Fork me on GitHub

@metasoarous Yes, I am finding the parameterized template system to be much more expressive than the approach used in Altair (Python VGL - and defacto 'standard' PL binding) with many added benefits - such as building libs of templates for domain specific visualizations. Completely user extensible and with no API - it's all data transformation. Hanami is what you use to build various (typically domain specific) visualization applications. Saite is one such example application that is for exploratory and ad hoc visualizations (basically a supercharged variation of what is in oz). Aeroviews will be another which is shaping up to be a many feature fully cross view interactive visualization application for RNA-Seq, Tn-Seq, Term-Seq, and WholeGenome-Seq (differential gene expression and fitness exploration).


Saite makes use of the framework aspect of Hanami (completely optional, but can be very useful) by exploiting the tab system for many simultaneous visualization 'pages' as well as automated grid layouts of separate visualizations per page. Also, it uses the session capabilities so you can send different visualization sets to multiple viewers. The simple, but expressive, websocket protocol (Hanasu) used also eliminates the complected and heavy weight Sente stuff. And it uses current, non deprecated, React components and protocols.


@jsa-aerial Very cool stuff. I'm excited to play around with Hanami, and have added a note about it in the Oz README:


Very cool to hear about Aeroviews as well! And in general that you're getting back into compbio (IIRC you did this some years ago?).


@metasoarous Nice. I've been in comp bio for many years now having come from previous ML stuff (mostly NLP)


I've also been using Vega/Vega-Lite for bio applications as well (B-cell deep sequencing analysis tools), though sadly not with Clojure, but in vanilla React.


WRT Hanami - Saite is the more relevant thing, since it is a self contained application which is built with Hanami. It is an exploratory graphics and ad hoc vis app


Yes, Hanami is built on reagent and re-com enabled


I am sending you a comment on the updates - basically, you must make sure that you have new externs; tooltips work fine for me


Oh another thing - you can't have a single PR for this - the dependencies will cause it to fail


Cool; thanks. Presumably its something in the oz code that is keeping the tooltips working. Not sure why one PR wouldn't be ok though; I've done it that way on cljsjs in the past. Has something changed?


Regarding Hanami: that clarifies things; thanks. I meant that the templating functionality could potentially be useful in either context. But good to know that it has it's own Reagent/etc logic.


Also, I was interested to see your websocket lib.


I've got a few bones to pick with Sente, and overall liked what I saw in Hanasu. In particular that the API seems more clj/cljs symmetric (would you agree with that statement?).


As for Oz versus Saite, I think we have somewhat different focuses. Saite seems useful in that you can a bit more efficiently update one viz at a time, and don't have to think as much about how to organize all the individual viz components. Oz approaches this space by letting you compose visualizations using hiccup, which gives you something a little more akin to a more general interactive notebook/document generation tool.


So I can see people using one or the other for different reasons.


Super exciting to see all the energy in this area and the emergence of a robust Clojure+Vega ecosystem for data-science & dataviz.