3.1 从规格到轨迹

#!/usr/bin/env python3
# file: sim/gd_mech_03_trajectory.py
# tier: S
# verify: /opt/homebrew/bin/python3 sim/gd_mech_03_trajectory.py
"""3.1 从规格到轨迹（S 层）。

`ashlib.run_spec` 只返回**快照**（终态）。本脚本把同一套规则拆成「一步」逐步执行，
每执行一步就记一行状态，得到**轨迹**：一条按步序排列的状态序列。

规则与 ashlib.run_spec 逐条对应（复用 ashlib 的洗牌与常数，只把每步状态留下来）：
出牌扣能量、加格挡、结算伤害；结束回合敌人攻击（先扣格挡再扣血）、抽牌、能量重置。

轨迹长度 = **实际执行的步数**——被跳过的非法动作（下标越界）与欠费动作（能量不足）
**不计入**，与快照里的 played 不是同一个量。

导出 out/mech03_trajectory.csv（step/turn/energy/playerHp/enemyHp/block/dealt/action），
打印首尾几行。整条轨迹的最后一个状态必须与 run_spec 的快照逐位一致。
"""
from __future__ import annotations

import os
import sys

import ashlib

HERE = os.path.dirname(os.path.abspath(__file__))
ROOT = os.path.dirname(HERE)

HEADER = ["step", "turn", "energy", "playerHp", "enemyHp", "block", "dealt", "action"]


def replay(spec: dict) -> list:
    """逐步重放 strategy，返回轨迹（每步一行状态）。跳过非法/欠费动作，不计行。"""
    p = spec["params"]
    deck: list = []
    for c in spec["cards"]:
        deck.extend([c] * c["count"])
    deck = ashlib.shuffle_deck(deck, spec["seed"])

    hand = deck[: p["handSize"]]
    draw = deck[p["handSize"]:]
    discard: list = []

    turn = 1
    energy = p["energyMax"]
    player_hp = p["playerHp"]
    enemy_hp = p["enemyHp"]
    block = 0
    dealt = 0
    over = False

    rows: list = []
    for cmd in spec["strategy"]:
        if over:
            break
        if turn > p["turnLimit"]:
            over = True
            break
        t = turn  # 本步发生在第几回合（end_turn 会把 turn 加一，故先记下）
        if cmd.startswith("p"):
            idx = int(cmd[1:])
            if idx < 0 or idx >= len(hand):
                continue  # 非法动作：不计步
            c = hand[idx]
            if c["cost"] > energy:
                continue  # 欠费动作：不计步
            hand.pop(idx)
            discard.append(c)
            energy -= c["cost"]
            block += c["block"]
            d = min(c["damage"], enemy_hp)
            enemy_hp -= d
            dealt += d
            action = "play:" + c["id"]
            if enemy_hp <= 0:
                over = True
        elif cmd == "e":
            absorbed = min(block, p["enemyAttack"])
            block -= absorbed
            player_hp = max(0, player_hp - (p["enemyAttack"] - absorbed))
            action = "end_turn"
            if player_hp <= 0:
                over = True
                rows.append(mk(len(rows) + 1, t, energy, player_hp, enemy_hp, block, dealt, action))
                break
            turn += 1
            block = 0
            energy = p["energyMax"]
            while len(hand) < p["handSize"] and draw:
                hand.append(draw.pop(0))
        else:
            continue  # 未知指令：不计步
        rows.append(mk(len(rows) + 1, t, energy, player_hp, enemy_hp, block, dealt, action))
    return rows


def mk(step, turn, energy, player_hp, enemy_hp, block, dealt, action):
    return [step, turn, energy, player_hp, enemy_hp, block, dealt, action]


def main() -> int:
    spec = ashlib.load_spec()
    rows = replay(spec)
    snap = ashlib.run_spec(spec)

    out = os.path.join(ROOT, "out", "mech03_trajectory.csv")
    n = ashlib.write_csv(out, HEADER, rows)

    # 轨迹末态必须与快照逐位一致：轨迹不是另跑一遍，是同一条时间线。
    last = rows[-1]
    same = (
        last[2] == snap["energy"]
        and last[3] == snap["playerHp"]
        and last[4] == snap["enemyHp"]
        and last[5] == snap["block"]
        and last[6] == snap["dealt"]
    )

    print(f"deckSize={ashlib.deck_size(spec)} strategyLen={len(spec['strategy'])}")
    print(f"trajectoryLen={n} playedInSnapshot={snap['played']} skipped={len(spec['strategy']) - n}")
    print("step turn energy playerHp enemyHp block dealt action")
    for r in rows[:4]:
        print(" ".join(str(x) for x in r))
    print("...")
    for r in rows[-4:]:
        print(" ".join(str(x) for x in r))
    print(f"snapshot turn={snap['turn']} enemyHp={snap['enemyHp']} won={snap['won']}")
    print(f"trajectoryLast==snapshot: {same}")
    print(f"wrote out/mech03_trajectory.csv rows={n}")
    return 0 if same else 1


if __name__ == "__main__":
    sys.exit(main())

deckSize=12 strategyLen=25
trajectoryLen=12 playedInSnapshot=9 skipped=13
step turn energy playerHp enemyHp block dealt action
1 1 3 60 47 0 3 play:jab
2 1 1 60 33 0 17 play:heavy
3 1 0 60 27 0 23 play:strike
4 1 3 54 27 0 23 end_turn
...
9 3 2 54 15 0 35 play:strike
10 3 1 54 9 0 41 play:strike
11 3 3 48 9 0 41 end_turn
12 4 1 48 0 50 play:heavy
snapshot turn=4 enemyHp=0 won=True
trajectoryLast==snapshot: True
wrote out/mech03_trajectory.csv rows=12

step,turn,energy,playerHp,enemyHp,block,dealt,action
1,1,3,60,47,0,3,play:jab
2,1,1,60,33,0,17,play:heavy
3,1,0,60,27,0,23,play:strike
4,1,3,54,27,0,23,end_turn
...
12,4,1,48,0,50,play:heavy