Fork me on GitHub

@hswick I’ll add that this covers the parts of gorilla-repl + huri that I was using them for (graphing), without the annoyances (notebook interface being subpar to just coding in emacs).


Awesome! I specifically designed it to not be coupled to any ide environment so it could fit in a variety of use cases.


call for comments


I'm happy to see neanderthal, but I think the author is a bit flippant about why people wouldn't install local libraries. Having worked at larger firms with more conservative IT departments, one is simply not allowed to just install things anywhere. I remember getting haskell 9 months after we asked for it, several versions out of date, all installed as root (which meant nobody was able to install any packages), and when we asked to remedy any of this, the answers ranged from "it'll take 6 months" to "no"


@c25l that is actually the most serious argument out there related to this topic. The conservative IT boss. However, it is a bit of a stretch from the main point, an this is "people starting out with ML and HPC". Anyway, to return to the issue of the conservative boss: how to install leiningen in such a company? I assume Java is already there. Or, to be more extreme, when it comes to machine learning, how do one install TensorFlow (CUDA + CUDNN installation) there? I agree: this is a huge issue, but isn't it much more general than the things we are dealing with here?


OTOH, I would not count on conservative IT departments to go anywhere with machine learning anyway 🙂


You'd be surprised. Insurance is a good example for this, extremely forward-looking R&D using all the techniques possible, but IT with a more 80s-90s ethos. Leiningen can run in user-space just fine, in my experience, so long as you have a jvm. As for Tensorflow, you can anaconda it into userspace without cuda, with the rest on a few dedicated machines with gpus. It's no different of an issue in genral, I'd think. I just find that the "how are you incapable of this as a person?" mentality to be unnecessarily ignorant of the legal constraints in a big business setting, and it's unfortunate to see it so widespread.


@c25l ah, there seems to be misunderstanding, at least when it comes to Neanderthal's need for MKL. MKL does offer a GUI installer, which I assume is the easiest way for the person who would like to start learning this on their personal computer without much fuss. However, MKL is just a bunch of *.so files. You absolutely do not need to install it. You can package them into a jar, or an uberjar, or a rpm or even a folder that gets added to a path by a bash script, or whatever best suits your need. But then, you are someone who knows what is the best option. The point is that you are already supported there, but you are not a beginner, by far.


Yeah, i see where you're coming from there.


what is the standard best practice for IO on large/medium tab-separated files? would like to do this in clojure'esk way


probably lazy eval like java’s iterator pattern