Skip to content

emidln/jug

 
 

Repository files navigation

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.

Website: http://luispedro.org/software/jug

Documentation: http://packages.python.org/Jug

Video: On vimeo or showmedo

Mailing List: http://groups.google.com/group/jug-users

Short Example

Here is a one minute example. Save the following to a file called primes.py:

from jug import TaskGenerator
from time import sleep

@TaskGenerator
def is_prime(n):
    sleep(1.)
    for j in xrange(2,n-1):
        if (n % j) == 0:
            return False
    return True

primes100 = map(is_prime, xrange(2,101))

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.

Now type jug status primes.py 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 primes.py &. 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 primes.py):

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 primes.py. 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]

What's New

version 0.9: - In the presence of a barrier(), rerun the jugfile. This makes barrier much

easier to use.
  • Add set_jugdir to public API
  • Added CompoundTaskGenerator
  • Support subclassing of Task
  • Avoid creating directories in file backend unless it is necessary
  • Add jug.mapreduce.reduce (which mimicks the builtin reduce)

version 0.8.1: - Fix redis backend for new version of client module - Faster file store for large files - Fix invalidate with Tasklets - Install tests and have them be runnable - Changed hash computation method. This has a special case on numpy arrays

(for speed) and is more extensible through a __jug_hash__ hook
  • Fix bug with Tasklet dependencies not being properly taken into account
  • Fix shell subcommand in newer versions of ipython
  • Add __file__ attribute to fake jugmodule

version 0.8: - Tasklets - Fix bugs in sleep-until and cleanup - Fix bugs with CompoundTask (you needed to run jug execute twice before)

version 0.7.4: - Fix case where ~/.jug/configrc does not exist - Print host name to lock file on file_store - Refactored implementation of options - Fix unloading tasks that have not run - Fix mapreduce for empty input

Version 0.7.3: - Parse ~/.jug/configrc - Fix bug with waiting times - Special case saving of numpy arrays - Add more expressive jugdir syntax - Save dict_store backend to disk

Version 0.7.2: - included missing files in the distribution

Version 0.7.1: - sleep-until subcommand - bugfixes

Roadmap

Version 1.0

Version 1.0 is just around the corner. After 0.8 is done, there really are not that many features left. More flexible configuration, a bit more caching, and we are done.

After version 1.0

I want to start adding bells&whistles through extensions. Things like timing, more active monitoring, &c.

About

Parallel programming with Python

Resources

License

Stars

Watchers

Forks

Packages

No packages published