Source code for firedrake.progress_bar
"""A module providing progress bars."""
from pyop2.mpi import COMM_WORLD
from progress.bar import FillingSquaresBar
class _NullProgressBar:
"""A placeholder class with the same interface as a progress bar."""
def __init__(self, *args, **kwargs):
pass
def __enter__(self):
pass
def __exit__(self, *args, **kwargs):
pass
def iter(self, iterator):
return iterator
[docs]
class ProgressBar(FillingSquaresBar):
"""A progress bar for simulation execution.
This is a subclass of ``progress.bar.FillingSquaresBar`` which is
configured to be suitable for tracking progress in forward and adjoint
simulations. It is also extended to only output on rank 0 in parallel.
Parameters
----------
message : str
An identifying string to be prepended to the progress bar. This
defaults to an empty string.
comm : mpi4py.MPI.Intracomm
The MPI communicator over which the simulation is run. Defaults to
`COMM_WORLD`
Notes
-----
Further parameters can be passed as per the `progress package documentation
<https://github.com/verigak/progress>`_, or you can customise
further by subclassing.
Examples
--------
To apply a progress bar to a loop, wrap the loop iterator in the
``iter`` method of a ``ProgressBar``:
>>> for t in ProgressBar("Timestep").iter(np.linspace(0.0, 1.0, 10)):
... sleep(0.2)
...
Timestep ▣▣▣▣▣▣▣▣▣▣▣▣▣▣▣▣▣▣▣▣▣▣▣▣▣▣▣▣▣▣▣▣▣▣▣▣▣▣▣▣▣▣▣▣▣▣▣▣▣▣ 10/10 [0:00:02]
To see progress bars for functional, adjoint and Hessian evaluations in an
adjoint simulation, set the ``progress_bar`` attribute of the tape to
`ProgressBar`:
>>> get_working_tape().progress_bar = ProgressBar
This use case is covered in the documentation for :class:`pyadjoint.Tape`.
"""
def __new__(cls, *args, comm=COMM_WORLD, **kwargs):
# Only print a progress bar on rank 0.
if comm.rank == 0:
return super().__new__(cls)
else:
return _NullProgressBar()
# Sensible default suffix for timestepping or adjoint loops.
suffix = "%(index)s/%(max)s [%(elapsed_td)s]"
# Required in order to have output in parallel.
check_tty = False
width = 50
