Fetching contributors…
Cannot retrieve contributors at this time
201 lines (148 sloc) 7.07 KB

Jug: A Task-Based Parallelization Framework

Jug allows you to write code that is broken up into tasks and run different tasks on different processors.

It uses the filesystem to communicate between processes and works correctly over NFS, so you can coordinate processes on different machines.

Jug is a pure Python implementation and should work on any platform.

Python 2.6/2.7 and Python 3.3+ are supported.



Video: On vimeo or showmedo

Mailing List:

Short Example

Here is a one minute example. Save the following to a file called (if you have installed jug, you can obtain a slightly longer version of this example by running jug demo on the command line):

from jug import TaskGenerator
from time import sleep

def is_prime(n):
    for j in range(2,n-1):
        if (n % j) == 0:
            return False
    return True

primes100 = [is_prime(n) for n in range(2,101)]

This is a brute-force way to find all the prime numbers up to 100. Of course, this is only for didactical purposes, normally you would use a better method. Similarly, the sleep function is so that it does not run too fast. Still, it illustrates the basic functionality of Jug for embarassingly parallel problems.

Type jug status to get:

Task name                  Waiting       Ready    Finished     Running
primes.is_prime                  0          99           0           0
Total:                           0          99           0           0

This tells you that you have 99 tasks called primes.is_prime ready to run. So run jug execute &. You can even run multiple instances in the background (if you have multiple cores, for example). After starting 4 instances and waiting a few seconds, you can check the status again (with jug status

Task name                  Waiting       Ready    Finished     Running
primes.is_prime                  0          63          32           4
Total:                           0          63          32           4

Now you have 32 tasks finished, 4 running, and 63 still ready. Eventually, they will all finish and you can inspect the results with jug shell This will give you an ipython shell. The primes100 variable is available, but it is an ugly list of jug.Task objects. To get the actual value, you call the value function:

In [1]: primes100 = value(primes100)

In [2]: primes100[:10]
Out[2]: [True, True, False, True, False, True, False, False, False, True]


"I've been using jug with great success to distribute the running of a reasonably large set of parameter combinations" - Andreas Longva

What's New

version 1.6.1 (Thu Aug 29 2017) - Fix bug with invalidate() in the shell

version 1.6.0 (Thu Aug 24 2017) - Add 'graph' subcommand - Generates a graph of tasks - 'jug execute --keep-going' now ends with non-zero exit code in case of failures - Fix bug with cleanup in dict_store not providing the number of removed records - Add 'jug cleanup --keep-locks' to remove obsolete results without affecting locks

version 1.5.0 (Sun Jul 16 2017) - Add 'demo' subcommand - Add is_jug_running() function - Fix bug in finding config files - Improved --debug mode: check for unsupported recursive task creation - Add invalidate() to shell environment - Use ~/.config/jug/jugrc as configuration file - Add experimental support for extensible commands, use

  • jugrc: execute_wait_cycle_time_secs is now execute_wait_cycle_time
  • Expose sync_move in jug.utils

version 1.4.0 (Tue Jan 3 2017) - Fix bug with writing very large objects to disk - Smarter handling of --aggressive-unload (do not unload what will be

immediately necessary)
  • Work around corner case in jug shell command
  • Add test-jug subcommand
  • Add return_tuple decorator

version 1.3.0 (Tue Nov 1 2016) - Update shell subcommand to IPython 5 - Use ~/.config/jugrc as configuration file - Cleanup usage string - Use bottle instead of for webstatus subcommand - Add jug_execute function - Add timing functionality

version 1.2.2 (Sat Jun 25 2016) - Fix bugs in shell subcommand and a few corner cases in encoding/decoding


version 1.2.1 (Mon Feb 15 2016) - Changed execution loop to ensure that all tasks are checked (issue #33 on

  • Fixed bug that made 'check' or 'sleep-until' slower than necessary
  • Fixed jug on Windows (which does not support fsync on directories)
  • Made Tasklets use slightly less memory

version 1.2 (Thu Aug 20 2015) - Use HIGHEST_PROTOCOL when pickle()ing - Add compress_numpy option to file_store - Add register_hook_once function - Optimize case when most (or all) tasks are already run - Add --short option to 'jug status' and 'jug execute' - Fix bug with dictionary order in kwargs (fix by Andreas Sorge) - Fix ipython colors (fix by Andreas Sorge) - Sort tasks in 'jug status'

version 1.1 (Tue Mar 3 2015) - Python 3 compatibility fixes - fsync(directory) in file backend - Jug hooks (still mostly undocumented, but already enabling internal code simplification)

version 1.0 (Tue May 20 2014) - Adapt status output to terminal width (by Alex Ford) - Add a newline at the end of lockfiles for file backend - Add --cache-file option to specify file for status --cache

version 0.9.7 (Tue Feb 18 2014)

  • Fix use of numpy subclasses
  • Fix redis URL parsing
  • Fix shell for newer versions of IPython
  • Correctly fall back on non-sqlite status
  • Allow user to call set_jugdir() inside jugfile

version 0.9.6 (Tue Aug 6 2013)

  • Faster decoding
  • Add jug-execute script
  • Add describe() function
  • Add write_task_out() function

version 0.9.5 (May 27 2013)

  • Added debug mode
  • Even better using blocked access
  • Python 3 support
  • Documentation improvements

For older version see ChangeLog file.

Join the chat at