creating pairs for network analysis

From my dataset I'm trying to make pairs based on a ranking. my data looks like

ID     grp      rank
1      grp1       1
1      grp2       1
1      grp3       2
2      grp1       1
2      grp2       2
2      grp2       2
2      grp2       2
2      grp3       2
2      grp1       3

The output I am aiming for is the following: for each ID

if rank = 1 then grp in source and destination are the same = grp
If rank is different from 1 then
- source = take grp from previous rank
- destination = take group from current rank If more then one group exist for the same ranking then an additional line needs to be created so that each pair is represented.

This looks then like the following

ID     rank     source  destination
 1       1       grp1        grp1
 1       1       grp2        grp2
 1       2       grp1        grp3
 1       2       grp2        grp3
 2       1       grp1        grp1
 2       2       grp1        grp2 
 2       2       grp1        grp2
 2       2       grp1        grp2
 2       2       grp1        grp3
 2       3       grp2        grp1
 2       3       grp3        grp1

I started with a for loop and if_else statements but I got stuck. Any help is appreciated! thx in advance.

Solution

We may do the following:

df %>% group_by(ID) %>% 
  do(map_dfr(1:nrow(.), function(i)
    data.frame(.[i, -2], source = if(.$rank[i] == 1) .$grp[i] else unique(.$grp[.$rank == .$rank[i] - 1]), 
               destination = .$grp[i])))
# A tibble: 11 x 4
# Groups:   ID [2]
#       ID  rank source destination
#    <int> <int> <fct>  <fct>      
#  1     1     1 grp1   grp1       
#  2     1     1 grp2   grp2       
#  3     1     2 grp1   grp3       
#  4     1     2 grp2   grp3       
#  5     2     1 grp1   grp1       
#  6     2     2 grp1   grp2       
#  7     2     2 grp1   grp2       
#  8     2     2 grp1   grp2       
#  9     2     2 grp1   grp3       
# 10     2     3 grp2   grp1       
# 11     2     3 grp3   grp1

We group by ID and then go over each row a given group. Then for each row we construct a new data frame according to your rules.