westpa 2022.12__cp313-cp313-macosx_10_13_x86_64.whl

westpa/core/propagators/executable.py

@@ -0,0 +1,719 @@
+import logging
+import os
+import shutil
+import signal
+import subprocess
+import sys
+import tempfile
+import time
+import tarfile
+import pickle
+from io import BytesIO
+
+import numpy as np
+from numpy.random import MT19937, Generator
+
+import westpa
+from westpa.core.extloader import get_object
+from westpa.core.propagators import WESTPropagator
+from westpa.core.states import BasisState, InitialState, return_state_type
+from westpa.core.segment import Segment
+from westpa.core.yamlcfg import check_bool
+
+from westpa.core.trajectory import load_trajectory
+from westpa.core.h5io import safe_extract
+
+log = logging.getLogger(__name__)
+
+# Get a list of user-friendly signal names
+SIGNAL_NAMES = {getattr(signal, name): name for name in dir(signal) if name.startswith('SIG') and not name.startswith('SIG_')}
+
+
+def pcoord_loader(fieldname, pcoord_return_filename, destobj, single_point):
+    """Read progress coordinate data into the ``pcoord`` field on ``destobj``.
+    An exception will be raised if the data is malformed.  If ``single_point`` is true,
+    then only one (N-dimensional) point will be read, otherwise system.pcoord_len points
+    will be read.
+    """
+
+    system = westpa.rc.get_system_driver()
+
+    assert fieldname == 'pcoord'
+
+    pcoord = np.loadtxt(pcoord_return_filename, dtype=system.pcoord_dtype)
+
+    if single_point:
+        expected_shape = (system.pcoord_ndim,)
+        if pcoord.ndim == 0:
+            pcoord.shape = (1,)
+    else:
+        expected_shape = (system.pcoord_len, system.pcoord_ndim)
+        if pcoord.ndim < 2:
+            pcoord.shape = expected_shape
+    if pcoord.shape != expected_shape:
+        raise ValueError(
+            'progress coordinate data has incorrect shape {!r} [expected {!r}] Check pcoord.err or seg_logs for more information.'.format(
+                pcoord.shape, expected_shape
+            )
+        )
+    destobj.pcoord = pcoord
+
+
+def aux_data_loader(fieldname, data_filename, segment, single_point):
+    data = np.loadtxt(data_filename)
+    segment.data[fieldname] = data
+    if data.nbytes == 0:
+        raise ValueError('could not read any data for {}'.format(fieldname))
+
+
+def npy_data_loader(fieldname, coord_file, segment, single_point):
+    log.debug('using npy_data_loader')
+    data = np.load(coord_file, allow_pickle=True)
+    segment.data[fieldname] = data
+    if data.nbytes == 0:
+        raise ValueError('could not read any data for {}'.format(fieldname))
+
+
+def pickle_data_loader(fieldname, coord_file, segment, single_point):
+    log.debug('using pickle_data_loader')
+    with open(coord_file, 'rb') as fo:
+        data = pickle.load(fo)
+    segment.data[fieldname] = data
+    if data.nbytes == 0:
+        raise ValueError('could not read any data for {}'.format(fieldname))
+
+
+def trajectory_loader(fieldname, coord_folder, segment, single_point):
+    '''Load data from the trajectory return. ``coord_folder`` should be the path to a folder
+    containing trajectory files. ``segment`` is the ``Segment`` object that the data is associated with.
+    Please see ``load_trajectory`` for more details. ``single_point`` is not used by this loader.'''
+    try:
+        data = load_trajectory(coord_folder)
+        segment.data['iterh5/trajectory'] = data
+    except Exception as e:
+        log.warning('could not read any {} data for HDF5 Framework: {}'.format(fieldname, str(e)))
+
+
+def restart_loader(fieldname, restart_folder, segment, single_point):
+    '''Load data from the restart return. The loader will tar all files in ``restart_folder``
+    and store it in the per-iteration HDF5 file. ``segment`` is the ``Segment`` object that
+    the data is associated with. ``single_point`` is not used by this loader.'''
