOutput):

    """
    Protocol file output action

    A LogOutput object can be used in the action list of any
    trajectory-generating operation. It writes any of the available
    data to a text file.
    """

    def __init__(self, file, data = None, first=0, last=None, skip=1):
        """
        :param file: a file object or a string, which is interpreted as the
                     name of a file that is opened in write mode
        :param data: a list of data categories. All variables provided by the
                     trajectory generator that fall in any of the listed
                     categories are written to the trajectory file. See the
                     descriptions of the trajectory generators for a list
                     of variables and categories. By default (data = None)
                     the categories "configuration", "energy",
                     "thermodynamic", and "time" are written.
        :param first: the number of the first step at which the action is run
        :type first: int
        :param last: the number of the step at which the action is suspended.
                     A value of None indicates that the action should
                     be applied indefinitely.
        :type last: int
        :param skip: the number of steps to skip between two action runs
        :type skip: int
        """
        TrajectoryOutput.__init__(self, file, data, first, last, skip)

    def _setupDestination(self, destination, universe):
        self.must_be_closed = open(destination, 'w')
        return self.must_be_closed

    spec_type = 'print'

    default_data = ['energy', 'time']

class StandardLogOutput(LogOutput):

    """
    Standard protocol output action

    A StandardLogOutput object can be used in the action list of any
    trajectory-generating operation. It is a specialization of
    LogOutput to the most common case and writes data in the categories
    "time" and "energy" to the standard output stream.

    :param skip: the number of steps to skip between two action runs
    :type skip: int
    """

    def __init__(self, skip=50):
        LogOutput.__init__(self, sys.stdout, None, 0, None, skip)

#
# Snapshot generator
#
class SnapshotGenerator(TrajectoryGenerator):

    """
    Trajectory generator for single steps

    A SnapshotGenerator is used for manual assembly of trajectory
    files. At each call it writes one step to the trajectory,
    using the current state of the universe (configuration, velocities, etc.)
    and data provided explicitly with the call.

    Each call to the SnapshotGenerator object produces one step.
    All the keyword options can be specified either when
    creating the generator or when calling it.
    """

    def __init__(self, universe, **options):
        """
        :param universe: the universe on which the generator acts
        :keyword data: a dictionary that supplies values for variables
                       that are not part of the universe state
                       (e.g. potential energy)
        :keyword actions: a list of actions to be executed periodically
                          (default is none)
        """
        TrajectoryGenerator.__init__(self, universe, options)
        self.available_data = []
        try:
            e, g = self.universe.energyAndGradients()
        except: pass
        else:
            self.available_data.append('energy')
            self.available_data.append('gradients')
        try:
            self.universe.configuration()
            self.available_data.append('configuration')
        except: pass
        if self.universe.cellVolume() is not None:
            self.available_data.append('thermodynamic')
        if self.universe.velocities() is not None:
            self.available_data.append('velocities')
            self.available_data.append('energy')
            self.available_data.append('thermodynamic')

    default_options = {'steps': 0, 'actions': []}

    def __call__(self, **options):
        self.setCallOptions(options)
        from MMTK_trajectory import snapshot
        data = copy.copy(options.get('data', {}))
        energy_terms = 0
        for name in data.keys():
            if name == 'time' and 'time' not in self.available_data:
                self.available_data.append('time')
            if  name[-7:] == '_energy':
                energy_terms = energy_terms + 1
                if 'energy' not in self.available_data:
                    self.available_data.append('energy')
            if (name == 'temperature' or name == 'pressure') \
               and 'thermodynamic' not in self.available_data:
                self.available_data.append('thermodynamic')
            if name == 'gradients' and 'gradients' not in self.available_data:
                self.available_data.append('gradients')
        actions = self.getActions()
        for action in actions:
            categories = action[-1]
            for c in categories:
                if c == 'energy' and not data.has_key('kinetic_energy'):
                    v = self.universe.velocities()
                    if v is not None:
                        m = self.universe.masses()
                        e = (v*v*m*0.5).sumOverParticles()
                        data['kinetic_energy'] = e
                        df = self.universe.degreesOfFreedom()
                        data['temperature'] = 2.*e/df/Units.k_B/Units.K
                if c == 'configuration':
                    if  data.has_key('configuration'):
                        data['configuration'] = data['configuration'].array
                    else:
                        data['configuration'] = \
                                         self.universe.configuration().array
                if c == 'velocities':
                    if  data.has_key('velocities'):
                        data['velocities'] = data['velocities'].array
                    else:
                        data['velocities'] = self.universe.velocities().array
                if c == 'gradients':
                    if  data.has_key('gradients'):
                        data['gradients'] = data['gradients'].array
                p = self.universe.cellParameters()
                if p is not None:
                    data['box_size'] = p
                volume = self.universe.cellVolume()
                if volume is not None:
                    data['volume'] = volume
                try:
                    m = self.universe.masses()
                    data['masses'] = m.array
                except: pass
        snapshot(self.universe, data, actions, energy_terms)

#
# Trajectory reader (not yet functional...)
#
if False:

    class TrajectoryReader(TrajectoryGenerator):

        def __init__(self, trajectory, options):
            TrajectoryGenerator.__init__(self, trajectory.universe, options)
            self.input = trajectory
            self.available_data = trajectory.variables()

        default_options = {'trajectory': None, 'log': None, 'options': []}

        def __call__(self, **options):
            self.setCallOptions(options)
            from MMTK_trajectory import readTrajectory
            readTrajectory(self.universe, self.input.trajectory,
                           [self.getOption('trajectory'),
                            self.getOption('log')] +
                           self.getOption('options'))

#
# Print information about trajectory file
#
def trajectoryInfo(filename):
    """
    :param filename: the name of a trajectory file
    :type filename: str
    :returns: a string with summarial information about the trajectory
    """
    from Scientific.IO import NetCDF
    file = NetCDF.NetCDFFile(filename, 'r')
    nsteps = file.variables['step'].shape[0]
    if 'minor_step_number' in file.dimensions.keys():
        nsteps = nsteps*file.variables['step'].shape[1]
    s = 'Information about trajectory file ' + filename + ':\n'
    try:
        s += file.comment + '\n'
    except AttributeError:
        pass
    s += `file.dimensions['atom_number']` + ' atoms\n'
    s += `nsteps` + ' steps\n'
    s += file.history
    file.close()
    return s