dcggraph 0.0.9

Directed Cyclic Graph Utilities for Python

Directed Cyclic Graph Utilities for Python

installing

pip install dcggraph

description

This package currently used for doing predicated searches from a given node in a directed graph that is allowed to be cyclic. Some points to clear up confusion:

• directed: means an edge between 2 nodes can only be travelled in one direction _(although there can be two edges going opposite directions between the same two nodes)
• predicated: means that even if a node is reachable, it still has to pass a prediate test (i.e. predicate(node) == True) in order to be returned.
  • this means there are really 2 search filters in the graph:
    1. the node has to be reachable without any filtering
    2. the node has to pass the predicate test
• cyclic: means that if you start from a node, and follow a bunch of paths, you can end up at the same node. This makes things a lot harder than if you were working with DAGs.

cycles are tricky

The main trick used to deal with cycles is having the predicate return 3 states:

1. PASS: means the node confidently passes the predicate
2. FAIL: means the node confidently fails the predicate
3. WAIT: means the node is unsure if it passes yet, and would like to defer giving a confident answer until more nodes are seen (since starting point matters in cyclical graphs)

Nodes that return the WAIT state from a predicate get put in the unresolved list, and are re-evaluated every time another node returns PASS or FAIL. After all reachable nodes have been evaluated, any remaining nodes in the unresolved list are demoted to the FAIL state, since they couldn't make up their mind, and there are no more chances for them to give a real answer.

This algorithm seems to work well for the cases I use predicated searches in a graph. I'm positive better algorithms exist than what is implemented here.

API

The main feature is predicated graph compression and search. In addition to the snippet below, look at the tests for example usage

import re
from dcggraph import DCGnode, DCGedge, DCGgraph
from dcggraph.search import PredicateResult, DCGsearch

g = DCGgraph()
n0 = g.create_node("n0")
n1 = g.create_node("n1")
n2 = g.create_node("n2")
n3 = g.create_node("n3")
n4 = g.create_node("n4")
g.create_edge(n0, n1)
g.create_edge(n0, n2)
g.create_edge(n1, n3)
g.create_edge(n1, n4)

def even(node, visited, passed, failed):
   if f re.match("[a-z][02468]", node.name):
       return PredicateResult.PASS
   else:
       return PredicateResult.FAIL

s1 = DCGsearch(g, even)

# reuse the search object to cache FAIL/PASS results between searches
assertEqual(len(s1.search(n0)), 3)
assertEqual(len(s1.search(n1)), 1)
assertEqual(len(s1.search(n2)), 1)
assertEqual(len(s1.search(n3)), 0)
assertEqual(len(s1.search(n4)), 1)