AWK: Parsing a directed graph / Extracting records from a file based on fields matching parts of an input file - awk

Question

Problem
So I have an input file with three fields. It is basically a list describing a type of directed graph. The first field is the starting node, second connection type (as in this case there is more than one), and last is the node projected onto.
The problem is, this is a very large and unruly direct graph, and I am only interested in some paths. So I want to provide an input file which has a name of nodes that I care about. If the nodes are mentioned in either the first or third field of the graph file, then I want that entire record (as the path type might vary).
Question
How to extract only certain records of a directed graph?
Bonus, how to extract only those paths which join nodes of interest by at most one neighbor (i.e. nodes of interest can be second nearest neighbors).
Request
I am trying to improve my AWK programming, which is why 1) I want to do this in AWK and 2) I would greatly appreciate a verbose explanation of the code :)
Example of the Problem
Input file:
A
C
D
File to parse:
A -> B
A -> C
A -> D
B -> A
B -> D
C -> E
D -> F
E -> B
E -> F
F -> C
...
Output:
A -> B
A -> C
A -> D
B -> A
B -> D
C -> E
D -> F
F -> C
Bonus Example:
A -> B -> D -> F -> C

Answer 1

If I understand your problem correctly, then this will do:
awk 'NR==FNR { data[$1] = 1; next } $1 in data || $3 in data { print }' graph[12]
How it works: while reading the first file, add all interesting nodes to data. While reading the second file, print only the lines where field one or field three is in data, i.e. is an interesting node.

Answer 2

Going for the bonus:
function left(str) { # returns the leftmost char of a given edge (A -> B)
return substr(str,1,1)
}
function right(str) { # returns the rightmost...
return substr(str,length(str),1)
}
function cmp_str_ind(i1, v1, i2, v2) # array travese order function
{ # this forces the start from the node in the beginning of input file
if(left(i1)==left(a)&&left(i2)!=left(a)) # or leftmost value in a
return -1
else if(left(i2)==left(a)&&left(i1)!=left(a))
return 1
else if(i1 < i2)
return -1
return (i1 != i2)
}
function trav(a,b,c,d) { # goes thru edges in AWK order
# print "A:"a," C:"c," D:"d
if(index(d,c)||index(d,right(c))) {
return ""
}
d=d", "c # c holds the current edge being examined
if(index(a,right(c))) { # these edges affect a
# print "3:"c
sub(right(c),"",a)
if(a=="") { # when a is empty, path is found
print d # d has the traversed path
exit
}
for (i in b) {
if(left(i)==right(c)) # only try the ones that can be added to the end
trav(a,b,i,d)
}
a=a""right(c)
} else {
# print "4:"c
for (i in b)
if(left(i)==right(c))
trav(a,b,i,d)
}
}
BEGIN { # playing with the traverse order
PROCINFO["sorted_in"]="cmp_str_ind"
}
NR==FNR {
a=a""$0 # a has the input (ADC)
next
}
{
b[$0]=$0 # b has the edges
}
END { # after reading in the data, recursively test every path
for(i in b) # candidate pruning the unfit ones first. CLR or Dijkstra
if(index(a,left(i))) { # were not consulted on that logic.
# print "3: "i
sub(left(i),"",a)
trav(a,b,i,left(i))
a=a""left(i)
}
else {
# print "2: "i
trav(a,b,i,left(i))
}
}
$ awk -f graph.awk input parse
A, A -> D, D -> F, F -> C
If you uncomment the BEGIN part, you get the A, A -> B, B -> D, D -> F, F -> C. I know, I should work on it more and comment it better but it's midnight over here. Maybe tomorrow.