+    try:
+        d = BytesIO()
+        with tarfile.open(mode='w:gz', fileobj=d) as t:
+            t.add(restart_folder, arcname='.')
+
+        segment.data['iterh5/restart'] = d.getvalue() + b'\x01'  # add tail protection
+    except Exception as e:
+        log.warning('could not read any {} data for HDF5 Framework: {}'.format(fieldname, str(e)))
+    finally:
+        d.close()
+
+
+def restart_writer(path, segment):
+    '''Prepare the necessary files from the per-iteration HDF5 file to run ``segment``.'''
+    try:
+        restart = segment.data.pop('iterh5/restart', None)
+        # Making an exception for start states in iteration 1
+        if restart is None:
+            raise ValueError('restart data is not present')
+
+        d = BytesIO(restart[:-1])  # remove tail protection
+        with tarfile.open(fileobj=d, mode='r:gz') as t:
+            safe_extract(t, path=path)
+
+    except ValueError as e:
+        log.warning('could not write HDF5 Framework restart data for {}: {}'.format(str(segment), str(e)))
+        d = BytesIO()
+        if segment.n_iter == 1:
+            log.warning(
+                'In iteration 1. Assuming this is a start state and proceeding to skip reading restart from per-iteration HDF5 file for {}'.format(
+                    str(segment)
+                )
+            )
+    except Exception as e:
+        log.warning('could not write HDF5 Framework restart data for {}: {}'.format(str(segment), str(e)))
+    finally:
+        d.close()
+
+
+def seglog_loader(fieldname, log_file, segment, single_point):
+    '''Load data from the log return. The loader will tar all files in ``log_file``
+    and store it in the per-iteration HDF5 file. ``segment`` is the ``Segment`` object that
+    the data is associated with. ``single_point`` is not used by this loader.'''
+    try:
+        d = BytesIO()
+        with tarfile.open(mode='w:gz', fileobj=d) as t:
+            t.add(log_file, arcname='.')
+
+        segment.data['iterh5/log'] = d.getvalue() + b'\x01'  # add tail protection
+    except Exception as e:
+
+        log.warning('could not read any data for {}: {}'.format(fieldname, str(e)))
+    finally:
+        d.close()
+
+
+# Dictionary with all the possible loaders
+data_loaders = {
+    'default': aux_data_loader,
+    'auxdata_loader': aux_data_loader,
+    'aux_data_loader': aux_data_loader,
+    'npy_loader': npy_data_loader,
+    'npy_data_loader': npy_data_loader,
+    'pickle_loader': pickle_data_loader,
+    'pickle_data_loader': pickle_data_loader,
+}
+
+
+class ExecutablePropagator(WESTPropagator):
+    ENV_CURRENT_ITER = 'WEST_CURRENT_ITER'
+
+    # Environment variables set during propagation
+    ENV_CURRENT_SEG_ID = 'WEST_CURRENT_SEG_ID'
+    ENV_CURRENT_SEG_DATA_REF = 'WEST_CURRENT_SEG_DATA_REF'
+    ENV_CURRENT_SEG_INITPOINT = 'WEST_CURRENT_SEG_INITPOINT_TYPE'
+    ENV_PARENT_SEG_ID = 'WEST_PARENT_ID'
+    ENV_PARENT_DATA_REF = 'WEST_PARENT_DATA_REF'
+
+    # Environment variables set during propagation and state generation
+    ENV_BSTATE_ID = 'WEST_BSTATE_ID'
+    ENV_BSTATE_DATA_REF = 'WEST_BSTATE_DATA_REF'
+    ENV_ISTATE_ID = 'WEST_ISTATE_ID'
+    ENV_ISTATE_DATA_REF = 'WEST_ISTATE_DATA_REF'
+
+    # Environment variables for progress coordinate calculation
+    ENV_STRUCT_DATA_REF = 'WEST_STRUCT_DATA_REF'
+
+    ENV_RAND16 = 'WEST_RAND16'
+    ENV_RAND32 = 'WEST_RAND32'
+    ENV_RAND64 = 'WEST_RAND64'
+    ENV_RAND128 = 'WEST_RAND128'
+    ENV_RANDFLOAT = 'WEST_RANDFLOAT'
+
+    def __init__(self, rc=None):
+        super().__init__(rc)
+
+        # A mapping of environment variables to template strings which will be
+        # added to the environment of all children launched.
