将data.tables按2列合并，以表示可变序对

Question

Suppose I have the following data.tables:

X1 X2 val1
A  B  1
B  C  2
C  A  3

X1 X2 val2
A  B  100
C  B  200
A  C  300

where every combination of (X1, X2) appears once in each data.table, but the ordering is uncertain. I'm aiming for this output:

X1 X2 val1 val2
A  B  1    100
B  C  2    200
C  A  3    300

What's the most efficient way to do this? Especially if there's a 3rd data.table containing a 3rd value column with the same situation re: X1, X2, etc.

Answer 1

How about something like this?

special_join <- function(x, y, xcols, ycols=xcols) {
    ix1 = y[x, on=structure(xcols, names=ycols), which=TRUE]
    ix2 = y[x, on=structure(rev(xcols), names=ycols), which=TRUE]
    pmax(ix1, ix2, na.rm=TRUE)
}
ix = special_join(dt1, dt2, names(dt1)[1:2])
dt1[, val2 := dt2$val2[ix]]

---

where,

dt1 = fread('X1 X2 val1
A  B  1
B  C  2
C  A  3')

dt2 = fread('X1 X2 val2
A  B  100
C  B  200
A  C  300')

I'll leave the part on adapting this for your 3rd data.table as an exercise.