juzi/h5/test/build_fixtures.py

#!/usr/bin/env python3
# -*- coding: utf-8 -*-
# ═════ 真值 fixture 构建脚本（零第三方新依赖：openpyxl 环境已装，仅用于建 fixture）═════
# 用法：python3 h5/test/build_fixtures.py
# 产出：
#   fixtures/case1.json  ← reference/facepp-skinanalyze/examples/ref_3view_raw.json
#                          （案例1 完整原始返回，精简坐标数组与 face_maps base64）
#   fixtures/case2.json  ← docs/皮肤检测测试结果_三视图_20260926.xlsx AY4 扁平化文本解析回嵌套
#   fixtures/truth.json  ← 同表第 3/4 行真值（P/S/档位/文案等断言目标）
# 两个 case fixture 均为服务端 /api/skin-analyze 响应形状：
#   { result, face_rectangle, side: { left: left_side_result, right: right_side_result } }

import json
import os
import re
import sys

import openpyxl

ROOT = os.path.normpath(os.path.join(os.path.dirname(os.path.abspath(__file__)), '..', '..'))
REF_JSON = os.path.join(ROOT, 'reference', 'facepp-skinanalyze', 'examples', 'ref_3view_raw.json')
XLSX = os.path.join(ROOT, 'docs', '皮肤检测测试结果_三视图_20260926.xlsx')
OUT_DIR = os.path.join(os.path.dirname(os.path.abspath(__file__)), 'fixtures')


# ─── 通用：递归精简（坐标数组→[]，face_maps→{}；评分只用标量与 count）───
def slim(node, key=''):
    if isinstance(node, dict):
        if key == 'face_maps':  # base64 整图，评分不需要，整体置空对象
            return {}
        return {k: slim(v, k) for k, v in node.items()}
    if isinstance(node, list):
        # 元素为 dict 的坐标/标记数组（rectangle/polygon/*_mark/*_outline）→ 空数组；
        # 纯数值数组（confidence 列表等）保留
        if node and all(isinstance(x, dict) for x in node):
            return []
        return [slim(x, key) for x in node]
    return node


# ─── 案例1：原始 pro 返回 → 服务端响应形状 ───
def build_case1():
    with open(REF_JSON, encoding='utf-8') as f:
        raw = json.load(f)
    fixture = {
        'request_id': raw.get('request_id'),
        'time_used': raw.get('time_used'),
        'result': slim(raw['result']),
        'face_rectangle': raw.get('face_rectangle'),
        'side': {
            'left': slim(raw.get('left_side_result')),
            'right': slim(raw.get('right_side_result'))
        }
    }
    return fixture


# ─── 案例2：AY4 扁平化「路径 = 值」文本 → 嵌套 JSON ───
SEC_RE = re.compile(r'^——\s*(.+?)\s*——$')
KV_RE = re.compile(r'^([^=]+?)\s*=\s*(.*)$')
NUMLIST_RE = re.compile(r'^[\d\s,.\-eE+]+$')
INT_RE = re.compile(r'^-?\d+$')
FLOAT_RE = re.compile(r'^-?(?:\d+\.\d*|\.\d+)(?:[eE][-+]?\d+)?$|^-?\d+[eE][-+]?\d+$')
MS_RE = re.compile(r'^(-?\d+(?:\.\d+)?)\s*ms$')

skipped = []  # [摘要] 跳过的路径（坐标数组/base64/深层结构）


def parse_value(path, s):
    s = s.strip()
    if s == '':
        return None, False
    if s.startswith('{'):  # 行内 JSON（face_rectangle）
        try:
            return json.loads(s), True
        except ValueError:
            skipped.append(path + '  [行内JSON解析失败]')
            return None, False
    if s.startswith('['):
        if s == '[]':
            return [], True
        inner = s[1:-1].strip() if s.endswith(']') else ''
        if inner and NUMLIST_RE.match(inner):  # 纯数值数组（confidence 列表）
            out = []
            for tok in inner.split(','):
                tok = tok.strip()
                if tok == '':
                    continue
                out.append(float(tok) if ('.' in tok or 'e' in tok or 'E' in tok) else int(tok))
            return out, True
        skipped.append(path + '  ' + s[:40])  # [N组坐标]/[N项]/[base64图,…]/[深层结构省略]
        return None, False
    m = MS_RE.match(s)  # time_used 的「751 ms」
    if m:
        return float(m.group(1)) if '.' in m.group(1) else int(m.group(1)), True
    if INT_RE.match(s):
        return int(s), True
    if FLOAT_RE.match(s):
        return float(s), True
    return s, True  # 字符串（request_id 等）


def set_path(root, path, value):
    segs = path.split('.')
    cur = root
    for i, seg in enumerate(segs[:-1]):
        nxt = segs[i + 1]
        if seg.isdigit():  # 数字段：上级应是 list（本数据仅 skin_type.details.N，均被摘要跳过）
            idx = int(seg)
            while len(cur) <= idx:
                cur.append({} if not nxt.isdigit() else [])
            if cur[idx] is None:
                cur[idx] = [] if nxt.isdigit() else {}
            cur = cur[idx]
        else:
            if seg not in cur or not isinstance(cur[seg], (dict, list)):
                cur[seg] = [] if nxt.isdigit() else {}
            cur = cur[seg]
    last = segs[-1]
    if last.isdigit():
        idx = int(last)
        while len(cur) <= idx:
            cur.append(None)
        cur[idx] = value
    else:
        cur[last] = value