+        self.addtl_child_environ = dict()
+
+        # A mapping of executable name ('propagator', 'pre_iteration', 'post_iteration') to
+        # a dictionary of attributes like 'executable', 'stdout', 'stderr', 'environ', etc.
+        self.exe_info = {}
+        self.exe_info['propagator'] = {}
+        self.exe_info['pre_iteration'] = {}
+        self.exe_info['post_iteration'] = {}
+        self.exe_info['get_pcoord'] = {}
+        self.exe_info['gen_istate'] = {}
+
+        # A mapping of data set name ('pcoord', 'coord', 'com', etc) to a dictionary of
+        # attributes like 'loader', 'dtype', etc
+        self.data_info = {}
+        self.data_info['pcoord'] = {}
+
+        # Validate configuration
+        config = self.rc.config
+
+        for key in [
+            ('west', 'executable', 'propagator', 'executable'),
+            ('west', 'data', 'data_refs', 'segment'),
+            ('west', 'data', 'data_refs', 'basis_state'),
+            ('west', 'data', 'data_refs', 'initial_state'),
+        ]:
+            config.require(key)
+
+        self.segment_ref_template = config['west', 'data', 'data_refs', 'segment']
+        self.basis_state_ref_template = config['west', 'data', 'data_refs', 'basis_state']
+        self.initial_state_ref_template = config['west', 'data', 'data_refs', 'initial_state']
+        store_h5 = config.get(['west', 'data', 'data_refs', 'iteration']) is not None
+
+        # Create a persistent RNG for each worker
+        self.rng = Generator(MT19937())
+
+        # Load additional environment variables for all child processes
+        self.addtl_child_environ.update({k: str(v) for k, v in (config['west', 'executable', 'environ'] or {}).items()})
+
+        # Load configuration items relating to child processes
+        for child_type in ('propagator', 'pre_iteration', 'post_iteration', 'get_pcoord', 'gen_istate', 'subgroup_walkers'):
+            child_info = config.get(['west', 'executable', child_type])
+            if not child_info:
+                continue
+
+            info_prefix = ['west', 'executable', child_type]
+
+            # require executable to be specified if anything is specified at all
+            config.require(info_prefix + ['executable'])
+
+            self.exe_info[child_type]['executable'] = child_info['executable']
+            self.exe_info[child_type]['stdin'] = child_info.get('stdin', os.devnull)
+            self.exe_info[child_type]['stdout'] = child_info.get('stdout', None)
+            self.exe_info[child_type]['stderr'] = child_info.get('stderr', None)
+            self.exe_info[child_type]['cwd'] = child_info.get('cwd', None)
+
+            if child_type not in ('propagator', 'get_pcoord', 'gen_istate'):
+                self.exe_info[child_type]['enabled'] = child_info.get('enabled', True)
+            else:
+                # for consistency, propagator, get_pcoord, and gen_istate can never be disabled
+                self.exe_info[child_type]['enabled'] = True
+
+            # apply environment modifications specific to this executable
+            self.exe_info[child_type]['environ'] = {k: str(v) for k, v in (child_info.get('environ') or {}).items()}
+
+        log.debug('exe_info: {!r}'.format(self.exe_info))
+
+        # Load configuration items relating to dataset input
+        self.data_info['pcoord'] = {'name': 'pcoord', 'loader': pcoord_loader, 'enabled': True, 'filename': None, 'dir': False}
+        self.data_info['trajectory'] = {
+            'name': 'trajectory',
+            'loader': trajectory_loader,
+            'enabled': store_h5,
+            'filename': None,
+            'dir': True,
+        }
+        self.data_info['restart'] = {
+            'name': 'restart',
+            'loader': restart_loader,
+            'enabled': store_h5,
+            'filename': None,
+            'dir': True,
+        }
+        self.data_info['log'] = {'name': 'seglog', 'loader': seglog_loader, 'enabled': store_h5, 'filename': None, 'dir': False}
+
+        # Grab config from west.executable.datasets, else fallback to west.data.datasets.
