Aggregate a data frame based on unordered pairs of columns

One way is to create extra columns with pmax and pmin of id1 and id2as follows. I'll use data.table solution here.

require(data.table)
DT <- data.table(DF)
# Following mnel's suggestion, g1, g2 could be used directly in by
# and it could be even shortened by using `id1` and id2` as their names
DT.OUT <- DT[, list(size=sum(size)), 
        by=list(id1 = pmin(id1, id2), id2 = pmax(id1, id2))]
#     id1  id2 size
# 1: 5400 5505   18
# 2: 5033 5458    1
# 3: 5452 2873   24
# 4: 5452 5213    2
# 5: 5452 4242   26
# 6: 4823 4823    4

an alternate method:

R> library(igraph)
R> DF
   id1  id2 size
1 5400 5505    7
2 5033 5458    1
3 5452 2873   24
4 5452 5213    2
5 5452 4242   26
6 4823 4823    4
7 5505 5400   11
R> g  <- graph.data.frame(DF, directed=F)
R> g  <- simplify(g, edge.attr.comb="sum", remove.loops=FALSE)
R> DF <- get.data.frame(g)
R> DF
   id1  id2 size
1 5400 5505   18
2 5033 5458    1
3 5452 2873   24
4 5452 5213    2
5 5452 4242   26
6 4823 4823    4