# -*- coding: utf-8 -*-
"""
补齐参数组合空间: 为"规则内合法但库中缺失"的组合生成零件号, 并指向系列级占位数模。

背景: 当前型号表由真实 STEP 文件扫描而来 —— 没有数模就没有型号, 用户选完也拿不到代号。
本脚本按规则反向生成缺失组合, 使其可选、可出号; 数模用每系列一个占位件, 后续 AI 补真实数模。

用法:
  PYTHONUTF8=1 python gen-missing-combos.py --report      # 只统计, 不写文件
  PYTHONUTF8=1 python gen-missing-combos.py --write       # 写入 CSV + meta 模板 + 占位清单
"""
import json, io, os, csv, itertools, sys, collections

CAT = r'D:/开发/OnebotCatalog/onebot-data/catalog'
REPORT = '--report' in sys.argv
WRITE = '--write' in sys.argv

# 只有"模板能 100% 对齐"的系列才补 (由 infer-sku-template.py 验证)。
# 其余系列 (旧码混用/编码结构特殊) 保持现状, 不生成可能不存在的零件号。
SAFE = {
    'KC', 'KCB', 'KCS', 'KS', 'KSB', 'KSU', 'M', 'MS', 'MSC', 'MAL', 'MALC',
    'ULA', 'ULL', 'ULM', 'ULO', 'ULP', 'ULT', 'ULX', 'ULY', 'ULZ', 'USL', 'USSL',
}

# 各系列的编码模板 (由 infer-sku-template.py 反推, 100% 对齐)
# {参数:补零位数}; 空值渲染为空串
TEMPLATES = {
    'KC':   ('KC{type}{bore}-{stroke}{magnet}{mount}', {}),
    'KCB':  ('KCB{bore}-{stroke}{magnet}{mount}', {}),
    'KCS':  ('KCS{type}{bore}-{stroke}{magnet}{mount}', {}),
    'KS':   ('KS{type}{bore}-{stroke}{magnet}{mount}', {}),
    'KSB':  ('KSB{bore}-{stroke1}x{stroke2}{magnet}{mount}', {}),
    'KSU':  ('KSU{type}{bore}-{stroke}{magnet}{mount}', {}),
    'M':    ('{action}{bore}-{stroke}{magnet}{rod}{mount}{cap}', {}),
    'MS':   ('{action}{bore}-{stroke}{magnet}{rod}{mount}{cap}', {}),
    'MSC':  ('{action}{bore}-{stroke}{magnet}{rod}{mount}{cap}', {}),
    'MAL':  ('{type}{bore}-{stroke}{magnet}{rod}{mount}{cap}', {}),
    'MALC': ('{type}{bore}-{stroke}{magnet}{rod}{mount}{cap}', {}),
    'ULA':  ('ULA{size}{stroke:pad3}{rod}{mount}{sensor}0', {}),
    'ULL':  ('ULL{size}{stroke:pad3}{rod}{orient}{sensor}0', {}),
    'ULM':  ('ULM{size}{stroke:pad3}{rod}{orient}{sensor}0', {}),
    'ULO':  ('ULO{size}{stroke:pad3}{rod}{sensor}0', {}),
    'ULP':  ('ULP{size}{stroke:pad3}{rod}{orient}{sensor}0', {}),
    'ULT':  ('ULT{size}{stroke:pad3}{rod}{sensor}0', {}),
    'ULX':  ('ULX{size}{stroke:pad3}{rod}{sensor}0', {}),
    'ULY':  ('ULY{size}{stroke:pad3}{rod}{orient}{sensor}0', {}),
    'ULZ':  ('ULZ{size}{stroke:pad3}{rod}{orient}{sensor}0', {}),
    'USL':  ('USL{size}{pcmodel}{pin_dia}{locating}{prog}{sensor}0', {}),
    'USSL': ('USSL{size}{pcmodel}{hook_pos}{pin_dia}{locating}{prog}{sensor}0', {}),
}

PLACEHOLDER_STEP = 'PLACEHOLDER.step'   # 每系列共用


def S(v):
    return '' if v is None else str(v)


def eq(a, b):
    return S(a).lower() == S(b).lower()


def cond_match(c, prm):
    if not c:
        return True
    val = prm.get(c['Param'])
    if val is None:
        return False
    op, cv = c['Op'], c['Value']
    if op == 'eq':
        return eq(val, cv)
    if op == 'ne':
        return not eq(val, cv)
    if op == 'in':
        return isinstance(cv, list) and any(eq(val, x) for x in cv)
    try:
        a, b = float(S(val)), float(S(cv))
    except ValueError:
        return False
    return {'gt': a > b, 'ge': a >= b, 'lt': a < b, 'le': a <= b}.get(op, False)


def violated(rules, prm):
    for r in rules or []:
        if cond_match(r.get('If'), prm) and not cond_match(r.get('Then'), prm):
            return True
    return False


