@hupan56/wlkj 2.5.0 → 2.7.0

package/templates/qoder/scripts/benchmark.py

@@ -1,209 +1,210 @@
-#!/usr/bin/env python3
-# -*- coding: utf-8 -*-
-"""
-benchmark.py - QODER 流水线性能基准 (性能优化 A7)
-
-测量关键操作的延迟和内存, 供优化前后对比。
-
-Usage:
-    python benchmark.py                    # 跑全部基准
-    python benchmark.py --search           # 只测搜索
-    python benchmark.py --build            # 只测构建 (耗时)
-    python benchmark.py --json             # 输出 JSON (供对比)
-
-指标:
-    - search_cold: 冷启动搜索 (Python 启动 + JSON 加载 + 扫描)
-    - search_warm: 热搜索 (倒排索引缓存命中)
-    - index_load: keyword-index.json 加载时间
-    - inverted_build: 倒排索引构建时间
-    - inverted_cache_load: 缓存加载时间
-    - cnmap_lookup: CN_MAP 查询 (含 auto-expand)
-"""
-
-import argparse
-import json
-import os
-import sys
-import time
-from datetime import datetime
-
-try:
-    sys.stdout.reconfigure(encoding='utf-8', errors='replace')
-except (AttributeError, TypeError, OSError):
-    pass
-
-THIS_DIR = os.path.dirname(os.path.abspath(__file__))  # .qoder/scripts
-sys.path.insert(0, THIS_DIR)
-# BASE = repo root = scripts 的上两级 (.qoder/scripts -> .qoder -> repo)
-BASE = os.path.dirname(os.path.dirname(THIS_DIR))
-
-INDEX_DIR = os.path.join(BASE, 'data', 'index')
-
-
-def measure(fn, repeat=1):
-    """测函数执行时间, 返回 (avg_ms, min_ms, max_ms)。"""
-    times = []
-    for _ in range(repeat):
-        t0 = time.perf_counter()
-        fn()
-        times.append((time.perf_counter() - t0) * 1000)
-    return sum(times) / len(times), min(times), max(times)
-
-
-def bench_search():
-    """搜索基准: 冷启动 + 热搜索对比。"""
-    results = {}
-
-    # 1. JSON 加载 (冷)
-    ki_path = os.path.join(INDEX_DIR, 'keyword-index.json')
-    def load_ki():
-        with open(ki_path, encoding='utf-8') as f:
-            return json.load(f)
-    results['index_load_ms'] = measure(load_ki)[0]
-    ki = load_ki()
-    results['keyword_count'] = len(ki)
-
-    # 2. 全扫描 (旧逻辑)
-    from search_index import _match_keyword
-    queries = ['insurance', 'attendance', 'salary', 'vehicle', 'user']
-    def full_scan():
-        for q in queries:
-            for kw in ki:
-                if _match_keyword(q, kw):
-                    pass
-    results['full_scan_5q_ms'] = measure(full_scan, repeat=5)[0] / 5
-
-    # 3. 倒排索引构建 (首次)
-    from common.search_engine import get_inverted_index, search_keywords_fast, build_inverted_index
-    # 清缓存测构建
-    cache_path = os.path.join(INDEX_DIR, '.inverted-cache.json')
-    if os.path.exists(cache_path):
-        os.remove(cache_path)
-    results['inverted_build_ms'] = measure(lambda: get_inverted_index(BASE))[0]
-
-    # 4. 缓存加载 (热)
-    results['inverted_cache_load_ms'] = measure(lambda: get_inverted_index(BASE), repeat=5)[0]
-
-    # 5. 倒排搜索 (热)
-    inverted = get_inverted_index(BASE)
-    def inverted_search():
-        for q in queries:
-            search_keywords_fast([q], inverted, ki)
-    results['inverted_scan_5q_ms'] = measure(inverted_search, repeat=5)[0] / 5
-
-    # 6. 提速比
-    if results['inverted_scan_5q_ms'] > 0:
-        results['speedup'] = round(results['full_scan_5q_ms'] / results['inverted_scan_5q_ms'], 1)
-    else:
-        results['speedup'] = float('inf')
-
-    return results
-
-
-def bench_cnmap():
-    """CN_MAP 查询基准。"""
-    from common.terms import CN_MAP, BUSINESS_PATH_MAP, get_cn_map_with_auto
