Are there any downsides to using virtualenv for scientific python and machine learning? -

i have received several recommendations use virtualenv clean python modules. concerned because seems true. has found downside related performance or memory issues in working multicore settings, starcluster, numpy, scikit-learn, pandas, or ipython notebook.

virtualenv best , easiest way keep sort of order when comes dependencies. python behind ruby (bundler!) when comes dealing installing , keeping track of modules. best tool have virtualenv.

so suggest create virtualenv directory each of applications, put file list 'pip install' commands need build environment , ensure have clean repeatable process creating environment.

i think nature of application makes little difference. there should not performance issue since virtualenv load libraries specific path rather load them directory saved default.

in case (this may irrelevant), if performance issue, perhaps ought looking @ compiled language. though, performance bottlenecks improved better coding.