step_holiday fails with NA values

[MrFlick]

The problem

The step_holiday function seems to return errors during baking/juicing in the presence of NA values. For example

Reproducible example

library(recipes)

dd <- data.frame(
  date = as.Date(c("2021-12-04", "2021-12-25", NA, "2021-12-27"))
)

recipe(~date, data=dd) %>% 
  step_holiday(date) %>% 
  prep() %>% 
  juice()

# Error in if (rng.nch[1] != rng.nch[2]) stop("'charvec' has non-NA entries of different number of characters") : 
#   missing value where TRUE/FALSE needed

Tested with recipes_0.1.16

Possible solution

The problem seems to be with the is_holiday function. It's passing a unique vector of years to the timeDate::holiday() function but that vector can include NA values which end up throwing the error the timeDate::holiday() code. It would be better to omit NA years (possibly checking to make sure there is at least one NA-year). It probably also makes sense to propagate the NA values in the returned indicator columns. One possible solution is

is_holiday <- function(hol, dt) {
  # ~~ add na.omit() here to drop NA values ~~
  hdate <- holiday(year = unique(year(na.omit(dt))), Holiday = hol)
  hdate <- as.Date(hdate)
  out <- rep(0, length(dt))
  out[dt %in% hdate] <- 1
  # ~~ carry forward missing values ~~
  out[is.na(dt)] <- NA
  out
}

This updated version of the function would return

# A tibble: 4 x 4
  date       date_LaborDay date_NewYearsDay date_ChristmasDay
  <date>             <dbl>            <dbl>             <dbl>
1 2021-12-04             0                0                 0
2 2021-12-25             0                0                 1
3 NA                    NA               NA                NA
4 2021-12-27             0                0                 0

If that seems like a reasonable solution, I could prepare a pull request.

[github-actions]

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