+        dataset_configs = config.get(["west", "executable", "datasets"]) or config.get(['west', 'data', 'datasets'], {})
+        for dsinfo in dataset_configs:
+            try:
+                dsname = dsinfo['name']
+            except KeyError:
+                raise ValueError('dataset specifications require a ``name`` field')
+
+            if dsname == 'pcoord':
+                # can never disable pcoord collection
+                dsinfo['enabled'] = True
+            else:
+                check_bool(dsinfo.setdefault('enabled', True))
+
+            loader_directive = dsinfo.get('loader', None)
+            if callable(loader_directive):
+                loader = loader_directive
+            elif loader_directive in data_loaders.keys():
+                if dsname not in ['pcoord', 'seglog', 'restart', 'trajectory']:
+                    loader = data_loaders[loader_directive]
+                else:
+                    loader = get_object(loader_directive)
+            elif dsname not in ['pcoord', 'seglog', 'restart', 'trajectory']:
+                loader = aux_data_loader
+            else:
+                # YOLO. Or maybe it wasn't specified.
+                loader = loader_directive
+
+            if loader:
+                dsinfo['loader'] = loader
+            self.data_info.setdefault(dsname, {}).update(dsinfo)
+
+        log.debug('data_info: {!r}'.format(self.data_info))
+
+    @staticmethod
+    def makepath(template, template_args=None, expanduser=True, expandvars=True, abspath=False, realpath=False):
+        template_args = template_args or {}
+        path = template.format(**template_args)
+        if expandvars:
+            path = os.path.expandvars(path)
+        if expanduser:
+            path = os.path.expanduser(path)
+        if realpath:
+            path = os.path.realpath(path)
+        if abspath:
+            path = os.path.abspath(path)
+        path = os.path.normpath(path)
+        return path
+
+    def random_val_env_vars(self):
+        '''Return a set of environment variables containing random seeds. These are returned
+        as a dictionary, suitable for use in ``os.environ.update()`` or as the ``env`` argument to
+        ``subprocess.Popen()``. Every child process executed by ``exec_child()`` gets these.'''
+
+        return {
+            self.ENV_RAND16: str(self.rng.integers(2**16, dtype=np.uint16)),
+            self.ENV_RAND32: str(self.rng.integers(2**32, dtype=np.uint32)),
+            self.ENV_RAND64: str(self.rng.integers(2**64, dtype=np.uint64)),
+            self.ENV_RAND128: str(int(self.rng.integers(2**64, dtype=np.uint64)) + int(self.rng.integers(2**64, dtype=np.uint64))),
+            self.ENV_RANDFLOAT: str(self.rng.random()),
+        }
+
+    def exec_child(self, executable, environ=None, stdin=None, stdout=None, stderr=None, cwd=None):
+        '''Execute a child process with the environment set from the current environment, the
+        values of self.addtl_child_environ, the random numbers returned by self.random_val_env_vars, and
+        the given ``environ`` (applied in that order). stdin/stdout/stderr are optionally redirected.
