Feature Request: `tstrsplit(... keep)` additional options

[JBrownArcGen]

The tstrsplit function has improved efficiency in a number of transform operations I have implemented and the keep parameter in particular has streamlined the process of getting string components into fields of the data.table.

At the moment keep can be applied by index which allows components to be accessed based on an anchor-point at the start of the string being split. It would be very useful to be able to anchor the keep to the end of the string as well because sometimes you won't know how many components are created, you only know you want the last one or last but one, or similar.

This outcome can be achieved at the moment using the standard strsplit, rev and lapply functions, combined with data.table::transpose, but is somewhat awkward to use in some situations:

x <- c("ABC-DEF", "ABC-DEF-GHI", "ABC-DEF-GHI-JKL", "ABC-DEF-GHI-JKL-MNO")
data.table::transpose(lapply(strsplit(x, "-", fixed = TRUE), rev))[1:3]

> [[1]]
> [1] "DEF" "GHI" "JKL" "MNO"
> 
> [[2]]
> [1] "ABC" "DEF" "GHI" "JKL"
> 
> [[3]]
> [1] NA    "ABC" "DEF" "GHI"

It would be valuable to be able to implement this "reversed" outcome natively as a parameter to tstrsplit itself, so something like:

data.table::tstrsplit(x, "-", fixed = TRUE, keep = 1:3, rev = TRUE)

> [[1]]
> [1] "DEF" "GHI" "JKL" "MNO"
> 
> [[2]]
> [1] "ABC" "DEF" "GHI" "JKL"
> 
> [[3]]
> [1] NA   "ABC" "DEF" "GHI"

[Mukulyadav2004]

Hi @MichaelChirico @tdhock ,
It seems like a valuable addition to 'tstrsplit()' when total number of splits() are unknown and there is requirement of last elements.
If you think its worth implementing, can I start implementing it by first defining reverse splitting parameter and when we want to anchor the 'keep' to the end of string, we'll set it to true and then reverse the split results before applying 'keep' indexing.?