Shiny: when shiny is embedded inside rmarkdown

[MelissaLe44]

I try to include interactive plots in my rmarkdown.
My dataset (mydata) comes from a .rdata that I load into a r chunk global
In another r chunk, I got my selectinput and my renderPlot that is based on plotting a subset of mydata depending on selectinput.
It does not work because it does not recognize mydata. However if instead I use renderPrint to compute the length of subset of mydata, this works! so it seems it is only a problem with renderPlot!! Using ggplot, I specify environnment=environnment() but this is not working!!
Any advice?

Of note, I put this issue in rmarkdown which says that it was more a shiny problem (and advice me to ask my problem here), but if I do an independant shiny app using load, it works, so the problem is according to me in rmarkdown!

Thanks,
Best
Mélissa

[mschilli87]

@MelissaLe44: As pointed out in the issue you mentioned, markdown != rmarkdown.
I cannot judge which package is 'to blame' for your trouble, but in case you don't get help here, try over there. ;)

[wch]

Can you provide a simple reproducible example?

[MelissaLe44]

1st step create a mydata and save it into a rdata

mydata<-data.table(A=c(rep("a",4),rep("b",3)),val=c(1:7))
save(mydata,"data.rdata")

2nd step have a rmarkdown template and put some rchunk

first r chunk code (to load the data)

load("data.rdata")

second r chunk code:

selectInput("selectinput", "",c("a","b"))
plotOutput("dist")

third r chunk: define dist as a renderplot

output$dist<-renderPlot({
chosen<-input$selectinput
p<-ggplot(mydata[A %in% chosen],aes(x=val,y=val))+geom_point()
})

When running the rmd, it does not work, error can't find mydata

[wch]

Fixed minimal example:

````

output: html_document

runtime: shiny

```{r}
library(data.table)
library(ggplot2)
mydata <- data.table(A=c("a", "a", "b", "c", "c"), val=c(1:5))

renderPlot({
ggplot(mydata[A %in% c("a", "c")], aes(x=val,y=val)) + geom_point()
})
```
````

It displays this:

[MelissaLe44]

Thanks a lot for your help in reproducing this example. It does not have access to mydata ?

[wch]

@MelissaLe44 There appears to be a problem with using the special data.table syntax inside of renderPlot when used with rmarkdown and runtime:shiny. The data.table syntax works fine with renderPrint in rmarkdown, and it works fine for a regular Shiny app.

For now I would suggest using the normal data frame subsetting syntax, like this:

mydata[mydata$A %in% c("a", "c"), ]

---

Example of a regular Shiny app that works with data.table:

library(shiny)
library(data.table)
mydata <- data.table(A=c("a", "a", "b", "c", "c"), val=c(1:5))

shinyApp(
  fluidPage(
    plotOutput("dist")
  ),
  function(input, output) {    
    output$dist<-renderPlot({
      chosen<-input$selectinput
      data <- mydata[A %in% c("a", "c")]
      plot(data$val, data$val)
    })
  }
)

[MelissaLe44]

Even while using a data frame it does not work in rmarkdown (I just converted my data table into a data frame)
I fully agree with you, inside a shiny app, it works (I have tried as well).

[wch]

Did you use the data frame syntax that I gave an example of?

[MelissaLe44]

yes!
I tried also to explicitily convert my data table into data frame
df1<-as.data.frame(mydata) (or I remove the as), and this outside the renderPlot but still does not work.

[wch]

You'll have to supply a reproducible example, because that code works for me.

[MelissaLe44]

Within a shiny app, there is no problem, but the same code as your "fixed minimal example" does still not work in rmarkdown, even by using
renderPlot({
ggplot(mydata[A %in% c("a", "c"), ], aes(x=val,y=val)) + geom_point()
})

[wch]

That's not the code I provided. Here it is again.

mydata[mydata$A %in% c("a", "c"), ]

[MelissaLe44]

Indeed it works! thanks a lot for your patience!!

[mschilli87]

@wch: Shouldn't this still be considered a bug? Why would some render-function handle DataTable-syntax but other's not?

[wch]

@mschilli87 yes, see the rmarkdown issue I filed that references this one: rstudio/rmarkdown#1162

[yihui]

I have fixed it in https://github.com/rstudio/rmarkdown/pull/1165.