-    results = {}
-    results['cn_map_static_count'] = len(CN_MAP)
-    results['business_path_map_count'] = len(BUSINESS_PATH_MAP)
-
-    # auto-expand (含 PRD 扫描)
-    def auto_expand():
-        return get_cn_map_with_auto()
-    results['auto_expand_ms'] = measure(auto_expand)[0]
-    results['cn_map_total_count'] = len(get_cn_map_with_auto())
-
-    # 查询扩展
-    test_terms = ['考勤', '保险', '库存', '采购', '发票']
-    def expand_query():
-        out = []
-        for t in test_terms:
-            if t in CN_MAP:
-                out.extend(CN_MAP[t].split())
-        return out
-    results['query_expand_5q_ms'] = measure(expand_query, repeat=10)[0] / 10
-
-    return results
-
-
-def bench_index_sizes():
-    """索引文件大小。"""
-    results = {}
-    for name in ['keyword-index.json', 'api-index.json', 'style-index.json',
-                 'prd-index.json', 'module-map.json', 'prd-code-map.json']:
-        p = os.path.join(INDEX_DIR, name)
-        if os.path.isfile(p):
-            results[name] = os.path.getsize(p)
-    # 缓存文件
-    for name in ['.inverted-cache.json', '.file-keys.json']:
-        p = os.path.join(INDEX_DIR, name)
-        if os.path.isfile(p):
-            results[name] = os.path.getsize(p)
-    return results
-
-
-def main():
-    parser = argparse.ArgumentParser(description='QODER 性能基准')
-    parser.add_argument('--search', action='store_true', help='只测搜索')
-    parser.add_argument('--build', action='store_true', help='测构建 (耗时)')
-    parser.add_argument('--json', action='store_true', help='JSON 输出')
-    args = parser.parse_args()
-
-    report = {
-        'timestamp': datetime.now().isoformat(timespec='seconds'),
-        'python': sys.version.split()[0],
-        'platform': sys.platform,
-        'base': os.path.basename(BASE),
-    }
-
-    if not args.build:
-        print('=' * 50)
-        print('QODER 性能基准')
-        print('=' * 50)
-
-    if not args.build or args.search:
-        if not args.json:
-            print('\n--- 搜索性能 ---')
-        sr = bench_search()
-        report['search'] = sr
-        if not args.json:
-            print('  keyword-index 加载: %.0f ms (%d keys)' % (sr['index_load_ms'], sr['keyword_count']))
-            print('  全扫描 (5 query): %.2f ms' % sr['full_scan_5q_ms'])
-            print('  倒排索引构建: %.0f ms (首次)' % sr['inverted_build_ms'])
-            print('  缓存加载: %.2f ms (热)' % sr['inverted_cache_load_ms'])
-            print('  倒排搜索 (5 query): %.2f ms' % sr['inverted_scan_5q_ms'])
-            print('  提速: %.1fx' % sr['speedup'])
-
-        if not args.json:
-            print('\n--- CN_MAP ---')
-        cr = bench_cnmap()
-        report['cnmap'] = cr
-        if not args.json:
-            print('  静态词: %d, 路径映射: %d' % (cr['cn_map_static_count'], cr['business_path_map_count']))
-            print('  auto-expand: %.0f ms, 总计 %d 词' % (cr['auto_expand_ms'], cr['cn_map_total_count']))
-
-    if not args.build:
-        if not args.json:
-            print('\n--- 索引大小 ---')
-        sz = bench_index_sizes()
-        report['index_sizes'] = sz
-        if not args.json:
-            for name, size in sorted(sz.items(), key=lambda x: -x[1]):
-                print('  %-30s %s' % (name, format_size(size)))
-
-    if args.json:
