R dplyr purrr error when passing dataframe to pmap_dfr()

Question

I'm preparing a function to graph trigonometric functions. I have written a function which successfully returns the correct value when given the parameters for single value of theta, but when I attempt to pass a data frame with a column for each parameter using purrr::pmap_dfr() in order to produce a set of values to graph, I get an error whose message doesn't appear to match the situation.

Here a reproducible example:

thetas <- seq(from = -3, to = 3, by = 0.05) 
amplitudes <- rep(1, times = length(thetas))
period_denoms<- rep(1, times = length(thetas))
horizontal_shifts <- rep(0, times = length(thetas))
vertical_shifts <- rep(0, times = length(thetas))

argList <- list(thetas,
                amplitudes,
                period_denoms,
                horizontal_shifts,
                vertical_shifts) %>%
    as_tibble(.name_repair = ~ c("theta",
                                "amplitude",
                                "period_denom",
                                "horizontal_shift",
                                "vertical_shift")
                )
xformed_sine <- function(theta, amplitude, period_denom, horizontal_shift, vertical_shift){
        period <- (2 * pi)/abs(period_denom)
        
        return(as.double(amplitude * sinpi(period(theta + horizontal_shift)) + vertical_shift))
}

pmap_dfr(argList, xformed_sine)

R returns this error:

"Error in dplyr::bind_rows(): ! Argument 1 must be a data frame or a named atomic vector."

As I read the error, R is complaining that the object argList is not a data frame, yet when I run this code: is.data.frame(argList) the returned value is TRUE.

The return from that last bit doesn't match the error message, which tells me that argList isn't a data frame. Both can't be true, and I suspect that either I misunderstand the error message, or there is a bug in the pmap_dfr() function.

Any thoughts?

Answer 1

The error is referring to what your xformed_sine() function returns. You can make it a tibble and your code will run:

xformed_sine <- function(theta, amplitude, period_denom, horizontal_shift, vertical_shift){
  period <- (2 * pi)/abs(period_denom)
  
  return(
    tibble(result = amplitude * sinpi(period(theta + horizontal_shift)) + vertical_shift)
  )
}

pmap_dfr(argList, xformed_sine)
# A tibble: 121 × 1
   result             
   <Period>           
 1 0S                 
 2 -0.15643446504023S 
 3 -0.309016994374948S
 4 -0.453990499739547S
 5 -0.587785252292474S
 6 -0.707106781186547S
 7 -0.809016994374947S
 8 -0.891006524188368S
 9 -0.951056516295154S
10 -0.987688340595138S
# … with 111 more rows
# ℹ Use `print(n = ...)` to see more rows

Alternately, just keep your xformed_sine() as is and use pmap_dbl().

A little extra explanation: There's a hint that purrr is hitting an error after it completes the mapping function, as it fails when trying to bind rows on res (res = result of map operation) in rlang::last_error():

rlang::last_error()
<error/rlang_error>
Error in `dplyr::bind_rows()`:
! Argument 1 must be a data frame or a named atomic vector.
---
Backtrace:
 1. purrr::pmap_dfr(argList, xformed_sine)
 2. dplyr::bind_rows(res, .id = .id)

It is kind of an ambiguous error message though, I can see how it is confusing without any additional context.