def render(tpl, prm):
    """模板渲染: {field} / {field:padN}; 空值渲染为空串。"""
    out, i = '', 0
    while i < len(tpl):
        if tpl[i] == '{':
            j = tpl.index('}', i)
            tok = tpl[i + 1:j]
            name, _, pad = tok.partition(':')
            name = name.strip()
            v = S(prm.get(name, ''))
            if pad.startswith('pad') and v != '':
                try:
                    v = v.rjust(int(pad[3:]), '0')
                except ValueError:
                    return None
            out += v
            i = j + 1
        else:
            out += tpl[i]
            i += 1
    return out


def load_meta(code):
    p = os.path.join(CAT, 'series', code + '.meta.json')
    return json.load(io.open(p, encoding='utf-8'))


def link_or_copy(src, dst):
    """建硬链接; 不支持则回退复制。返回 True 表示已创建。
    注: 必须先建链接再考虑失败回退, 绝不能先删目标再建 (删了又建失败 = 文件丢失)。"""
    if os.path.exists(dst):
        return False
    try:
        os.link(src, dst)
        return True
    except OSError:
        pass
    try:
        import shutil
        shutil.copy2(src, dst)
        return True
    except OSError:
        return False


def main():
    summary = []
    plan = {}
    for code in sorted(SAFE):
        meta = load_meta(code)
        s = meta['series']
        params = s['parameters']
        csvp = os.path.join(CAT, 'csv', code + '.csv')
        if not os.path.exists(csvp):
            print(f'[{code}] 无 CSV, 跳过')
            continue
        rows = list(csv.DictReader(io.open(csvp, encoding='utf-8-sig')))
        existing = {r['model_code'] for r in rows}

        # 值域: 数值参数取 CSV 实际出现过的值 (不用 meta.values, 避免含非法/未出现值)
        doms = collections.OrderedDict()
        for p in params:
            pc = p['code']
            if p['type'] == 'range':
                doms[pc] = None          # 连续域不枚举
                continue
            vals = list(dict.fromkeys(r.get(pc, '') for r in rows))
            vals = [v for v in vals if v != '' or True]
            doms[pc] = vals
        if any(v is None for v in doms.values()):
            print(f'[{code}] 含 range 参数, 跳过 (连续域无法枚举)')
            continue

        tpl, _ = TEMPLATES[code]
        codes = list(doms.keys())
        combos = list(itertools.product(*[doms[k] for k in codes]))
        added = []
        for combo in combos:
            prm = dict(zip(codes, combo))
            if violated(s.get('rules'), prm):
                continue
            mc = render(tpl, prm)
            if not mc or mc in existing:
                continue
            existing.add(mc)
            added.append((mc, prm))

        summary.append((code, len(rows), len(combos), len(added)))
        if added:
            plan[code] = (tpl, params, added, rows)

    # 输出
    print(f'{"系列":7s} {"现有":>7s} {"笛卡尔":>9s} {"待补":>7s}')
    print('-' * 36)
    tot_new = 0
    for code, n, cart, add in sorted(summary, key=lambda x: -x[3]):
        tot_new += add
        print(f'{code:7s} {n:7d} {cart:9,d} {add:7,d}')
    print(f'\n合计待补 {tot_new:,} 个型号')

    if not WRITE:
        print('\n(--report 模式, 未写文件)')
        return

    # ---------- 写入 ----------
    step_dir = os.path.join(CAT, 'step')
    placeholder = os.path.join(step_dir, PLACEHOLDER_STEP)
    manifest = []          # 待 AI 补真实数模的清单
    n_csv = n_ph = 0
    for code, (tpl, params, added, rows) in plan.items():
        csvp = os.path.join(CAT, 'csv', code + '.csv')
        header = list(rows[0].keys())
        dims = [h for h in header if h not in ('model_code', 'step_file')]
        # 新增行: step_file 统一指向 PLACEHOLDER.step ——
        # 构建器按 v.step 的路径复制 (同路径只存一份), 3157 个型号共用一个占位件,
        # 既不占磁盘也不撑大数据包。AI 补真实数模后把该行改成 "<型号>.step" 即可。
        with io.open(csvp, 'a', encoding='utf-8-sig', newline='') as fp:
            w = csv.writer(fp)
            for mc, prm in added:
                w.writerow([mc] + [S(prm.get(d, '')) for d in dims] + [PLACEHOLDER_STEP])
                manifest.append(f'{code},{mc},{",".join(d + "=" + S(prm.get(d, "")) for d in dims)}')
                n_csv += 1
        print(f'[{code}] CSV +{len(added)}')

    # 待补清单 (给 AI 补数模用)
    mf = os.path.join(CAT, '_pending_models.csv')
    with io.open(mf, 'w', encoding='utf-8-sig', newline='') as fp:
        fp.write('series,model_code,params\n')
        fp.write('\n'.join(manifest) + '\n')
    print(f'\n写入完成: CSV 新增 {n_csv} 行, 占位 STEP {n_ph} 个')
    print(f'待补数模清单: {mf}  ({len(manifest)} 个型号)')


if __name__ == '__main__':
    main()