New Game: Chess Golf

[donkirkby]

Same movement mechanic as #20 Neighbour Chess, but as a puzzle race game. Are Neighbour Chess and Chess Golf too similar to include both?

• Add rules.
• Add some examples.
• Add pieces back before dealing.
• Dealer chooses par 2, 3, or maybe 4 by dealing more cards.
• If a colour has no legal moves, then you may make a king move with any piece of that colour.
• Generate interesting problems to solve.
• Reduce memory usage in GolfGraph, or dump state to a db.
• Simulate a bunch of games, and give quartile scores as targets for solitaire play.

[donkirkby]

• Try switching to capturing pieces.
• Check for colour with no neighbours, add highest rank off board next to highest rank on board.
• Deal over captured card.
• For back 9, add pieces in any order, in first square w/ no neighbours of same colour.

[donkirkby]

Try putting pieces back before dealing, and add the option of dealing more cards to make it harder.

[donkirkby]

start_text = dedent("""\
    . . . R . . b .
    . Q . . . K . N
    . . . R . . . .
    b . . . . . n .
    N . . . . . . B
    . . . n . . k .
    . B . . q . . r
    . r . . . . . .
    chosen: bb""")

start_state = GolfState(start_text)
graph = GolfGraph()
tracemalloc.start()

try:
    graph.walk(start_state, size_limit=50_000)
finally:
    snapshot = tracemalloc.take_snapshot()
    display_top(snapshot)

Produced this report:

Top 3 lines
#1: chess-kit/golf.py:67: 16280.9 KiB
    return '\n'.join(sections)
#2: classes/digraph.py:709: 10447.5 KiB
    datadict.update(attr)
#3: classes/digraph.py:711: 5775.9 KiB
    self._pred[v][u] = datadict
56 other: 35030.2 KiB
Total allocated size: 67534.5 KiB

[donkirkby]

After converting nodes from a text display to compact bytes, I still can't get beyond depth 5 with a breadth-first search. Last report was 1.5 million nodes before oom killer struck.

Here's the previous test after converting to compact bytes:

Top 3 lines
#1: classes/digraph.py:709: 10890.9 KiB
    datadict.update(attr)
#2: classes/digraph.py:711: 7965.6 KiB
    self._pred[v][u] = datadict
#3: classes/digraph.py:708: 5809.2 KiB
    datadict = self._adj[u].get(v, self.edge_attr_dict_factory())
57 other: 38197.3 KiB
Total allocated size: 62863.2 KiB

[donkirkby]

Compressing move descriptions made very little difference.

Top 3 lines
#1: classes/digraph.py:709: 10890.9 KiB
    datadict.update(attr)
#2: classes/digraph.py:711: 7965.6 KiB
    self._pred[v][u] = datadict
#3: classes/digraph.py:708: 5809.2 KiB
    datadict = self._adj[u].get(v, self.edge_attr_dict_factory())
57 other: 36568.5 KiB
Total allocated size: 61234.3 KiB