DT[, list(by_var), by = by_var)

[matthieugomez]

I'm not sure whether it is a bug or not ; I expected the following commands to return a data.table with the same number of rows than DT

DT <- data.table(id = c(1,1,2,2))
DT[, list(id), by=id]
DT[, list(id, .N), by=id]

[arunsrinivasan]

by arguments are by default available as length=1 variables in j. The kind of operation you perform is unnecessary, as they (by arguments) are returned by default. The idea is to be able to perform operations like:

DT[, list(sum(y)*x), by=x]

without having to do x[1L].

But there's a bug that needs to be fixed #497 which is a consequence of this feature.

[matthieugomez]

Thanks. I did no know that. I actually wanted to do DT[, list(id, .N), by = id]

[arunsrinivasan]

!Oops.. misunderstood!

Can you explain why you wanted to do that? It'd duplicate id column.

[matthieugomez]

Right - I was wrong for two reasons. the syntax is good as it is.