rlang interferes with dcast ...~metric API

[l-jaye]

Hello!

I am not sure if this is a Shiny issue, or a data.table issue, but a recent update to either package causes an error Error in check_formula: '...' used in an incorrect context when using ... in a formula for dcast.data.table() inside a shiny app. Outside of the shiny app, no issue. Please see a reproducible example below:

library(data.table)
library(shiny)

shinyApp(
  
  ui = fluidPage(
    actionButton("do.btn","Go!")
  ),
  
  server = function(input, output){
    
    observeEvent(input$do.btn, {
      
      metrics = c("col.x","col.y")
      dt = data.table(x = 1:10, y = "test", col.x = 3, col.y = 4)
      melt.dt = melt(dt, measure.vars = metrics, variable.name = "metric")
      dt = dcast(melt.dt, ... ~ metric, value.var = "value", fill = NA) # this throws error
      
    })
    
  }
)

[l-jaye]

Update:

Replacing ... ~ metric with as.formula("...~metric") works:

library(data.table)
library(shiny)

shinyApp(
  
  ui = fluidPage(
    actionButton("do.btn","Go!")
  ),
  
  server = function(input, output){
    
    observeEvent(input$do.btn, {
      
      metrics = c("col.x","col.y")
      dt = data.table(x = 1:10, y = "test", col.x = 3, col.y = 4)
      melt.dt = melt(dt, measure.vars = metrics, variable.name = "metric")
      
      dt = dcast(melt.dt, as.formula("...~metric"), value.var = "value", fill = NA)
      print(dt)
      
    })
    
  }
)

[avimallu]

From the linked issue, it looks like this is not something that can be fixed on data.table's end, correct? The comment indicates that it's an interaction between rlang and data.table. Might want to close if my interpretation is accurate.

[ColeMiller1]

I would leave it open in case we would want to provide an additional API as a workaround. Maybe .SD ~ metric which would not interfere with rlang - I know it took me awhile to remember that all variables was ....

Also note, it's not interaction between rlang and data.table. It's rlang and base R, it's just that data.table NSE happens to use the .... Thanks to the shiny maintainers for helping isolate the issue.

rlang::eval_tidy(...~x)
## Error in rlang::eval_tidy(... ~ x) : '...' used in an incorrect context

As far as workarounds, quoting just the formula works. E.g.

dt = dcast(melt.dt, "...~metric", value.var = "value", fill = NA)

[lionel-]

This is a bug in rlang that we will fix on our side.

[MichaelChirico]

Thanks Lionel! Closing this side then.