Another option would be to use Bio::Coordinate::Graph, it's part of the BioPerl package of Perl Modules. It uses Djikstra's algorithm to solve the shortest path problem, and you just need to provide a hash of hashes for it to compute the paths, like this:

# this hash represents a directed graph with vertices {A,B,C,D,E} and 
# these edges:
# A -> B (weight = 3)
# B -> D (weight = 1)
# B -> C (weight = 1)
# D -> E (weight = 1)

my $hash= {
             'A' => {
                     'B' => 3
                    },
             'B' => {
                     'D' => 1,
                     'C' => 1
                    },
             'C' => undef,
             'D' => {
                     'E' => 1
                    },
             'E' => undef
            };
my $graph = Bio::Coordinate::Graph->new(-graph => $hash);
my @node_names = shortest_path( 'A', 'E' );

# node_names now contains a list of the vertex names that make up the
# shortest path (A,B,D,E)
[download]

So, in the context of your question you could build a hash of hashes as above, where the node names are SQL table names and there is a path from A->B if there is a foreign key relationship between tables A and B.

Hope this helps. :)