Identifying pairs in two columns in R

Question

I have a data frame with bilateral relations between countries:

C1 C2
US FR
FR US
US DE
DE US
US RU
US FI
RU FI
FI RU

The links are directional and some of them are missing (e.g. I have US>RU but not RU>US). I would like to identify all unique pairs; to have something like this:

C1 C2 PairID
US FR 1
FR US 1
US DE 2
DE US 2
US RU -
US FI -
RU FI 3
FI RU 3

Any suggestions?

Answer 1

Here is one option assuming you also want to count relations that are not bidirectional like US>RU:

library(dplyr)
df %>%
        mutate(relation = paste(pmin(C1, C2), pmax(C1, C2), sep = "-"), #define the relation no matter the direction
               PairID = cumsum(c(1, head(relation, -1) != tail(relation, -1)))) %>% 
        select(-relation)
# output
  C1 C2 PairID
1 US FR      1
2 FR US      1
3 US DE      2
4 DE US      2
5 US RU      3
6 US FI      4
7 RU FI      5
8 FI RU      5

# Data: df
structure(list(C1 = c("US", "FR", "US", "DE", "US", "US", "RU", 
"FI"), C2 = c("FR", "US", "DE", "US", "RU", "FI", "FI", "RU")), .Names = c("C1", 
"C2"), class = "data.frame", row.names = c(NA, -8L))

Answer 2

We can create a string identifier that captures a given pair of countries independent of their ordering:

library( tidyverse )

# Original data
X <- data_frame(C1 = c("US", "FR", "US", "DE", "US", "US", "RU", "FI"), 
            C2 = c("FR", "US", "DE", "US", "RU", "FI", "FI", "RU"))

# Creates an order-independent string ID for each entry
Y <- X %>% mutate( S = map2_chr( C1, C2, ~str_flatten(sort(c(.x,.y))) ) )
# # A tibble: 8 x 3
#   C1    C2    S    
#   <chr> <chr> <chr>
# 1 US    FR    FRUS 
# 2 FR    US    FRUS 
# 3 US    DE    DEUS 
# 4 DE    US    DEUS 
# 5 US    RU    RUUS 
# ...

We can then use these string identifiers to find pairs of countries that occur in both directions (e.g., US > FR and FR > US). These pairs will have two matching string IDs.

# Identify string IDs with both orderings and assign an integer ID to each
Z <- Y %>% group_by(S) %>% filter( n() == 2 ) %>% ungroup %>%   # Keep groups of size 2
  select(S) %>% distinct %>% mutate( PairID = 1:n() )           # Annotate unique values
# # A tibble: 3 x 2
#   S     PairID
#   <chr>  <int>
# 1 FRUS       1
# 2 DEUS       2
# 3 FIRU       3

All that's left to do is join the new string ID -> integer ID map against the original data, and replace NAs with "-":

left_join( Y, Z ) %>% select(-S) %>% mutate_at( "PairID", replace_na, "-")
# # A tibble: 8 x 3
#   C1    C2    PairID
#   <chr> <chr> <chr> 
# 1 US    FR    1     
# 2 FR    US    1     
# 3 US    DE    2     
# 4 DE    US    2     
# 5 US    RU    -     
# 6 US    FI    -     
# 7 RU    FI    3     
# 8 FI    RU    3