def build_case2(flat_text):
    doc = {}
    section = None  # 当前 —— section —— 名；None=顶层
    for line in flat_text.split('\n'):
        line = line.strip()
        if not line:
            continue
        m = SEC_RE.match(line)
        if m:
            section = m.group(1)
            continue
        m = KV_RE.match(line)
        if not m:
            continue
        path, raw_v = m.group(1).strip(), m.group(2)
        full_path = (section + '.' + path) if section else path
        value, ok = parse_value(full_path, raw_v)
        if not ok:
            continue
        target = doc
        if section:
            if section not in doc or not isinstance(doc[section], dict):
                doc[section] = {}
            target = doc[section]
        set_path(target, path, value)
    fixture = {
        'request_id': doc.get('request_id'),
        'time_used': doc.get('time_used'),
        'result': slim(doc.get('result', {})),
        'face_rectangle': doc.get('face_rectangle'),
        'side': {
            'left': slim(doc.get('left_side_result')),
            'right': slim(doc.get('right_side_result'))
        }
    }
    return fixture


# ─── 真值表：Excel 第 3/4 行 → truth.json ───
def col(ws, row, letter):
    return ws['%s%d' % (letter, row)].value


def build_truth(ws, row):
    def cat(p_col, s_col, tier_col, copy_col):
        return {
            'P': col(ws, row, p_col), 'S': col(ws, row, s_col),
            'tier': col(ws, row, tier_col), 'copy': col(ws, row, copy_col)
        }
    def dim(p_col, s_col, tier_col):
        return {'P': col(ws, row, p_col), 'S': col(ws, row, s_col), 'tier': col(ws, row, tier_col)}
    return {
        'profile': {'realAge': col(ws, row, 'AS'), 'skinType': '混合性'},  # 综合文案实测输出「混合性」
        'skinAge': col(ws, row, 'E'),
        'totalScore': col(ws, row, 'F'),
        'cats': {
            'contour': cat('I', 'J', 'K', 'L'),
            'wrinkles': cat('M', 'N', 'O', 'P'),
            'smooth': cat('Q', 'R', 'S', 'T'),
            'tone': cat('U', 'V', 'W', 'X'),
            'sens': cat('Y', 'Z', 'AA', 'AB'),
            'acne': cat('AC', 'AD', 'AE', 'AF')
        },
        'dims': {
            'firm': dim('AG', 'AH', 'AI'),
            'vita': dim('AJ', 'AK', 'AL'),
            'stab': dim('AM', 'AN', 'AO')
        },
        'jawline': {'left': col(ws, row, 'AP'), 'right': col(ws, row, 'AQ')},
        'group': col(ws, row, 'AT'),
        'essence': col(ws, row, 'AU'),
        'avgS': col(ws, row, 'AV'),
        'summary': col(ws, row, 'AW'),
        'lifestyle': col(ws, row, 'AX')
    }


def dump(name, obj):
    path = os.path.join(OUT_DIR, name)
    with open(path, 'w', encoding='utf-8') as f:
        json.dump(obj, f, ensure_ascii=False, indent=1)
    print('  写出 %s（%d 字节）' % (os.path.relpath(path, ROOT), os.path.getsize(path)))


def main():
    os.makedirs(OUT_DIR, exist_ok=True)
    print('== 案例1 fixture ← ref_3view_raw.json ==')
    dump('case1.json', build_case1())

    print('== 案例2 fixture ← Excel AY4 扁平化解析 ==')
    wb = openpyxl.load_workbook(XLSX, data_only=True)
    ws = wb['三视图检测汇总']
    case2 = build_case2(ws['AY4'].value)
    dump('case2.json', case2)
    print('  [摘要] 跳过 %d 个路径（坐标数组/base64/深层结构，评分只用标量，不影响）：' % len(skipped))
    for s in skipped:
        print('    - ' + s)

    print('== 真值表 ← Excel 第 3/4 行 ==')
    dump('truth.json', {'case1': build_truth(ws, 3), 'case2': build_truth(ws, 4)})

    # 构建期自检：fixture 关键标量 vs 真值
    c1 = json.load(open(os.path.join(OUT_DIR, 'case1.json'), encoding='utf-8'))
    c2 = json.load(open(os.path.join(OUT_DIR, 'case2.json'), encoding='utf-8'))
    t = json.load(open(os.path.join(OUT_DIR, 'truth.json'), encoding='utf-8'))
    for tag, fx, tr in (('案例1', c1, t['case1']), ('案例2', c2, t['case2'])):
        la = fx['side']['left']['left_jawline_info']['left_jawline_angle']
        ra = fx['side']['right']['right_jawline_info']['right_jawline_angle']
        sa = fx['result']['skin_age']['value']
        assert abs(la - tr['jawline']['left']) < 1e-9, (tag, '左下颌角', la, tr['jawline']['left'])
        assert abs(ra - tr['jawline']['right']) < 1e-9, (tag, '右下颌角', ra, tr['jawline']['right'])
        assert sa == tr['skinAge'], (tag, '肤龄', sa, tr['skinAge'])
        print('  %s 自检通过：下颌角 %.2f/%.2f、肤龄 %d 与真值一致' % (tag, la, ra, sa))


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