OnebotCatalog/onebot-data/tools/make-pending-list.py

# -*- coding: utf-8 -*-
"""
重写待补数模清单, 增加「核对」相关列, 便于 AI 逐个核实后补真实数模。

背景: v2.1 按编码模板外推生成了 3157 个组合, 但它们是【候选代号】而非已确认的产品 ——
      规律在已有型号上 100% 成立, 外推区域是否有真实产品未经核实。
      本清单把每个候选与其"最近的真实型号"并列, 便于逐个比对核实。

输出列:
  series        系列码
  model_code    候选零件号 (已进目录, 选型可选)
  params        完整参数
  nearest_real  同系列中最接近的【真实数模】型号 (按参数相同个数最多, 平手取编码最近)
  diff          与 nearest_real 不同的参数 (供人工/AI 判断该差异是否真实存在)
  step_file     当前指向 PLACEHOLDER.step; 补完真实数模后改为 <候选型号>.step
  status        固定为 pending (待核对+待补数模)

用法: PYTHONUTF8=1 python make-pending-list.py
"""
import csv, io, os, json, glob, collections

CAT = r'D:/开发/OnebotCatalog/onebot-data/catalog'
OUT = os.path.join(CAT, '_pending_models.csv')
PLACEHOLDER = 'PLACEHOLDER.step'


def load_real_models():
    """返回 {系列: [(model_code, {参数})]} —— 只含指向真实数模的型号。"""
    real = {}
    for f in glob.glob(os.path.join(CAT, 'csv', '*.csv')):
        code = os.path.basename(f)[:-4]
        rows = []
        for r in csv.DictReader(io.open(f, encoding='utf-8-sig')):
            if r.get('step_file') == PLACEHOLDER:
                continue                     # 跳过占位行
            prm = {k: v for k, v in r.items() if k not in ('model_code', 'step_file')}
            rows.append((r['model_code'], prm))
        if rows:
            real[code] = rows
    return real


def dist(params_a, params_b):
    """参数相同个数 (越大越接近)。"""
    keys = set(params_a) | set(params_b)
    return sum(1 for k in keys if str(params_a.get(k, '')) == str(params_b.get(k, '')))


def main():
    real = load_real_models()
    total_real = sum(len(v) for v in real.values())
    print(f'真实数模型号: {total_real} 个, 覆盖 {len(real)} 个系列')

    # 收集候选项
    pending = []
    for f in sorted(glob.glob(os.path.join(CAT, 'csv', '*.csv'))):
        code = os.path.basename(f)[:-4]
        for r in csv.DictReader(io.open(f, encoding='utf-8-sig')):
            if r.get('step_file') != PLACEHOLDER:
                continue
            prm = {k: v for k, v in r.items() if k not in ('model_code', 'step_file')}
            pending.append((code, r['model_code'], prm))

    print(f'待补候选: {len(pending)} 个')

    rows = []
    for code, mc, prm in pending:
        cands = real.get(code) or []
        near, ndiff = '', ''
        if cands:
            # 最接近的真实型号: 参数相同个数最多; 平手取编码字典序 (稳定)
            best = max(cands, key=lambda x: (dist(prm, x[1]), -len(x[0])))
            near = best[0]
            diffs = [f'{k}={prm.get(k, "")}≠{best[1].get(k, "")}'
                     for k in sorted(set(prm) | set(best[1]))
                     if str(prm.get(k, '')) != str(best[1].get(k, ''))]
            ndiff = '; '.join(diffs)
        rows.append([code, mc,
                     ' '.join(f'{k}={v}' for k, v in prm.items()),
                     near, ndiff, PLACEHOLDER, 'pending'])

    with io.open(OUT, 'w', encoding='utf-8-sig', newline='') as fp:
        w = csv.writer(fp)
        w.writerow(['series', 'model_code', 'params', 'nearest_real',
                    'diff_from_nearest', 'step_file', 'status'])
        w.writerows(rows)
    print(f'\n写入 {OUT}  ({len(rows)} 行)')

    # 覆盖度: 多少候选能找到同系列真实型号作参照
    withref = sum(1 for r in rows if r[3])
    print(f'  有同系列真实型号可参照: {withref}/{len(rows)}')
    nop = [r[0] for r in rows if not r[3]]
    if nop:
        print(f'  无参照的系列: {sorted(set(nop))}')

    # 按系列 + 缺失维度统计
    print('\n各系列待补 (按"与最近真实型号差异的维度数"分布):')
    bycat = collections.defaultdict(collections.Counter)
    for r in rows:
        n = len([x for x in r[4].split('; ') if x]) if r[4] else -1
        bycat[r[0]][n] += 1
    for code in sorted(bycat):
        dists = bycat[code]
        parts = ', '.join(f'{k}维差:{v}' for k, v in sorted(dists.items()) if k >= 0)
        print(f'  {code:6s} 共{sum(dists.values()):5d}   {parts}')


if __name__ == '__main__':
    main()