Submitted by: Damian Betebenner; Assigned to: Arun ; R-Forge link

Consider the following situation:

for (tmp.merge.variable in variables.to.merge) {

slot.data[tmp.data[,key(slot.data), with=FALSE], tmp.merge.variable := tmp.data[[tmp.merge.variable]], with=FALSE, mult="first"]

}

When the join of tmp.data[,key(slot.data),with=FALSE] to slot.data has some NA in the result we get the error (sometimes)

"Error during wrapup: Internal error: NA exist in 'rows' passed to C assign"

When with=TRUE and there are NAs it works ok, but not always when with=FALSE.

Still a problem in latest v1.8.11 :

X = data.table(A=1:3,B=1:6,key="A")
X
A B
1: 1 1
2: 1 4
3: 2 2
4: 2 5
5: 3 3
6: 3 6
X[J(2:5),B:=21L] # Ok. NB: values 4 and 5 join missing, but this is fine.
var = "B"
X[J(2:5),var:=22L,with=FALSE] # should work as well
Error in [.data.table(X, J(2:5), :=(var, 21L), with = FALSE) :
i[5] is NA. Can't assign by reference to row 'NA'.
X[J(2:5),(var):=22L] # but this is ok and good
X
A B
1: 1 1
2: 1 4
3: 2 22
4: 2 22
5: 3 22
6: 3 22