Progress Reporter

This library provides a simple interface to register different workloads and visualize their progress with either a plain text progress bar or a jupyter widget, depending on the current environment.

It optionally depends on Jupyter widgets to draw nice progress bars in the interactive Jupyter notebook environment.

Installation

with pip:

pip install progress_reporter

If you use IPython/Jupyter, you are strongly encourage to also install the jupyter widgets:

pip install ipywidgets

Examples

Image you have a class doing some heavy calculations, which are split into several jobs/tasks/threads etc.

In order to visualize the progress, one just needs to derive the worker class from progress_reporter.ProgressReporter and invoke the _progress_register method to tell the reporter how many pieces of work have to be done. Then the reporter is instructed by _progress_update(n) how many of pieces of work have been dispatched.

Note that these are "private" to use this class as a mixin class and not polute the public interface.

from progress_reporter import ProgressReporter
import time

class ExampleWorker(ProgressReporter):
    def __init__(self, n_jobs=100):
        self.n_jobs = n_jobs
        """ register the amount of work with the given description """
        self._progress_register(n_jobs, description='Dispatching jobs')

    def work(self):
        """ do some fake work (sleep) and update the progress via the reporter
        """
        for job in (lambda: time.sleep(0.1) for _ in range(self.n_jobs)):
            job()
            # indicate we've finished one job, to update the progress bar
            self._progress_update(1)

It also supports multi-stage sequential work loads by setting the parameter stage. This is just the dictionary key to the underlying process:

class MultiStageWorker(ProgressReporter):
    def __init__(self, n_jobs_init, n_jobs):
        self.n_jobs_init = n_jobs_init
        self.n_jobs = n_jobs
        """ register an expensive initialization routine """
        self._progress_register(self.n_jobs_init, description='initializing', stage=0)
        """ register the main computation """
        self._progress_register(self.n_jobs, description='main computation', stage=1)

    def work(self):
        """ do the initialization """
        for job in (lambda: time.sleep(0.1) for _ in range(self.n_jobs_init)):
            job()
            self._progress_update(1, stage=0)

        """ perform the next stage of the algorithm """
        for job in (lambda: time.sleep(0.2) for _ in range(self.n_jobs)):
            job()
            self._progress_update(1, stage=1)

Since version 2.0 there is also a version of the this class suitable for compositions.

from progress_reporter import ProgressReporter_

class Estimator(object):
    def fit(self, X, y=None):
        pg = ProgressReporter_()
        pg.register(100, description='work')
        with pg.context(): # ensure progress bars are closed if an exception occurs.
            pg.update(50)
            # ...