#!/usr/bin/env python3 """对比 InputFieldBuilder 生成的 formData vs baseline 真实 formData. 读取: - scripts/InputFieldBuilder-output.json (Java 生成产物, mvn test dumpToFile 写入) - scripts/input-baseline-samples.jsonl (baseline 真实 82 sample) 输出关键字段差异率 (允许 formId/renderKey/on 模板/fullNameI18nCode 不同 - 这些是 baseline 自带差异). """ import json, sys from collections import defaultdict # 加载生成产物 try: generated = json.load(open("scripts/InputFieldBuilder-output.json")) except FileNotFoundError: print("ERR: scripts/InputFieldBuilder-output.json 不存在. 跑 mvn test -Dtest=InputFieldBuilderTest#testInputField_dumpToFile 先生成.") sys.exit(1) # 加载 baseline 样本 samples = [json.loads(l) for l in open("scripts/input-baseline-samples.jsonl")] # 允许的差异字段 (formId/renderKey 是每次生成都不同, on 默认模板可变) ALLOW_DIFF = {"formId", "renderKey", "tableName", "on", "__vModel__", "id"} def compare(g, b, path=""): """递归对比两 dict, 返回差异列表 [(path, g_val, b_val)]""" diffs = [] if not isinstance(b, dict): return [(path, g, b)] for k, bv in b.items(): sub = f"{path}.{k}" if path else k if sub.split('.')[-1] in ALLOW_DIFF: continue gv = g.get(k) if isinstance(g, dict) else None if isinstance(bv, dict): diffs.extend(compare(gv if isinstance(gv, dict) else {}, bv, sub)) elif gv != bv: diffs.append((sub, gv, bv)) return diffs # 取 baseline 中 1 个最常见的"标准" input 控件 (排除 on=JS 模板的特例) # 选择 "624164673918337925.json 工号" 作为参考 baseline (有 maskConfig, on 默认模板) ref = None for s in samples: if s['formData'].get('useMask') is False and s['formData'].get('maskConfig'): ref = s break if not ref: ref = samples[0] print(f"=== InputFieldBuilder 0 误差对比 ===") print(f"生成产物: scripts/InputFieldBuilder-output.json (43 字段)") print(f"Baseline 参考: {ref['source']} (label={ref['label']})") print() # 1. 字段名 (key) 对比 g_keys_top = set(generated.keys()) b_keys_top = set(ref['formData'].keys()) print(f"--- 顶层 key 对比 ---") print(f" 生成: {sorted(g_keys_top)}") print(f" baseline: {sorted(b_keys_top)}") print(f" 多出: {sorted(g_keys_top - b_keys_top)}") print(f" 缺失: {sorted(b_keys_top - g_keys_top)}") g_keys_cfg = set(generated.get('__config__', {}).keys()) b_keys_cfg = set(ref['formData'].get('__config__', {}).keys()) print(f"--- __config__ key 对比 ---") print(f" 生成: {sorted(g_keys_cfg)}") print(f" baseline: {sorted(b_keys_cfg)}") print(f" 多出: {sorted(g_keys_cfg - b_keys_cfg)}") print(f" 缺失: {sorted(b_keys_cfg - g_keys_cfg)}") # 2. 值差异 (排除 ALLOW_DIFF) print() print(f"--- 值差异 (排除 {ALLOW_DIFF}) ---") diffs = compare(generated, ref['formData']) if not diffs: print(" ✓ 0 误差 - 全部关键字段值匹配") else: print(f" {len(diffs)} 处差异:") for path, gv, bv in diffs[:20]: print(f" {path}: 生成={gv!r}, baseline={bv!r}") if len(diffs) > 20: print(f" ... 还有 {len(diffs)-20} 处") # 3. 跨 82 baseline 样本的覆盖率 print() print(f"--- 82 baseline sample 覆盖率 ---") all_top = set() all_cfg = set() for s in samples: all_top |= set(s['formData'].keys()) if isinstance(s['formData'].get('__config__'), dict): all_cfg |= set(s['formData']['__config__'].keys()) print(f" 全集顶层 key ({len(all_top)}): 生成覆盖 {len(g_keys_top & all_top)}/{len(all_top)}") print(f" 全集 __config__ key ({len(all_cfg)}): 生成覆盖 {len(g_keys_cfg & all_cfg)}/{len(all_cfg)}") print(f" 生成多余 (baseline 全无): {(g_keys_top - all_top) | (g_keys_cfg - all_cfg)}")