Would it be beneficial for na_if() to accept NaN as a y argument?

[hlynurhallgrims]

I got to thinking. I was using tidyr::fill(), expecting it to replace some NaNs in the same way as it does for NAs. Seeing that it doesn't and understanding that it's not desireable for it to do so I wanted to raise the question if dplyr::na_if() should perhaps be made a more convenient alternative to dplyr::if_else() when you want to explicitly turn NaNs to NAs.

In the cases where you'd want to carry forward the last observation before a NaN you could use na_if(x, y = NaN) before using tidyr::fill().

A reprex follows:

vector <- c(1:5, rep(NaN, 5))
vector
#>  [1]   1   2   3   4   5 NaN NaN NaN NaN NaN

#This changes nothing
dplyr::na_if(x = vector, y = NaN)
#>  [1]   1   2   3   4   5 NaN NaN NaN NaN NaN

# Perhaps this is desireable? Perhaps not?
na_if <- function(x, y) {
# check_length() function left out for convenience here

if (is.nan(y)) {
  x[is.nan(x)] <- NA
} else {
  x[x == y] <- NA
}
x
}

# Our updated na_if function:
na_if(x = vector, y = NaN)
#>  [1]  1  2  3  4  5 NA NA NA NA NA

Created on 2019-11-22 by the reprex package (v0.2.0).

[hadley]

Minimal reprex:

dplyr::na_if(c(NA, NaN, 1), 1)
#> [1]  NA NaN  NA

^{Created on 2019-12-10 by the reprex package (v0.3.0)}

[hadley]

Moving this issue to funs, which is where the fix will eventually happen.

[lock]

This old issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with reprex) and link to this issue. https://reprex.tidyverse.org/

[lock]

This old issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with reprex) and link to this issue. https://reprex.tidyverse.org/