networkx-astar-path

Alternative A* implementation, which provides the current and previous edge to the weight function.

Requirements

• Python 3.6.1 or newer

Installation

pip install networkx-astar-path

Usage

networkx's implementation of A* limit the arguments of the weight function to the current edge onlye. Some scenarious require that the cost for the current edge is dependent on the prevous edge. An example is the bearing between two edges. Surely, the bearing could be pre-calculate and stored on the node, but this is sometimes not possible.

The API of this implementation is the same, with the one exception that the weight-funciton signature has changed to

def weight(graph: nx.Graph, prev_edge: Optional[Tuple[Any, Any]], cur_edge: Tuple[Any, Any]) -> float:
    ...

If the weight-function is called for the first time, the value for prev_edge is None as we haven't visited any other edge yet.

Example

NOTE This example is not very practical, but shows how this function can be used.

Given the following graph

import networks as nx

graph = nx.DiGraph()

graph.add_edges_from([('S', 'A1')], weight=-2)
graph.add_edges_from([('A1', 'T')], weight=7)
graph.add_edges_from([('S','A2'), ('A2','B2'),('B2','C2'),('C2','T')], weight=1)

we are searching for the shortest path from S to T.

The weights have been chosen in a way that the path ['S', 'A2', 'B2', 'C2', 'T'] is shorter when we simply sum up the weights, but longer if the weight of the current edge is divided by the weight of the previous edge. The shortest path for the latter is ['S', 'A1', 'T'].

Let's find the shortest path by only looking at the weights of the current edge.

from networkx_astar_path import astar_path

path = astar_path(graph, source='S', target='T', weight="weight")
# ['S', 'A2', 'B2', 'C2', 'T']

We now define a "simple" weight function, which takes the previous edge into account:

If we already visited an edge, the weight is the weight of the current edge divided by the weight of the previous edge. Otherwise, the weight of the current edge is returned.

from networkx_astar_path import astar_path

def dependent_weight(graph: nx.Graph, prev_edge: Optional[Tuple[Any, Any]], cur_edge: Tuple[Any, Any]) -> float:
    if prev_edge is None:
        return graph.get_edge_data(*cur_edge)['weight']

    prev_weight = graph.get_edge_data(*prev_edge)['weight']
    cur_weight = graph.get_edge_data(*cur_edge)['weight']
    return cur_weight / prev_weight

path = astar_path(graph, source='S', target='T', weight=dependent_weight)
# ['S', 'A1', 'T']

Development

This project uses poetry for packaging and managing all dependencies and pre-commit to run flake8, isort, mypy and black.

Clone this repository and run

poetry install
poetry run pre-commit install

to create a virtual enviroment containing all dependencies. Afterwards, You can run the test suite using

poetry run pytest

This repository follows the Conventional Commits style.

Cookiecutter template

This project was created using cruft and the cookiecutter-pyproject template. In order to update this repository to the latest template version run

cruft update

in the root of this repository.