[Request] Verbose message "Assigning to .N row subset of .N rows" seems wrong when x[i] is empty

[franknarf1]

I was looking at...

library(data.table)
DT = data.table(a = 1:10)

DT[a < 5, a := 5L, verbose=TRUE]
# Detected that j uses these columns: a 
# Assigning to 4 row subset of 10 rows

DT[a < 5, a := 5L, verbose=TRUE]
# Detected that j uses these columns: a 
# Assigning to 10 row subset of 10 rows

In the second instance, there is (I reckon) no assignment taking place, so it shouldn't look like all rows were modified. Similarly...

DT[0, d := 1, verbose=TRUE]
# Detected that j uses these columns: <none> 
# Assigning to 1 row subset of 10 rows

Here, I would expect to see numbers that better reflect what happened: d was initialized to NA on all rows. Ditto for DT[.(a=11L), on="a", f := 1, verbose=TRUE].

I'm interested in this because my colleagues are coming to R from Stata, which has this sort of feature (reporting counts of rows modified). I don't know if this should be characterized as a FR or bug.