data.table::fcase with vectorized default

[r2evans]

https://markfairbanks.github.io/tidytable/articles/speed_comparisons.html correctly states that data.table::fcase(..., default) must be length 1, but you can work around this using a catch-all rep(TRUE, .N):

as.data.table(mtcars)[, a := fcase(
  cyl == 4     , "_4",
  rep(TRUE, .N), as.character(cyl)
)][1:3,]
#     mpg cyl disp  hp drat    wt  qsec vs am gear carb  a
# 1: 21.0   6  160 110 3.90 2.620 16.46  0  1    4    4  6
# 2: 21.0   6  160 110 3.90 2.875 17.02  0  1    4    4  6
# 3: 22.8   4  108  93 3.85 2.320 18.61  1  1    4    1 _4

(A similar approach was mentioned several months ago in Rdatatable/data.table#4258 (comment).)

It's a little hack-y, but if you feel your speed comparison would be a better representation with vectorized defaults, then this is an option. Or perhaps as an additional comparison, demonstrating why perhaps having vectorized defaults (or non-vectorized "when" values) could be a good thing for data.table.

(More of a suggestion than an issue, feel free to close.)

[markfairbanks]

you can work around this using a catch-all

Hmm it might be a good idea to change the speed tests to use this. I'll think about it a bit.

As far as vectorized defaults, I was a little surprised something like this wasn't possible using fcase():

pacman::p_load(tidytable)

test_df <- data.table(x = 1:5)

test_df[, new_x := case.(x <= 3, 1, default = x)][]
#>    x new_x
#> 1: 1     1
#> 2: 2     1
#> 3: 3     1
#> 4: 4     4
#> 5: 5     5

This one could be covered by fifelse(), but it seems like you should be able to do a default of "leave the value alone". Oh well.

Either way, thanks for pointing out this workaround 👍

[r2evans]

The data.table issue I linked to (it's still open) is discussing that at length, and it is on the "Master list of most-requested issues" (Rdatatable/data.table#3189). I agree, though my expectation of safe recycling is different than R's default behavior: I'd expect length .N or 1, nothing else (not .N/2, for instance).