"""
altgraph.GraphUtil - Utility classes and functions
==================================================
"""

import random
from collections import deque

from altgraph import Graph, GraphError


def generate_random_graph(node_num, edge_num, self_loops=False, multi_edges=False):
    """
    Generates and returns a :py:class:`~altgraph.Graph.Graph` instance with
    *node_num* nodes randomly connected by *edge_num* edges.
    """
    g = Graph.Graph()

    if not multi_edges:
        if self_loops:
            max_edges = node_num * node_num
        else:
            max_edges = node_num * (node_num - 1)

        if edge_num > max_edges:
            raise GraphError("inconsistent arguments to 'generate_random_graph'")

    nodes = range(node_num)

    for node in nodes:
        g.add_node(node)

    while 1:
        head = random.choice(nodes)
        tail = random.choice(nodes)

        # loop defense
        if head == tail and not self_loops:
            continue

        # multiple edge defense
        if g.edge_by_node(head, tail) is not None and not multi_edges:
            continue

        # add the edge
        g.add_edge(head, tail)
        if g.number_of_edges() >= edge_num:
            break

    return g


def generate_scale_free_graph(steps, growth_num, self_loops=False, multi_edges=False):
    """
    Generates and returns a :py:class:`~altgraph.Graph.Graph` instance that
    will have *steps* \\* *growth_num* nodes and a scale free (powerlaw)
    connectivity. Starting with a fully connected graph with *growth_num*
    nodes at every step *growth_num* nodes are added to the graph and are
    connected to existing nodes with a probability proportional to the degree
    of these existing nodes.
    """
    # The code doesn't seem to do what the documentation claims.
    graph = Graph.Graph()

    # initialize the graph
    store = []
    for i in range(growth_num):
        for j in range(i + 1, growth_num):
            store.append(i)
            store.append(j)
            graph.add_edge(i, j)

    # generate
    for node in range(growth_num, steps * growth_num):
        graph.add_node(node)
        while graph.out_degree(node) < growth_num:
            nbr = random.choice(store)

            # loop defense
            if node == nbr and not self_loops:
                continue

            # multi edge defense
            if graph.edge_by_node(node, nbr) and not multi_edges:
                continue

            graph.add_edge(node, nbr)

        for nbr in graph.out_nbrs(node):
            store.append(node)
            store.append(nbr)

    return graph


def filter_stack(graph, head, filters):
    """
    Perform a walk in a depth-first order starting
    at *head*.

    Returns (visited, removes, orphans).

    * visited: the set of visited nodes
    * removes: the list of nodes where the node
      data does not all *filters*
    * orphans: tuples of (last_good, node),
      where node is not in removes, is directly
      reachable from a node in *removes* and
      *last_good* is the closest upstream node that is not
      in *removes*.
    """

    visited, removes, orphans = {head}, set(), set()
    stack = deque([(head, head)])
    get_data = graph.node_data
    get_edges = graph.out_edges
    get_tail = graph.tail

    while stack:
        last_good, node = stack.pop()
        data = get_data(node)
        if data is not None:
            for filtfunc in filters:
                if not filtfunc(data):
                    removes.add(node)
                    break
            else:
                last_good = node
        for edge in get_edges(node):
            tail = get_tail(edge)
            if last_good is not node:
                orphans.add((last_good, tail))
            if tail not in visited:
                visited.add(tail)
                stack.append((last_good, tail))

    orphans = [(lg, tl) for (lg, tl) in orphans if tl not in removes]

    return visited, removes, orphans