+
+        This function waits on the child process to finish, then returns
+        (rc, rusage), where rc is the child's return code and rusage is the resource usage tuple from os.wait4()'''
+
+        all_environ = dict(os.environ)
+        all_environ.update(self.addtl_child_environ)
+        all_environ.update(self.random_val_env_vars())
+        all_environ.update(environ or {})
+
+        stdin = open(stdin, 'rb') if stdin else sys.stdin
+        stdout = open(stdout, 'wb') if stdout else sys.stdout
+        if stderr == 'stdout':
+            stderr = stdout
+        else:
+            stderr = open(stderr, 'wb') if stderr else sys.stderr
+
+        # close_fds is critical for preventing out-of-file errors
+        proc = subprocess.Popen(
+            [executable],
+            cwd=cwd,
+            stdin=stdin,
+            stdout=stdout,
+            stderr=stderr if stderr != stdout else subprocess.STDOUT,
+            close_fds=True,
+            env=all_environ,
+        )
+
+        # Wait on child and get resource usage
+        (_pid, _status, rusage) = os.wait4(proc.pid, 0)
+        # Do a subprocess.Popen.wait() to let the Popen instance (and subprocess module) know that
+        # we are done with the process, and to get a more friendly return code
+        rc = proc.wait()
+
+        return (rc, rusage)
+
+    def exec_child_from_child_info(self, child_info, template_args, environ):
+        for key, value in child_info.get('environ', {}).items():
+            environ[key] = self.makepath(value)
+        return self.exec_child(
+            executable=self.makepath(child_info['executable'], template_args),
+            environ=environ,
+            cwd=self.makepath(child_info['cwd'], template_args) if child_info['cwd'] else None,
+            stdin=self.makepath(child_info['stdin'], template_args) if child_info['stdin'] else os.devnull,
+            stdout=self.makepath(child_info['stdout'], template_args) if child_info['stdout'] else None,
+            stderr=self.makepath(child_info['stderr'], template_args) if child_info['stderr'] else None,
+        )
+
+    # Functions to create template arguments and environment values for child processes
+    def update_args_env_basis_state(self, template_args, environ, basis_state):
+        new_template_args = {'basis_state': basis_state}
+        new_env = {
+            self.ENV_BSTATE_ID: str(basis_state.state_id if basis_state.state_id is not None else -1),
+            self.ENV_BSTATE_DATA_REF: self.makepath(self.basis_state_ref_template, new_template_args),
+        }
+        template_args.update(new_template_args)
+        environ.update(new_env)
+        return template_args, environ
+
+    def update_args_env_initial_state(self, template_args, environ, initial_state):
+        new_template_args = {'initial_state': initial_state}
+        new_env = {
+            self.ENV_ISTATE_ID: str(initial_state.state_id if initial_state.state_id is not None else -1),
+            self.ENV_ISTATE_DATA_REF: self.makepath(self.initial_state_ref_template, new_template_args),
+        }
+
+        if initial_state.basis_state is not None:
+            basis_state = initial_state.basis_state
+        elif initial_state.istate_type == InitialState.ISTATE_TYPE_START:
+            basis_state = BasisState(
+                label=f"sstate_{initial_state.state_id}", pcoord=initial_state.pcoord, probability=0.0, auxref=""
+            )
+        else:
+            basis_state = self.basis_states[initial_state.basis_state_id]
+
+        self.update_args_env_basis_state(new_template_args, new_env, basis_state)
+
+        template_args.update(new_template_args)
+        environ.update(new_env)
+        return template_args, environ
+
+    def update_args_env_iter(self, template_args, environ, n_iter):
+        environ[self.ENV_CURRENT_ITER] = str(n_iter if n_iter is not None else -1)
+        template_args['n_iter'] = int(n_iter)
+        return template_args, n_iter
+
+    def update_args_env_segment(self, template_args, environ, segment):
+        template_args['segment'] = segment
+
+        environ[self.ENV_CURRENT_SEG_INITPOINT] = Segment.initpoint_type_names[segment.initpoint_type]
+
+        if segment.initpoint_type == Segment.SEG_INITPOINT_CONTINUES:
+            # Could use actual parent object here if the work manager cared to pass that much data
+            # to us (we'd need at least the subset of parents for all segments sent in the call to propagate)
+            # that may make a good west.cfg option for future crazy extensibility, but for now,
+            # just populate the bare minimum
+            parent = Segment(n_iter=segment.n_iter - 1, seg_id=segment.parent_id)
