Skip to content

replicate evaluates its second argument in wrong environment

4 messages · Gabor Grothendieck, David Winsemius, Hadley Wickham

Original
Gabor Grothendieck
# 
Currently replicate used within sapply within a function can fail
because it gets the environment for its second argument, which is
currently hard coded to be the parent frame, wrong.  See this link for
a full example of how it goes wrong and how it could be made to work
if it were possible to pass an envir argument to it.

https://stackoverflow.com/questions/66184446/sapplya-replicate-b-expression-no-longer-works-inside-a-function/66185079#66185079
2 days later
Hadley Wickham
# 
This is a nice example of the motivation for tidy evaluation ? since
enquo() captures the environment in which the promise should be
evaluated, there's no need for an additional explicit argument.

library(rlang)

replicate2 <- function (n, expr, simplify = "array") {
  exnr <- enquo(expr)
  sapply(integer(n), function(i) eval_tidy(expr), simplify = simplify)
}

doRep2 <- function(a, b) sapply(a, replicate2, b)
doRep2(3, 2)
#>      [,1]
#> [1,]    2
#> [2,]    2
#> [3,]    2

Hadley

On Sat, Feb 13, 2021 at 7:09 AM Gabor Grothendieck
<ggrothendieck at gmail.com> wrote:

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      David Winsemius
    

    

      
      #
    

  


  

      
On 2/15/21 1:10 PM, Hadley Wickham wrote:

            

      
      
      
    

        
It does not appear that the line above would accomplish anything given 
the succeeding line. Or am I missing something? Taking it out doesn't 
seem to affect results. Whatever magic there is seems to be in the 
`eval_tidy` function, whose mechanism or rules seem opaque. Was "exnr" 
supposed to be passed to `eval_tidy`?

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      Hadley Wickham
    

    

      
      #
    

  


  

      
On Monday, February 15, 2021, David Winsemius <dwinsemius at comcast.net>
wrote:

            

      
      
      
    

        
Oops, yes, obviously that was supposed to be expr. It doesn?t matter in
Gabor?s example because it evaluates to a constant but obviously would
matter in other cases.

Hadley