Source code for openmdao.recorders.dump_recorder
""" Class definition for DumpRecorder, a recorder that prints
human-readable text output to a stream."""
import sys
from six import string_types, iteritems
from openmdao.core.mpi_wrap import MPI
from openmdao.recorders.base_recorder import BaseRecorder
from openmdao.util.record_util import format_iteration_coordinate
[docs]class DumpRecorder(BaseRecorder):
"""Dumps cases in a "pretty" form to `out`, which may be a string or a
file-like object (defaults to ``stdout``). If `out` is ``stdout`` or
``stderr``, then that standard stream is used. Otherwise, if `out` is a
string, then a file with that name will be opened in the current
directory. If `out` is None, cases will be ignored. When called under
MPI, the dumprecorder writes to a separate file for each rank, with the
rank number appended to each filename. In this case, only variables that
exist on all processes can be printed.
"""
def __init__(self, out='stdout'):
super(DumpRecorder, self).__init__()
self._parallel = True
if isinstance(out, string_types):
if out == 'stdout':
out = sys.stdout
elif out == 'stderr':
out = sys.stderr
else:
# Dump to separate file for each process if we are under MPI
if MPI:
if '.' in out:
parts = out.split('.')
parts[-2] += '_' + str(MPI.COMM_WORLD.rank)
out = '.'.join(parts)
else:
out += str(MPI.COMM_WORLD.rank)
out = open(out, 'w')
self.out = out
[docs] def startup(self, group):
""" Write out info that applies to the entire run.
Args
----
group : `Group`
Group that owns this recorder.
"""
super(DumpRecorder, self).startup(group)
[docs] def record_iteration(self, params, unknowns, resids, metadata):
"""Dump the given run data in a "pretty" form.
Args
----
params : `VecWrapper`
`VecWrapper` containing parameters. (p)
unknowns : `VecWrapper`
`VecWrapper` containing outputs and states. (u)
resids : `VecWrapper`
`VecWrapper` containing residuals. (r)
metadata : dict
Dictionary containing execution metadata (e.g. iteration coordinate).
"""
if not self.out: # if self.out is None, just do nothing
return
iteration_coordinate = metadata['coord']
timestamp = metadata['timestamp']
params, unknowns, resids = self._filter_vectors(params, unknowns, resids, iteration_coordinate)
write = self.out.write
fmat = "Timestamp: {0!r}\n"
write(fmat.format(timestamp))
fmat = "Iteration Coordinate: {0:s}\n"
write(fmat.format(format_iteration_coordinate(iteration_coordinate)))
if self.options['record_params']:
write("Params:\n")
for param, val in sorted(iteritems(params)):
write(" {0}: {1}\n".format(param, str(val)))
if self.options['record_unknowns']:
write("Unknowns:\n")
for unknown, val in sorted(iteritems(unknowns)):
write(" {0}: {1}\n".format(unknown, str(val)))
if self.options['record_resids']:
write("Resids:\n")
for resid, val in sorted(iteritems(resids)):
write(" {0}: {1}\n".format(resid, str(val)))
# Flush once per iteration to allow external scripts to process the data.
self.out.flush()
