We've transitioned our local/dev/prod instances to use conda on Heroku, and couldn't be happier. It was a tiny bit of work to get it set up, but now everything is consistent, and we can set up new local environments in seconds.
So I have been considering this. does conda track pypi or does it lag it? I have been concerned about moving over my requirements.text for a webapp with lots of dependencies
It's also pretty straightforward to set up your own Conda package tree. Nice for packaging your app for deployment or making sure you have very precise dependencies.
I think deployment is a solved problem with docker. Its libraries like blas,etc that are a huge pain. I'm not sure why static linked bumpy is not possible - even anaconda could not achieve it.
If you've ever tried to dive into the NumPy build process you'd see why. It's unbelievably complicated... not that they really could do it better given that they are compiling about a billion scientific libraries and support alternatives and optimizations (like MKL).
Yes - unfortunately I have and I failed miserably.
These days I'm trying to see if there's a docker build that can build a great numpy (with all optimizations). Interestingly there are even docker images to call cuda APIs from python.
We have to use a mix of pypi and conda since quite a few of our dependencies are not in conda. We have a script which checks conda first, then falls back to pypi, all from one requirements.txt
