search

Coorsaa / shinyMlr

shiny-mlr: Integration of the mlr package into shiny
Other
93 stars 20 forks source link

switch plots from ggplot to ggvis? #39

Open Coorsaa opened 7 years ago

Coorsaa commented 7 years ago

ggvis has a similar structure to ggplot but produces interactive plots.

jakob-r commented 7 years ago

What are the advantages over using plotly? For plotly you just have to change plotOutput to plotlyOutput and renderPlot to renderPlotly. It works flawless if renderPlot returns a ggplot.

Where would you want to have interactive plots the most?

Coorsaa commented 7 years ago

Thanks @jakob-r , I was hoping for that kind of input. After briefly looking into plotly, it seems that it would work pretty well for us. I was thinking of starting with it for the data summary plots with possible extensions for the prediction and benchmark plots. What do u think @florianfendt ?

larskotthoff commented 7 years ago

AFAIK plotly plots can't be embedded as nicely (i.e. they're hosted on plotly's website). This may be something to consider.

SteveBronder commented 7 years ago

What about Highcharter?

Haven't used ggvis in a while, but I remember it's development was slowed due to vega missing some pieces. That was a year ago so it may be much better now.

Coorsaa commented 7 years ago

So Highcharter doesn't have plotly's embedding disadvantage? It seems to be very flexible and it has a broad variety of different plot types.

SteveBronder commented 7 years ago

I'm a big fan of highcharts! Though I've never used this particular package, I do not see why you would have to embed it like you do with plotly. It should output the html needed to integrate with shiny. I believe it's backend is partly based on htmlwidgets which is supposed to integrate with shiny very well.

Coorsaa commented 7 years ago

ok sounds great! Thanks for your support and ideas!

larskotthoff commented 7 years ago

Hmm, from the htmlwidgets page that you link to it looks like you can actually embed plotly plots (they do it there). The big advantage of plotly is that you can export and save the plots.

jakobbossek commented 7 years ago

@larskotthoff Right, you can export and save. However, exporting to PDF format is only possible if you pay money. E.g., https://plot.ly/~chris/1638/ (interactive version) https://plot.ly/~chris/1638.png (possible) https://plot.ly/~chris/1638.pdf (payment required ...)

larskotthoff commented 7 years ago

Ok, but at least it gives you that option. Is there an easy way in shiny to make plots available as downloadable PDFs?

jakob-r commented 7 years ago

AFAIK plotly plots can't be embedded as nicely (i.e. they're hosted on plotly's website). This may be something to consider.

No. Plotly works totally fine offline. It just relies on JavaScript. You can publish but this is just an additional feature. Try the examples here. It uses htmlwidgets as well as highcharts and so on.

Is there an easy way in shiny to make plots available as downloadable PDFs?

Just a quick search result: http://shiny.rstudio.com/gallery/download-knitr-reports.html This obviously would not work directly with things that use htmlwidgets.

florianfendt commented 7 years ago

@Coorsaa plotly really looks great but if we can't export pdfs easily I'd vote for staying with ggplot for now and at least make pdf download available soon. If we have time later we can try make the plots interactive. What do you reckon?

jakob-r commented 7 years ago

I worked on this today a bit and if you go with the approach in the link I provided it is easy to do ggplot and plotly together. The major disadvantage is that you need a separate Rmd file that serves as a template like for a report to generate a pdf. You can also extract the intermediate markdown files of the pdf generation and offer a zip download of the single plots.

This leads me to think that with some cleverness you call ggsave on all ggplot objects and than download them.