Fork me on GitHub
#data-science
<
2019-02-15
>
David Pham10:02:45

@jsa-aerial @metasoarous Thanks a lot for your answers! So you mean that vega(-lite) creates a description of your visualization in a data format? It is then similar to plotly where you describe your visualization through a json object/nested maps?

David Pham10:02:23

I am interested a lot into using one or the other library for doing visualization, but I am not sure I can convince my peers about the advantages over a simple wrapper around plotly (or react-plotly).

David Pham10:02:27

(in cljs at least)

Daniel Slutsky12:02:48

Some important discussion is going on in Clojureverse about learning data-science: https://clojureverse.org/t/learning-data-science-with-clojure/3697 Opinions would be helpful!

jsa-aerial15:02:53

@neo2551 If plotly works for you, go for it. For myself it didn't seem to have a set of foundational principles and separation of concerns. So, it always seemed (oddly) easier but more complicated. Not sure why you need a wrapper for it if it is really just all data all the time. Robin Linacre, a UK data scientist posted a short doc on why VG/VGL on Vega Slack for feedback. When finished he posted it: https://medium.com/@robin.linacre/why-im-backing-vega-lite-as-our-default-tool-for-data-visualisation-51c20970df39. It's a short, pretty nice description of how I look at it as well. YMMV.

👍 15
David Pham15:02:33

What do you mean by separation of concern?

David Pham15:02:49

Thanks a lot for the reading 🙂

metasoarous17:02:28

I haven't used plotly at all, so I can't really speak to it, but Vega-Lite/Vega give you this sense of things having been extremely well thought out and composable.

☝️ 5
metasoarous17:02:53

The post @jsa-aerial points to is a great summary, as far as I can tell.

metasoarous17:02:22

One of the things I'll highlight there that may help tip the boat for you is Voyager: https://vega.github.io/voyager2

metasoarous17:02:50

I haven't used Tableau, but I'm told by folks who have that it is very similar. But unlike Tableau, which sort of locks you in, when you create a plot in Voyager, it will spit out the Vega-Lite/Vega for you, so you can take that an run with it in any way you like.

metasoarous17:02:08

I haven't heard of anything like that for plotly

jsa-aerial18:02:27

Also, wrt composable, I haven't seen (doesn't mean it doesn't exist) any inter-chart/plot interactivity in plotly. In Vega(Lite) views, layers, selections and transforms are all well thought out separate things that can be consistently combined in all manner of ways to create very impressive cross chart/pot interactive visualizations. You don't really need a special extra thing like Dash, dashboards just fall out 'for free'.

✔️ 10