+            parent_template_args = dict(template_args)
+            parent_template_args['segment'] = parent
+
+            environ[self.ENV_PARENT_SEG_ID] = str(segment.parent_id if segment.parent_id is not None else -1)
+            environ[self.ENV_PARENT_DATA_REF] = self.makepath(self.segment_ref_template, parent_template_args)
+        elif segment.initpoint_type == Segment.SEG_INITPOINT_NEWTRAJ:
+            # This segment is initiated from a basis state; WEST_PARENT_SEG_ID and WEST_PARENT_DATA_REF are
+            # set to the basis state ID and data ref
+            initial_state = self.initial_states[segment.initial_state_id]
+
+            if initial_state.istate_type == InitialState.ISTATE_TYPE_START:
+                basis_state = BasisState(
+                    label=f"sstate_{initial_state.state_id}", pcoord=initial_state.pcoord, probability=0.0, auxref=""
+                )
+
+            else:
+                basis_state = self.basis_states[initial_state.basis_state_id]
+
+            if self.ENV_BSTATE_ID not in environ:
+                self.update_args_env_basis_state(template_args, environ, basis_state)
+            if self.ENV_ISTATE_ID not in environ:
+                self.update_args_env_initial_state(template_args, environ, initial_state)
+
+            assert initial_state.istate_type in (
+                InitialState.ISTATE_TYPE_BASIS,
+                InitialState.ISTATE_TYPE_GENERATED,
+                InitialState.ISTATE_TYPE_START,
+            )
+            if initial_state.istate_type == InitialState.ISTATE_TYPE_BASIS:
+                environ[self.ENV_PARENT_DATA_REF] = environ[self.ENV_BSTATE_DATA_REF]
+
+            elif initial_state.istate_type == InitialState.ISTATE_TYPE_START:
+                # This points to the start-state PDB
+                environ[self.ENV_PARENT_DATA_REF] = environ[self.ENV_BSTATE_DATA_REF] + '/' + initial_state.basis_auxref
+            else:  # initial_state.type == InitialState.ISTATE_TYPE_GENERATED
+                environ[self.ENV_PARENT_DATA_REF] = environ[self.ENV_ISTATE_DATA_REF]
+
+        environ[self.ENV_CURRENT_SEG_ID] = str(segment.seg_id if segment.seg_id is not None else -1)
+        environ[self.ENV_CURRENT_SEG_DATA_REF] = self.makepath(self.segment_ref_template, template_args)
+        return template_args, environ
+
+    def template_args_for_segment(self, segment):
+        template_args, environ = {}, {}
+        self.update_args_env_iter(template_args, environ, segment.n_iter)
+        self.update_args_env_segment(template_args, environ, segment)
+        return template_args
+
+    def exec_for_segment(self, child_info, segment, addtl_env=None):
+        '''Execute a child process with environment and template expansion from the given
+        segment.'''
+        template_args, environ = {}, {}
+        self.update_args_env_iter(template_args, environ, segment.n_iter)
+        self.update_args_env_segment(template_args, environ, segment)
+        environ.update(addtl_env or {})
+        self.prepare_file_system(segment, environ)
+        child_info['cwd'] = environ[self.ENV_CURRENT_SEG_DATA_REF]
+        return self.exec_child_from_child_info(child_info, template_args, environ)
+
+    def exec_for_iteration(self, child_info, n_iter, addtl_env=None):
+        '''Execute a child process with environment and template expansion from the given
+        iteration number.'''
+        template_args, environ = {}, {}
+        self.update_args_env_iter(template_args, environ, n_iter)
+        environ.update(addtl_env or {})
+        return self.exec_child_from_child_info(child_info, template_args, environ)
+
+    def exec_for_basis_state(self, child_info, basis_state, addtl_env=None):
+        '''Execute a child process with environment and template expansion from the
+        given basis state'''
+        template_args, environ = {}, {}
+        self.update_args_env_basis_state(template_args, environ, basis_state)
+        environ.update(addtl_env or {})
+        return self.exec_child_from_child_info(child_info, template_args, environ)
+
+    def exec_for_initial_state(self, child_info, initial_state, addtl_env=None):
+        '''Execute a child process with environment and template expansion from the given
+        initial state.'''
+        template_args, environ = {}, {}
+        self.update_args_env_initial_state(template_args, environ, initial_state)
+        environ.update(addtl_env or {})
+        return self.exec_child_from_child_info(child_info, template_args, environ)
+
+    def prepare_file_system(self, segment, environ):
+        try:
+            # If the filesystem is properly clean.
