Fork me on GitHub
#data-science
<
2017-08-12
>
yogidevbear10:08:36

@elise_huard, I saw an email with a YouTube video link the other day for one of your talks, but the link doesn't work. Do you know if/why the video disappeared? Is there a new link for it? Thank you simple_smile

aaelony22:08:12

@elise_huard my question was not about classic gorilla-repl. Rather it was about the utility of the dockerized version you linked to, i.e. "pimping gorilla", due to the limitation that you cannot, to my knowledge, add dependencies there in the dockerized "pimped" version once you fire up the docker container. It's a show stopper for the Dockerized version if you can't bring in new libraries you need, hence the source and lein or boot seem to be a better option. A good notebook ideally should allow you to bring in new dependencies, so perhaps a boot option might be the way to pursue. @blueberry, I'm also a an emacs user and a big fan of org-mode but isn't the main advantage of a "notebook" that you can have a literate programming environment where you can not only repeat the analysis, but also see the input code and the output code without necessarily running anything apriori? Not only code but previous results and visualizations. In this sense, I think the browser wins. I think that the first step is to read a new notebook, and once read to proceed to re-running the code in the notebook, and finally editing it and re-running to test one's comprehension or to try it out on different data. no? per the emacs side, are you successfully syntax highlighting clojure code (via cider or clojure-mode) and retaining org-mode navigation and highlighting as well? A hybrid "clojure-org-mode" that does both might be nice if there isn't another way to do that.

blueberry22:08:45

What stops you from doing that in org-mode? In my opinion, browser is a rather poor editing environment. It clearly wins by being available everywhere, of course. But, once those analyses start being more complex and requiring a bit more complex setup of libraries, they also stop working, or are quite hard to set up. Emacs, (given that the user can use it, of course) is already there, and can do everything. When it comes to being able to play with analyses one can find in notebooks online, I think it looks much more useful as a demo, than as a practical tool. But, of course, I vote for freedom: everyone should use what they think is best for them 🙂

blueberry22:08:14

Regarding clojure + org-mode, everything works: coloring, editing, shortcuts, etc. No need for some additional magic on top of what's available with cider and org-babel

aaelony22:08:45

well, as you mentioned initially, the saved org-mode files don't have interpretable visualizations unless a png or pdf was saved separately. Other than this, I agree with all your points. The initial idea, though, was to try to assemble a listing of past "notebooks" done in clojure that represent analyses that new users can peruse and learn from. I think this is what made python successful for data science, that the "iPython Notebook" could be viewed and learned from by people that knew little else but how to use a browser. I myself prefer emacs for editing, but the reach of the browser being available everywhere in an instantly viewable form means that many more people can learn and create new analyses in such manner. no?