Closed calebgeniesse closed 9 years ago
Not to do anything to delay or nay-say this merge, but I just want to make a comment that I feel features in files/scripts which assume or are special cased for running on a particular machine or cluster (or worse, with a particular user's directory) are highly sub-optimal. It makes running programs and scripts on other machines or in other labs difficult, which really can hurt scientific reproducibility. (Worse case scenario, the script is set up such that any attempts by other users or other labs are an exercise in frustration.)
Generally speaking, it's best to specify cluster/machine/user specific information via command line options, environment variables or config files, so there's an explicit and (relatively) straightforward way of controlling the behavior for people who wish to use it on other machines.
(We now return to our regularly-scheduled pull request, already in progress.)
@roccomoretti thanks for the input--I totally agree! I don't have time to refactor all of the hard-coding now, but I will make sure to generalize the code introduced in this PR.
@roccomoretti its a little hacky, but hopefully my last commit will prevent some user issues down the road.. thanks again!
@cyucheng i've updated rosetta_submit.py so that individual logs are saved for jobs submitted via ppserver.py
@rhiju, @kkappel1, @cyucheng, @jyesselm: we own nodes on sherlock... let's use them!