[R-Forge #5289] .I should always reflect the original row local

[arunsrinivasan]

Submitted by: Michele Carriero; Assigned to: Nobody; R-Forge link

Please consider the following:

dt <- data.table(a=sample(letters, 100, T), b=rnorm(100))
dt[ a=="c", list(.N, .I)]
   N .I
1: 4  1
2: 4  2
3: 4  3
4: 4  4

dt[a=="c", list(.N, .I), by=a]
   a N .I
1: c 4 54
2: c 4 67
3: c 4 71
4: c 4 86

The first statement subsets the table first and then calculates locations. The second shows the original location of all rows having a=="c".

Unless there are reasons (which I'm currently missing) for which the above needs to happen, I think that both expressions should give the second result.

Regards,
Michele.

[jangorecki]

closing as duplicate of #2598