Currently you mark "bad things" on influence map, but setting a higher default weight would also allow marking "good things" on the influence map, so we can tell the pathing to "prefer this".

Copy pasted discord conversation as a note:

jensiiiToday at 6:33 PM
What if I want to mark good things on the map? 😄
EladYaniv01Today at 6:33 PM
ok i see what you mean
jensiiiToday at 6:33 PM
That's basically my question
EladYaniv01Today at 6:33 PM
hold on
jensiiiToday at 6:33 PM
😄
EladYaniv01Today at 6:34 PM
you can take the grid that the lib returns (get_pyastar_grid)
and do
CUSTOM_DEFAULT_WEIGHT = 300
grid = np.where(grid == 1, CUSTOM_DEFAULT_WEIGHT, np.inf).astype(np.float32)
it will switch all 1's to 300
and leave the np.inf for non pathable

fault tolerant , especially the closest_toward_point method

need a test that makes sure that any radius change to be made wont break the pathing

https://visualstudio.microsoft.com/visual-cpp-build-tools/

This is an attempt to proxy hatch (in the nat) using the region connectivity which led to unexplained location

that fix really demonstrates the weight impact on the path

very robust method of writing test suites, should be adopted

ERROR:sc2.main:Error: get_clean_air_grid() got an unexpected keyword argument 'default_weight'

Looks like the clean air grid does not take a default weight.

It's sometimes useful to find the "safest" position for a unit within a restricted radius. For example, a unit kiting while its weapon is on call down. Example method signature:

def self.find_lowest_cost(self, from_pos: Point2, radius: int, grid: np.ndarray) -> Point2:

Traceback (most recent call last):
File "/home/aiarena/aiarena-client/arenaclient/bots/Jensiiibot/sc2/main.py", line 174, in _play_game_ai
await ai.on_step(iteration)
File "/home/aiarena/aiarena-client/arenaclient/bots/Jensiiibot/Jensiiibot.py", line 606, in on_step
self.map_data = MapData(self)
File "/home/aiarena/aiarena-client/arenaclient/bots/Jensiiibot/MapAnalyzer/MapData.py", line 40, in init
self.compile_map() # this is called on init, but allowed to be called again every step
File "/home/aiarena/aiarena-client/arenaclient/bots/Jensiiibot/MapAnalyzer/MapData.py", line 179, in compile_map
self._calc_grid()
File "/home/aiarena/aiarena-client/arenaclient/bots/Jensiiibot/MapAnalyzer/MapData.py", line 106, in _calc_grid
vision_blockers_array[vision_blockers_indices] = 1
IndexError: arrays used as indices must be of integer (or boolean) type

• tests

Get dead airspace from map analyzer. Maybe even group it to regions of dead airspace, so one could use that to navigate to specific areas of dead space in the map.

Regular path finding via the pyastar lib works well, but it would also be useful to calculate paths through the regions this library calculates, where each point in the path is perhaps the center location of the interconnected region.

This could be a far more efficient way to calculate a path for many use cases, while pathfinding via the pyastar lib is most useful when precise movements are required. Some use cases I can think of, but there are probably more:

• Routing an army around the edge of the map
• Routing an army while avoiding a particular region the enemy currently occupies (avoiding regions with influence?)
• Find all available paths from a start region to the target region
• Find path to region while avoiding a certain choke / ramp

At the moment MapData.in_region only accepts Tuple(int, int) as its parameter, any other type and the method returns None. It would be useful if it could also handle:

• Point2 from python-sc2
• Tuple(float, float)

also , make it fault tolerant and not crash the bot running

Currently def get_pyastar_grid() only returns the ground pathing grid.
There should be an option to return the air pathing grid too.

From my limited testing of using the sc2pathlibp air grid, and passing it to the existing pathfind() method in this library it works just as well as the ground path finding.

When influence blocks path on playable area path will be drawn outside of playable area.
Grid was created with command "viking_grid = self.map_data.get_clean_air_grid(default_weight=10)"

Corner parameters could made adjustable, so the users can play around with numbers and see if they can improve on this.

Traceback (most recent call last):
File "C:\Starcraft 2 botti\Ladder\Bots\Jensiiibot\sc2\main.py", line 174, in _play_game_ai
await ai.on_step(iteration)
File "C:\Starcraft 2 botti\Ladder\Bots\Jensiiibot\Jensiiibot.py", line 606, in on_step
self.map_data = MapData(self)
File "C:\Starcraft 2 botti\Ladder\Bots\Jensiiibot\MapAnalyzer\MapData.py", line 40, in init
self.compile_map() # this is called on init, but allowed to be called again every step
File "C:\Starcraft 2 botti\Ladder\Bots\Jensiiibot\MapAnalyzer\MapData.py", line 182, in compile_map
self._calc_chokes()
File "C:\Starcraft 2 botti\Ladder\Bots\Jensiiibot\MapAnalyzer\MapData.py", line 157, in _calc_chokes
region.region_chokes.append(new_choke)
AttributeError: 'NoneType' object has no attribute 'region_chokes'

Region should also have the following

• Region.border_towards_point(Point2) this is very helpful if you want to set your army on the border, facing a direction ( the enemy's army direction, for example)

• Region.is_on_map_edge this will indicate if the region is on the edge of the map, meaning there is a fly zone between it and the map-edge , without any other regions touching that border, this could be useful for drops, nydus, etc

• Region.drop_areas will return areas that are recommended for drops / turret placements

• Region.siege_walls will return a list of points and direction from where to get to that spot, for sieging, example of such spot could be seen here https://www.youtube.com/watch?v=NUQsAWIBTSk&start=458 ( a few seconds in TY is tucking the tank in a
  very strategic spot)

test for compatability with aiarena , sc2ai

right now the problem is that polygon is initialized with Region,
and that makes it hard inheriting it onto ChokeArea which doesn't require a region to init

ramps should get the normal default weight that all pathable cells get

self.get_air_vs_ground_grid() creates a grid where by default, the ground pathable tiles have a cost of 100 and non pathable of 1. Pretend we add cost to the grid because we see a marine:

air_vs_ground_grid = self.get_air_vs_ground_grid()
pos = marine.position
air_vs_ground_grid = self.add_cost(
            position=(int(pos.x), int(pos.y)),
            radius=marine.range,
            grid=air_vs_ground_grid,
            weight=10,
        )

Then the cost values over ground pathable tiles will correctly update to 110 and non pathable will now be 11

Everything works as it should, but the problem with this, is that units will favour the non pathable tile with a value of 11 (which is now dangerous) over a perfectly safe ground tile that has a default value of 100

My suggestion is that when cost is added over a non pathable area because the radius spans over a non pathable area, the default value for the ground pathable tiles is also added.

Highlighted in red on the picture is an example of what the air vs ground grid thinks is safe, but really it's a dangerous area (in this picture the default weight for pathable areas was set to 50).

Terran may not be able to find a path from their main base if they walled off, when the depots are lowered.

• make sure that wall off points are included in buildable points

• assert that once we wall off those points are removed from buildable points

• test it

• tests

Class ChokePoint

with available attributes :

• get_regions ( usually a chokepoint will divide two regions, meaning it is directly connected to two or more )

• get_sides

• get_center

• get_area (or width )