Referring to dplyr group by, carry forward value from previous group to next and Get the last row of a previous group in data.table on how to code a recursive calculation, I noticed that the order of evaluation is not sequential within j
Example using a simple sum of previous and current row to demonstrate the issue hence I am not asking about using shift:
library(data.table) #data.table_1.12.8 R-3.6.1 Win x64
DT <- data.table(rn=1:4, X=c(0,0,1,0), Y=c(0,0,0,0))
DT[, {
A <- X[1L]
B <- X[1L]
.SD[-1L,
{
print(.BY$rn)
print(c(X=X, A=A, B=B))
#expecting the next 4 lines to be evaluated in sequence but seems like B <- X is evaluated before these 2 lines
sA <- A + Y #line 1
sB <- B + Y #line 2
B <- X #line 3
A <- copy(X) #line 4
print(c(X=X, A=A, B=B))
.(sA, sB)
},
rn]
}]
output:
[1] 2
X A B
0 0 0
X A B
0 0 0
[1] 3
X A B
1 0 1<- this should be 0
X A B
1 1 1
[1] 4
X A B
0 1 0<- this should be 1
X A B
0 0 0
I am expecting the 4 lines to be evaluated in sequence but seems like B <- X is evaluated before the first 2 lines. Using copy solves the issue (i.e. A <- copy(X)). Should this be a bug or desired behaviour?
