a_star 3.0.0 copy "a_star: ^3.0.0" to clipboard
a_star: ^3.0.0 copied to clipboard

A comprehensive A* algorithm. Suitable for anything from physical grids to abstract search spaces.

A* path finding with Dart #

A simple A* algorithm implemented in Dart. An example of path finding.

Last updated 2021-12-29.

The original 2D algorithm was ported from this JavaScript example. No effort has been made to optimize it. A more generic A* algorithm was added in November 2013. That one is fairly optimized.

See LICENSE file for license details.

See an (old) running example at http://sethladd.github.io/dart-a-star/deploy/

Example #

There are two separate A* algorithms in this package. One of them, aStar2D, is specific to 2D grid maps. The usage can be as simple as:

import 'package:a_star/a_star_2d.dart';
main() {
  String textMap = """
        sooooooo
        oxxxxxoo
        oxxoxooo
        oxoogxxx      
        """;
  Maze maze = new Maze.parse(textMap);
  Queue<Tile> solution = aStar2D(maze);
}

The second algorithm is generic and works on any graph (e.g. 3D grids, mesh networks). The usage is best explained with an example (details below):

import 'package:a_star/a_star.dart';

class TerrainTile extends Object with Node {
  // ...
}
class TerrainMap implements Graph<TerrainTile> {
  // Must implement 4 methods.
  Iterable<T> get allNodes => /* ... */
  num getDistance(T a, T b) => /* ... */
  num getHeuristicDistance(T a, T b) => /* ... */
  Iterable<T> getNeighboursOf(T node) => /* ... */
}

main() {
  var map = new TerrainMap();
  var pathFinder = new AStar(map);
  var start = /* ... */
  var goal = /* ... */
  pathFinder.findPath(start, goal)
  .then((path) => print("The best path from $start to $goal is: $path"));
}

Explanation: Here, we have a TerrainMap of TerrainTile nodes. The only requirements are that TerrainMap implements Graph (4 methods) and TerrainTile is extended with the Node mixin (no additional work). Then, we can simply instantiate the A* algorithm by new AStar(map) and find paths between two nodes by calling the findPath(start, goal) method. Normally, we would only create the AStar instance once and then reuse it throughout our program. This saves performance.

You can also use findPathSync(start, goal) if you don't need to worry about blocking.

All the three classes (AStar, Graph and Node) are well documented in lib/a_star.dart. For a complete example, see the minimal unit tests or one of the two generalized benchmarks (benchmark.dart or benchmark_generalized_2d).

Reporting bugs #

Please file bugs at https://github.com/sethladd/dart-a-star/issues

Contributors #

2
likes
135
pub points
60%
popularity

Publisher

unverified uploader

A comprehensive A* algorithm. Suitable for anything from physical grids to abstract search spaces.

Repository (GitHub)
View/report issues

Documentation

API reference

License

unknown (license)

Dependencies

collection, meta

More

Packages that depend on a_star