set_context function

[andrewheusser]

seaborn has a nice way to set the plot styling for a particular context (like figures for a talk, poster, paper or notebook)

I was thinking we could take a similar approach with superEEG, setting a context to run the code on a single computer, cluster, or some other setups. I idea would be to:

import superEEG
superEEG.set_context('google-cluster')

This would change an internal config object that determines the behavior of our functions (how the jobs are split up etc). thoughts?

[jeremymanning]

That a great idea! I like it.

[andrewheusser]

Here's how I currently have it implemented. Example code:

import superEEG

# print default context
print(superEEG.context)

# change context to preset cluster default
superEEG.set_context('cluster')

# print updated context
print(superEEG.context)

# change context to custom dict
google_cluster = {
    'environment' : 'google-cluster',
    'nodes' : 1000000000,
    'memory' : 3000000000,
}
superEEG.set_context(google_cluster)
print(superEEG.context)

Result:

{'environment': 'single', 'nodes': 1, 'memory': 8}
{'environment': 'cluster', 'nodes': 4, 'memory': 30}
{'environment': 'google-cluster', 'nodes': 1000000, 'memory': 3000000000}

We can then use this module attribute to make decisions. What do you think of this setup? What other attributes should we add to this dict?

[jeremymanning]

Job submission command, code directory, scratch directory, data directory, and probably others as we go...