Scope of .N in i

[claytontstanley]

I noticed that .N was made available in i recently, so I started using the feature. However, I got into trouble when trying to use .N in nested data tables. Here's an example:

require(data.table)
tbl = data.table(1:5)
tbl[1:.N]
#    V1
#1:  1
#2:  2
#3:  3
#4:  4
#5:  5

tbl[, head(.SD, n=2)[1:.N]]
#    V1
#1:  1
#2:  2
#3: NA
#4: NA
#5: NA

tbl[, head(.SD, n=2)[, 1:.N]]
# [1] 1 2

I was expecting .N in both i and j to bind to the most local data.table environment. In this case, the nested table, so .N=2.

I brought this issue up with Arun. He looked at the code and found that .N was locked to the top-level data.table. He thought this would be a good issue to bring to the community for discussion.

-Clayton

[eantonya]

Seems "obvious" to me that it should be equal to 2 in your second expression, instead of 5, so I'd vote yay for the change.

[jangorecki]

+1 for change

[franknarf1]

+1

[renkun-ken]

+1

[arunsrinivasan]

Thanks @claytonstanley for catching this. Will fix.

[rsaporta]

I like the change as well, but just to be sure all bases are covered, what about when toplevel .N is desired? (example that comes to mind might be when there is a by clause used in conjunction)

---

that being said, I do believe that the value of .N should match across i and j

[franknarf1]

@rsaporta It's just a scalar, so I'd manually make a copy if needed: DT[,{g.N=.N;....SD...},by=g]