How to do a join using awk - awk

Question

Here is my Input file
Identifier Relation
A 1
A 2
A 3
B 2
B 3
C 1
C 2
C 3
I want to join this file to itself based on the "Relation" field.
Sample Output file
A 1 C 1
A 2 B 2
A 2 C 2
B 2 C 2
A 3 B 3
A 3 C 3
B 3 C 3
I used the following awk script:
awk 'NR==FNR {a[NR]=$0; next} { for (k in a) if (a[k]~$2) print a[k],$0}' input input > output
However, I had to do another awk step to delete lines which did a join with itself i.e, A 1 A 1 ; B 2 B 2 etc.
The second issue with this file is it prints both directions of the join, thus
A 1 C 1 is printed along with C 1 A 1 on another line.
Both these lines display the same relation and I would NOT like to include this.I want to see just one or the other i.e, "A 1 C 1" or "C 1 A 1" not both.
Any suggestions/directions are highly appreciated.

Answer 1

alternative solution using awk with join and sort support
$ join -j 2 <(sort -k2 -k1,1 file){,}
| awk '$2!=$3 && !($3 FS $2 in a){a[$2 FS $3]; print$2,$1,$3,$1}'
A 1 C 1
A 2 B 2
A 2 C 2
B 2 C 2
A 3 B 3
A 3 C 3
B 3 C 3
create the cross product, eliminate the diagonal and one of the symmetrical pairs.

Answer 2

Here is an awk-only solution:
awk 'NR>1{ar[$2]=(ar[$2]$1);}\
END{ for(key in ar){\
for(i=1; i<length(ar[key]); i++) {\
for(j=i+1; j<length(ar[key])+1; j++) {\
print substr(ar[key],i,1), key, substr(ar[key],j,1), key;\
}\
}\
}}' infile
Each number in the second column of the input serves as a key of an awk-array. The value of the corresponding array-element is a sequence of first-column letters (e.g., array[1]=ABC).
Then, we built all two-letter combinations for each sequence (e.g., "ABC" gives "AB", "AC" and "BC")
Output:
A 1 C 1
A 2 B 2
A 2 C 2
B 2 C 2
A 3 B 3
A 3 C 3
B 3 C 3
Note:
If a number occurs only once, no output is generated for this number.
The order of output depends on the order of input. (No sorting of letters!!). That is if the second input line was C 1, then array[1]="CAB" and the first output line would be C 1 A 1
First line of input is ignored due to NR>1

Answer 3

There is surely a solution with awk only, but I'm going to propose a solution using awk and sort because I think it's quite simple and does not require storing the entire file content in awk variables. The idea is as follows:
rewrite the input file so that the "relation" field is first (A 1 -> 1 A)
use sort -n to put together all lines with same "relation"
use awk to combine consecutive lines having the same "relation"
That would translate to something like:
awk '{print $2 " " $1}' input | sort -n |
awk '{if ($1==lastsel)printf " "; else if(lastsel) printf "\n"; lastsel=$1; printf "%s %s", $2, $1;}END{if(lastsel)printf"\n"}'
A 1 C 1
A 2 B 2 C 2
A 3 B 3 C 3
EDIT: If you want only one i-j relation per line:
awk '{print $2 " " $1}' input | sort -n |
awk '$1!=rel{rel=$1;item=$2;next;} {printf "%s %s %s %s\n", item, rel, $2, $1;}'
A 1 C 1
A 2 B 2
A 2 C 2
A 3 B 3
A 3 C 3
Note the following limitations with this solution:
In case a given n has only one entry, nothing will be output (no output such as D 1)
All relations always have the lexicographically first item in the first column (e.g. A 1 C 1 but never B 1 C 1)