add compression methods for decks/replays

This commit is contained in:
Maximilian Keßler 2023-03-02 20:16:26 +01:00
parent 0bd0b42230
commit 174e8276f2
Signed by: max
GPG key ID: BCC5A619923C0BA5

91
compress.py Normal file
View file

@ -0,0 +1,91 @@
import json
from enum import Enum
from typing import List, Optional
BASE62 = "abcdefghijklmnopqrstuvwxyz0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ";
COLORS = 'rygbp'
class DeckCard():
def __init__(self, suitIndex: int, rank: int):
self.suitIndex: int = suitIndex
self.rank: int = rank
def __eq__(self, other):
return self.suitIndex == other.suitIndex and self.rank == other.rank
def __repr__(self):
return COLORS[self.suitIndex] + str(self.rank)
class ActionType(Enum):
Play = 0
Discard = 1
ColorClue = 2
RankClue = 3
EndGame = 4
class Action():
def __init__(self, type_: ActionType, target: int, value: Optional[int] = None):
self.type = type_
self.target = target
self.value = value
def compress_actions(actions: List[Action]) -> str:
minType = 0
maxType = 0
if len(actions) != 0:
minType = min(map(lambda a: a.type.value, actions))
maxType = max(map(lambda a: a.type.value, actions))
typeRange = maxType - minType + 1
def compress_action(action):
value = 0 if action.value is None else action.value
a = BASE62[typeRange * value + (action.type.value - minType)]
b = BASE62[action.target]
return a + b
out = str(minType) + str(maxType)
out += ''.join(map(compress_action, actions))
return out
def compress_deck(deck: List[DeckCard]) -> str:
assert(len(deck) != 0)
minRank = min(map(lambda c: c.rank, deck))
maxRank = max(map(lambda c: c.rank, deck))
rankRange = maxRank - minRank + 1
def compress_card(card):
return BASE62[rankRange * card.suitIndex + (card.rank - minRank)]
out = str(minRank) + str(maxRank)
out += ''.join(map(compress_card, deck))
return out
def decompress_deck(deck_str: str) -> List[DeckCard]:
assert(len(deck_str) >= 2)
minRank = int(deck_str[0])
maxRank = int(deck_str[1])
assert(maxRank >= minRank)
rankRange = maxRank - minRank + 1
def decompress_card(card_char):
index = BASE62.index(card_char)
suitIndex = index // rankRange
rank = index % rankRange + minRank
return DeckCard(suitIndex, rank)
return [decompress_card(c) for c in deck_str[2:]]
## test
deck = [DeckCard(0,1), DeckCard(2,4), DeckCard(4,5)]
c = compress_deck(deck)
l = decompress_deck(c)
print(deck, l)
f = [Action(ActionType.Discard, 2), Action(ActionType.Play, 3, 8)]
a = compress_actions(f)
print(a)