+            os.makedirs(environ[self.ENV_CURRENT_SEG_DATA_REF])
+        except Exception:
+            # If the filesystem is NOT properly clean.
+            shutil.rmtree(environ[self.ENV_CURRENT_SEG_DATA_REF])
+            os.makedirs(environ[self.ENV_CURRENT_SEG_DATA_REF])
+        if self.data_info['restart']['enabled']:
+            restart_writer(environ[self.ENV_CURRENT_SEG_DATA_REF], segment=segment)
+
+    def setup_dataset_return(self, segment=None, subset_keys=None):
+        '''Set up temporary files and environment variables that point to them for segment
+        runners to return data. ``segment`` is the ``Segment`` object that the return data
+        is associated with. ``subset_keys`` specifies the names of a subset of data to be
+        returned.'''
+        if subset_keys is None:
+            subset_keys = self.data_info.keys()
+
+        addtl_env = {}
+        return_files = {}
+        del_return_files = {}
+
+        for dataset in self.data_info:
+            if dataset not in subset_keys:
+                continue
+
+            if not self.data_info[dataset].get('enabled', False):
+                continue
+
+            return_template = self.data_info[dataset].get('filename')
+            if return_template:
+                if segment is None:
+                    raise ValueError('segment needs to be provided for dataset return')
+                return_files[dataset] = self.makepath(return_template, self.template_args_for_segment(segment))
+                del_return_files[dataset] = False
+            else:
+                isdir = self.data_info[dataset].get('dir', False)
+                if isdir:
+                    rfname = tempfile.mkdtemp()
+                else:
+                    (fd, rfname) = tempfile.mkstemp()
+                    os.close(fd)
+                return_files[dataset] = rfname
+                del_return_files[dataset] = True
+
+            addtl_env['WEST_{}_RETURN'.format(dataset.upper())] = return_files[dataset]
+
+        return addtl_env, return_files, del_return_files
+
+    def retrieve_dataset_return(self, state, return_files, del_return_files, single_point):
+        '''Retrieve returned data from the temporary locations directed by the environment variables.
+        ``state`` is a ``Segment``, ``BasisState`` , or ``InitialState``object that the return data is
+        associated with. ``return_files`` is a ``dict`` where the keys are the dataset names and
+        the values are the paths to the temporarily files that contain the returned data.
+        ``del_return_files`` is a ``dict`` where the keys are the names of datasets to be deleted
+        (if the corresponding value is set to ``True``) once the data is retrieved.'''
+
+        state_name, state_id = return_state_type(state)
+
+        for dataset in self.data_info:
+            if dataset not in return_files:
+                continue
+
+            # pcoord is always enabled (see __init__)
+            if not self.data_info[dataset].get('enabled', False):
+                continue
+
+            filename = return_files[dataset]
+            loader = self.data_info[dataset]['loader']
+            try:
+                loader(dataset, filename, state, single_point=single_point)
+            except Exception as e:
+                log.error('could not read {} for {} {} from {!r}: {!r}'.format(dataset, state_name, state_id, filename, e))
+                if isinstance(state, Segment):
+                    state.status = state.SEG_STATUS_FAILED
+                break
+            else:
+                if del_return_files.get(dataset, False):
+                    try:
+                        if os.path.isfile(filename):
+                            os.unlink(filename)
+                        else:
+                            shutil.rmtree(filename)
+                    except Exception as e:
+                        log.warning(
+                            'could not delete {} file {!r} for {} {}: {!r}'.format(dataset, filename, state_name, state_id, e)
+                        )
+                    else:
+                        log.debug('deleted {} file {!r} for {} {}'.format(dataset, filename, state_name, state_id))
+
+    # Specific functions required by the WEST framework
+    def get_pcoord(self, state):
+        '''Get the progress coordinate of the given basis or initial state.'''
