How do I use Graph::Traversal?

[thor]

Greetings fellow monks,

I find myself in a situation where I have a data set that's well represented by a Graph. Knowing that CPAN has Graph, I use it. I now want to traverse the graph, but I'm finding the results are not what I expect. Here's a test script:

use strict;
use warnings;

use Graph;
use Graph::Traversal::DFS;

my $graph = Graph->new();
$graph->add_edges( ['A','B'], ['B','C'], ['C','D'], ['D','E'], 
                   ['A','b'], ['b','c'], ['c','d'], ['d','e']);

my $trav = Graph::Traversal::DFS->new($graph);
print join("\n", $trav->dfs());
__END__
E
D
e
d
c
b
C
B
A

I realize that the algorithm is non-deterministic, but it should at least return the correct result. I'd expect that it'd print 'A' and then all the capital letters followed by all the lower-case letters, or vice versa. The docs are unusually sparse. Am I missing something here?

[runrig]

You are doing a Depth First Search, so 'E' is the first result. Perhaps what you want is a Breadth First Search?

Update: that still wouldn't give you what you expect though...I think BFS would return A, B, b, C, c, D, d, E, e).

Last Update?: What you want is DFS, but use $t->preorder instead of $t->dfs. This returns (A, b, c, d, e, B, C, D, E) which maybe acceptably close to what you are expecting.

Aha, and use the next_alphabetic option:

my $trav = Graph::Traversal::DFS->new($graph, next_alphabetic=>1);
my $v; print "$v\n" while $v = $trav->preorder;
__END__
A
B
C
D
E
b
c
d
e

[thor]

I wish I could use the next_alphabetic. However, my example script is a paper tiger; it doesn't reflect the nature of my data (which is not ordered alphabetically). Looking at my actual data and the Graph module, topological_sort might get the job done.

[runrig]

A topo sort is not strict enough for what you want (it just guarantees parents are returned before children), and in this example returns (A, B, C, D, b, c, d, e, E). Is there any way to force an alphabetical order on your vertices? I'd try adding a fixed length prefix to every vertex as you're adding edges, if at all possible. E.g., start with $p = "0000" (or however many places necessary), and just do $p++ before adding vertex "${p}_$v". And then strip the prefix as each vertex is returned before you need to use it.

[Joost]

I can't really help you, but for future reference:

Graph is a module that was introduced in the mastering algorithms in perl book, and used to be best documented in that book. However, browsing at the index of the book, Graph::Traversal seems to be a later addition to the distribution (it's not in the index).

I can only recommend you read the code if the docs aren't clear enough.

[melmoth]

Greetings, I'm exploring the Graph::Traversal::DFS in order to compute all possible paths of a directed graph. This solution here: http://www.perlmonks.org/?node_id=382052 Seems to do what I want: provides all paths of a $tree variable as an array of array references. The problem is that the code is too obscure for me understand at the moment and I've already implemented Graph.pm in my code elsewhere and would like to continue using it. All paths shouldn't be a major computational problem in this case because the graphs will be relatively simple and I could get the components first as a simplification. Here's a simple example for a set of vertices called Group 26:

 print qq[$_->{dna}\t$_->{score}\n] foreach @{$RNA_SEQ{$RNA[0]}->{alignments}};

opera_scaffold_55153 28.9909638554217

opera_scaffold_12001 26.5813253012048

opera_scaffold_221202 19.6159638554217

opera_scaffold_220117 15.210843373494

opera_scaffold_184033 1.35542168674699

# The relationship among the vertices is in an array of hash reference called @E. # Each element of @E looks like this:

{v1 => $vertex1, v2 => $vertex2, rna => $rna, state => 1 }


# build a directed graph with this data, vertices have a weight ( scor
+e )
 
my $g = Graph->new();

for my $edge ( @E )
{
    my %weights = map { $_->{dna} => $_->{score} } @{$RNA_SEQ{$edge->{
+rna}}->{alignments}};
    $g->add_weighted_vertices( $edge->{v1}, $weights{$edge->{v1}}, $ed
+ge->{v2}, $weights{$edge->{v2}});
    $g->add_edge ( $edge->{v1}, $edge->{v2} );
}

# Print out the graph to see what it looks like  

foreach ( @E )
{
    my $w1 = $g->get_vertex_weight($_->{v1});
    my $w2 = $g->get_vertex_weight($_->{v2});
    print qq[$_->{v1} --> $_->{v2}\t$w1 $w2\n];
}

# Output the edges of the graph with weights for each vertex

opera_scaffold_55153 --> opera_scaffold_221202 28.9909638554217 19.6159638554217

opera_scaffold_221202 --> opera_scaffold_184033 19.6159638554217 1.35542168674699

opera_scaffold_184033 --> opera_scaffold_220117 1.35542168674699 15.210843373494

opera_scaffold_220117 --> opera_scaffold_12001 15.210843373494 26.5813253012048

# traverse the graph using DFS 
my $t = Graph::Traversal::DFS->new($g);
my @v = $t->preorder;
print qq[Preorder:\n];
print qq[$_\t] foreach @v;

Preorder: opera_scaffold_221202 opera_scaffold_184033 opera_scaffold_220117 opera_scaffold_12001 opera_scaffold_55153

@v = $t->postorder;
print qq[Postorder:\n];
print qq[$_\t] foreach @v;

Postorder: opera_scaffold_12001 opera_scaffold_220117 opera_scaffold_184033 opera_scaffold_221202 opera_scaffold_55153

my @r = $t->roots;
print qq[Roots:\n];
print qq[$_\t] foreach @r;

Roots: opera_scaffold_221202 opera_scaffold_55153

Postorder gives me what I want in this example, albeit in the reverse order. However this is a simple example in which the graph has just one single path.

Group 16 looks like this:

Graph Edges:

opera_scaffold_131077 --> opera_scaffold_45770 2.43611177454024 54.8841652734655

opera_scaffold_102945 --> opera_scaffold_23837 13.8762837353714 19.871029376642

Preorder:

opera_scaffold_45770 opera_scaffold_102945 opera_scaffold_23837 opera_scaffold_131077

Postorder:

opera_scaffold_45770 opera_scaffold_23837 opera_scaffold_102945 opera_scaffold_131077

Roots:

opera_scaffold_45770 opera_scaffold_102945 opera_scaffold_131077

Clearly not what I'm looking for. There's two paths in this graph represented by the two edges: opera_scaffold_131077 --> opera_scaffold_45770 and opera_scaffold_102945 --> opera_scaffold_23837 Again, I need a way that, given a graph, I can get back all possible paths. At that point I need to determine the cumulative score for each path, and then choose the one with the highest score. Graph::Traversal seems to offer a lot of options but I'm not sure how to use them all and the documentation isn't very good. Can anyone offer some help on this problem PLEASE. THANKS!.

[Anonymous Monk]

o.k that last one was a poor example as I could have computed the components first. The solution is required for other logical possibilities, such as:

v1 -> v2

v2 -> v3

v2 -> v5

v3 -> v4

The paths would then be:

v1 -> v2 ->v3 -> v4

v1 -> v2 -> v3