Adding folder clearing function to staNMF.py, adjusting staNMF.py

[amyecampbell]

In this pull request (to be followed with one with cluster driving scripts), I add the ClearDirectory() method to fulfill the enhancement suggested by @gwaygenomics, and update staNMF.py to allow it to be more easily parallelized. I do so with the following changes:

• The constructor now allows the user to input either an integer or a range of integers(in the form of a python list) for the replicates variable. This way, the user can either calculate all factorizations of a given K at once(ex: replicates=100 would output factorization files 1 through 99), or break them into separate calls (ex: replicates=[0,50] in one call, replicates=[50,100] in another).
• The instability() function now outputs an instability.csv file for each separate K if the user sets instability(parallel=True), so that instability can be calculated for each K independently, and then be used by the same call to plot().
• The plot() function can now either plot instability based on the staNMF object's existing self.instabilitydict, or can collect instability measures from each K folder and plot those.
• The call to import matplotlib specifies the 'Agg' backend in order to be able to save plots on remote machines. I do know that including the matplotlib.use() call in the section of imported files conflicts with pep8 style, but I do want to import plt as a shortcut for matplotlib.pyplot (and have to do it after specifying the backend). Any ideas for workarounds for this?

[amyecampbell]

@dhimmel @gwaygenomics how do I follow a new pull request for the newest script? would I have had to merge this one first, then file a new one for the driver script I added?

[dhimmel]

@dhimmel @gwaygenomics how do I follow a new pull request for the newest script? would I have had to merge this one first, then file a new one for the driver script I added?

You can have multiple open pull requests at once. You need to use branches. It's easiest if the pull requests don't modify the same files and especially the same lines within the same file.

[amyecampbell]

@dhimmel ok, so If I've already pushed a commit with the staNMF_driver.py script, I can branch that and file a PR on it? Or do I have to first pull a branch from the repo, push the driver script to it, and then file a PR on that?

[dhimmel]

ok, so If I've already pushed a commit with the staNMF_driver.py script, I can branch that and file a PR on it?

If you create a branch from amyecampbell:master (which has your commit a185ff909590cf004a1d6cd65cfd39d3fc624319 that added staNMF_driver.py) and create a pull request from the branch, that pull request will be identical to this one.

What you probably want is to create a branch from the latest commit in greenelab:master and then cherry pick a185ff909590cf004a1d6cd65cfd39d3fc624319. This way the two pull requests will be independent.

However, note the branch names will get confusing. I never make commits directly to my master branch. Instead I fast forward my master branch to match the upstream master. Since you based this pull request off your master, things will be a bit more confusing.

Maybe we should wait until Monday. It may help for you to familiarize yourself with https://gist.github.com/Chaser324/ce0505fbed06b947d962.