+
+        template_args, environ = {}, {}
+
+        if isinstance(state, BasisState):
+            execfn = self.exec_for_basis_state
+            self.update_args_env_basis_state(template_args, environ, state)
+            struct_ref = environ[self.ENV_BSTATE_DATA_REF]
+        elif isinstance(state, InitialState):
+            execfn = self.exec_for_initial_state
+            self.update_args_env_initial_state(template_args, environ, state)
+            struct_ref = environ[self.ENV_ISTATE_DATA_REF]
+        else:
+            raise TypeError('state must be a BasisState or InitialState')
+
+        child_info = self.exe_info.get('get_pcoord')
+        addtl_env, return_files, del_return_files = self.setup_dataset_return(
+            subset_keys=['pcoord', 'trajectory', 'restart', 'log']
+        )
+        addtl_env[self.ENV_STRUCT_DATA_REF] = struct_ref
+
+        rc, rusage = execfn(child_info, state, addtl_env)
+        if rc != 0:
+            log.error('get_pcoord executable {!r} returned {}'.format(child_info['executable'], rc))
+
+        self.retrieve_dataset_return(state, return_files, del_return_files, True)
+
+    def gen_istate(self, basis_state, initial_state):
+        '''Generate a new initial state from the given basis state.'''
+        child_info = self.exe_info.get('gen_istate')
+        rc, rusage = self.exec_for_initial_state(child_info, initial_state)
+        if rc != 0:
+            log.error('gen_istate executable {!r} returned {}'.format(child_info['executable'], rc))
+            initial_state.istate_status = InitialState.ISTATE_STATUS_FAILED
+            return
+
+        # Determine and load the progress coordinate value for this state
+        try:
+            self.get_pcoord(initial_state)
+        except Exception:
+            log.exception('could not get progress coordinate for initial state {!r}'.format(initial_state))
+            initial_state.istate_status = InitialState.ISTATE_STATUS_FAILED
+            raise
+        else:
+            initial_state.istate_status = InitialState.ISTATE_STATUS_PREPARED
+
+    def prepare_iteration(self, n_iter, segments):
+        child_info = self.exe_info.get('pre_iteration')
+        if child_info and child_info['enabled']:
+            try:
+                rc, rusage = self.exec_for_iteration(child_info, n_iter)
+            except OSError as e:
+                log.warning('could not execute pre-iteration program {!r}: {}'.format(child_info['executable'], e))
+            else:
+                if rc != 0:
+                    log.warning('pre-iteration executable {!r} returned {}'.format(child_info['executable'], rc))
+
+    def finalize_iteration(self, n_iter, segments):
+        child_info = self.exe_info.get('post_iteration')
+        if child_info and child_info['enabled']:
+            try:
+                rc, rusage = self.exec_for_iteration(child_info, n_iter)
+            except OSError as e:
+                log.warning('could not execute post-iteration program {!r}: {}'.format(child_info['executable'], e))
+            else:
+                if rc != 0:
+                    log.warning('post-iteration executable {!r} returned {}'.format(child_info['executable'], rc))
+
+    def propagate(self, segments):
+        child_info = self.exe_info['propagator']
+
+        for segment in segments:
+            starttime = time.time()
+
+            addtl_env, return_files, del_return_files = self.setup_dataset_return(segment)
+
+            # Spawn propagator and wait for its completion
+            rc, rusage = self.exec_for_segment(child_info, segment, addtl_env)
+
+            if rc == 0:
+                segment.status = Segment.SEG_STATUS_COMPLETE
+            elif rc < 0:
+                log.error('child process for segment %d exited on signal %d (%s)' % (segment.seg_id, -rc, SIGNAL_NAMES[-rc]))
+                segment.status = Segment.SEG_STATUS_FAILED
+                continue
+            else:
+                log.error('child process for segment %d exited with code %d' % (segment.seg_id, rc))
+                segment.status = Segment.SEG_STATUS_FAILED
+                continue
+
+            # Extract data and store on segment for recording in the master thread/process/node
+            self.retrieve_dataset_return(segment, return_files, del_return_files, False)
+
+            if segment.status == Segment.SEG_STATUS_FAILED:
+                continue
+
+            # Record timing info
+            segment.walltime = time.time() - starttime
+            segment.cputime = rusage.ru_utime
+        return segments