.NROW as a way to access the nrow after subsetting in i but before entering a subgroup

[zerweck]

I use data.table a lot to calculate absolute frequencies on the fly with, while also applying a subsetting flag in i in the same call.

iris[Sepal.Width<3.7,.N,.(Species)]

If i want to change this to relative frequencies within the subset, I need to save the number of rows after filtering as denominator in another call, like this:

nrow_after_filtering <- iris[Sepal.Width<3.7,.N]
iris[Sepal.Width<3.7,.(rel_freq=.N/nrow_after_filtering) ,.(Species)]

This gets old after a while when doing this a lot.
It would be great if a symbol .NROW similar to .N existed, returning the nrow of the data.table after applying i, but before going into the subgroup.

The name .NROW is probably bad since it would be unclear what the difference to .N is, maybe .NI or .NFILT?

[ben-schwen]

Your intended behaviour can simply be achieved by chaining
iris[Sepal.Width<3.7, .N, .(Species)][, rel_freq := N/sum(N)][]

[ColeMiller1]

Related: #1206

I like the concept but I always think of @MichaelChirico's comment from the thread as I sometimes wonder if there are too many symbols already, some of which are only triggered via NSE.

Not sure when symbol overload kicks in...

Finally, here was my code golf.

library(data.table)
as.data.table(iris)[Sepal.Width < 3.7,
                    {
                      N = .N
                      .SD[, .(rel_freq = .N / N), Species]
                    }
]

[zerweck]

Related: #1206

As usual when I post something here, a better issue exists already that I somehow didn't find when searching :)

Also thanks to both of you @ben-schwen and @ColeMiller1 for the suggestions, I have used chaining in the past but somehow forgot about it again.