Introducing Radiant - an interface for business analytics

495 views
Skip to first unread message

Vincent

unread,
Mar 21, 2015, 4:02:48 AM3/21/15
to shiny-...@googlegroups.com

Radiant is an open-source interface for business analytics in R, based on the Shiny package.

Radiant is available on CRAN. To install the latest version with complete documentation for offline access, open R(studio) and copy-and-paste the commands below:

options(repos = 'http://vnijs.github.io/radiant_miniCRAN/')
install.packages("radiant")

Once all packages are installed use the commands below to launch the app:

library(radiant)
radiant("marketing")

For more information see: http://vnijs.github.io/radiant/news.html







ZJ

unread,
Mar 23, 2015, 1:47:23 AM3/23/15
to shiny-...@googlegroups.com
How awesome is this!!!

I will learn from many parts of the code for my own projects looks.

Vincent

unread,
Mar 23, 2015, 2:23:18 PM3/23/15
to shiny-...@googlegroups.com
Thanks ZJ. Comments and suggestions welcome.

Ben Hunter

unread,
Mar 24, 2015, 11:34:59 AM3/24/15
to shiny-...@googlegroups.com
Hi Vincent,

I got interested in Radiant quite a while back and have made an app using the basic outline of it that was put together by someone else. I think it's great and really appreciate it. I feel like I have a good grasp of how it works, but I'm having troubles with plots disappearing after switching views. My javascript plots get pushed over and the lattice plots in the 'predict' tab disappear when you leave the "page" and then come back. 

As a suggestion, as cool as Radiant is, I think the community would really benefit from a library that provides the shell of a multi page shiny app.

No big deal, of course, but if you have a hunch on what I'm missing, I'd really appreciate it.

This is the link to the google groups post with a few more details:

Vincent

unread,
Mar 24, 2015, 1:50:31 PM3/24/15
to shiny-...@googlegroups.com
Hi Ben,

Nice app! I haven't noticed any of the issues you mentioned. I assume this is because I don't currently use any javascript plots in my app. 

I assume the app is open-source? If so, could you point me to the code and I will take a look.

Vincent

Ben Hunter

unread,
Mar 24, 2015, 3:11:22 PM3/24/15
to shiny-...@googlegroups.com
Hey thanks Vincent. I'm more than happy to post the code - the repo is kinda big because there are some large data files being used and some in the history, too.

I'm surprised that the bugs aren't showing up. Here are the steps after loading the page.

1. Click 'predict'
2. Click any of the 'period' tabs, 1, 3, or 3
3. Observer the ROC plot, add and remove models
4. Go to 'explore' or 'models'
5. Upon returning to 'predict' and then the period tabs, the ROC plots don't appear.

I thought you might have insight into what's wrong with the lattice plots, but I want to hit up the RStudio guys or HTMLWidget folk about the JS problems.

There's a lot of code related to page state in the radiant repo that I'm not using and don't fully understand, could that be it?

Ben

Vincent

unread,
Mar 25, 2015, 1:36:51 AM3/25/15
to shiny-...@googlegroups.com
I just meant I had not seen these types of issues in Radiant. It seems I spoke too soon. When zooming in/out in chrome some plots disappear (e.g., the data > visualize tab). Refresh gets them back but that shouldn't happen. It doesn't happen with all plots. Just those where grid.arrange or par is used for multiple plots. I am trying to track this issue down.

I did now notice that some of your latice plots disappear when changing tabs. It doesn't happen consistently however. I have not seen this at all in Radiant but I doubt that has anything to do with the state code.

No idea about the layout changes for the js plots.

If you can reproduce the issues in a simpler (single-page) app that will be helpful for the Shiny developer if it turns out that they do need to take a look at this.


Vincent

Vincent

unread,
Mar 26, 2015, 8:41:26 PM3/26/15
to shiny-...@googlegroups.com
Ben,

FYI I just filled a shiny issue on github related to plots disappearing when you zoom in/out in chrome or firefox. I don't think this related to your issue however.

Vincent

unread,
Mar 29, 2015, 5:19:41 PM3/29/15
to shiny-...@googlegroups.com
There is a fix/workaround for the plots disappearing when zooming in/out: https://github.com/rstudio/shiny/issues/781

Ben Hunter

unread,
Apr 1, 2015, 9:07:27 PM4/1/15
to shiny-...@googlegroups.com
Vincent, 

I know nothing of writing R packages and I'm going through Hadley's book on it. Would you mind if I make a simple package that creates a directory and handles the boilerplate required to do something like Radiant? I figure I'd call it "radshiny" - it's got an "R", references your project and Shiny. Plus "rad" is a funny word.

-Ben

Ben Hunter

unread,
Apr 2, 2015, 1:52:05 PM4/2/15
to shiny-...@googlegroups.com
Anyway, here it is:

https://github.com/benjh33/radshiny

If anyone wants to pipe in with some ideas to extend it, I'd be happy to collaborate or just do it myself.

Greg L

unread,
Apr 7, 2015, 7:38:26 PM4/7/15
to shiny-...@googlegroups.com
hey, I took a look at your app for a bit. Didn't read any of your code but I noticed a few things.
your 3 plots in Models have fixed widths and have a hard time fitting on one row for many resolutions. 1680x1050 here, and I only see 2 plots aligned properly from the beginning. I still get the problem of 1 plot on top and 2 on bottom when i switch tabs, but these might be related. Try dynamic widths on your plots or scaling them down a bit or extending the width of whatever container they're in.
second, switching tabs makes things real funky. the input seems to change randomly for explore and predict every time you switch tabs. the 3 teams in explore change and the scatter matrix variables in predict change too. do some debugging like with print statements or something and make sure each time you switch tabs, your 3 inner plots in predict are actually trying to plot what you intend them to plot. it's likely not showing up because whatever happens when you switch tabs causes it to plot some invalid data.
Reply all
Reply to author
Forward
0 new messages