-        print(json.dumps(report, indent=2, ensure_ascii=False))
-    else:
-        print('\n' + '=' * 50)
-
-
-def format_size(n):
-    for unit in ['B', 'KB', 'MB']:
-        if n < 1024:
-            return '%.1f %s' % (n, unit)
-        n /= 1024
-    return '%.1f GB' % n
-
-
-if __name__ == '__main__':
-    main()
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""
+benchmark.py - QODER 流水线性能基准 (性能优化 A7)
+
+测量关键操作的延迟和内存, 供优化前后对比。
+
+Usage:
+    python benchmark.py                    # 跑全部基准
+    python benchmark.py --search           # 只测搜索
+    python benchmark.py --build            # 只测构建 (耗时)
+    python benchmark.py --json             # 输出 JSON (供对比)
+
+指标:
+    - search_cold: 冷启动搜索 (Python 启动 + JSON 加载 + 扫描)
+    - search_warm: 热搜索 (倒排索引缓存命中)
+    - index_load: code-keyword.json 加载时间
+    - inverted_build: 倒排索引构建时间
+    - inverted_cache_load: 缓存加载时间
+    - cnmap_lookup: CN_MAP 查询 (含 auto-expand)
+"""
+
+import argparse
+from common.paths import BUILD_CACHE_DIR
+import json
+import os
+import sys
+import time
+from datetime import datetime
+
+try:
+    sys.stdout.reconfigure(encoding='utf-8', errors='replace')
+except (AttributeError, TypeError, OSError):
+    pass
+
+THIS_DIR = os.path.dirname(os.path.abspath(__file__))  # .qoder/scripts
+sys.path.insert(0, THIS_DIR)
+# BASE = repo root = scripts 的上两级 (.qoder/scripts -> .qoder -> repo)
+BASE = os.path.dirname(os.path.dirname(THIS_DIR))
+
+INDEX_DIR = os.path.join(BASE, 'data', 'index')
+
+
+def measure(fn, repeat=1):
+    """测函数执行时间, 返回 (avg_ms, min_ms, max_ms)。"""
+    times = []
+    for _ in range(repeat):
+        t0 = time.perf_counter()
+        fn()
+        times.append((time.perf_counter() - t0) * 1000)
+    return sum(times) / len(times), min(times), max(times)
+
+
+def bench_search():
+    """搜索基准: 冷启动 + 热搜索对比。"""
+    results = {}
+
+    # 1. JSON 加载 (冷)
+    ki_path = os.path.join(BUILD_CACHE_DIR, 'code-keyword.json')
+    def load_ki():
+        with open(ki_path, encoding='utf-8') as f:
+            return json.load(f)
+    results['index_load_ms'] = measure(load_ki)[0]
+    ki = load_ki()
+    results['keyword_count'] = len(ki)
+
+    # 2. 全扫描 (旧逻辑)
+    from search_index import _match_keyword
+    queries = ['insurance', 'attendance', 'salary', 'vehicle', 'user']
+    def full_scan():
+        for q in queries:
+            for kw in ki:
+                if _match_keyword(q, kw):
+                    pass
+    results['full_scan_5q_ms'] = measure(full_scan, repeat=5)[0] / 5
+
+    # 3. 倒排索引构建 (首次)
+    from common.search_engine import get_inverted_index, search_keywords_fast, build_inverted_index
+    # 清缓存测构建
+    cache_path = os.path.join(INDEX_DIR, '.inverted-cache.json')
+    if os.path.exists(cache_path):
+        os.remove(cache_path)
+    results['inverted_build_ms'] = measure(lambda: get_inverted_index(BASE))[0]
+
+    # 4. 缓存加载 (热)
+    results['inverted_cache_load_ms'] = measure(lambda: get_inverted_index(BASE), repeat=5)[0]
+
+    # 5. 倒排搜索 (热)
+    inverted = get_inverted_index(BASE)
+    def inverted_search():
+        for q in queries:
+            search_keywords_fast([q], inverted, ki)
+    results['inverted_scan_5q_ms'] = measure(inverted_search, repeat=5)[0] / 5
+
+    # 6. 提速比
+    if results['inverted_scan_5q_ms'] > 0:
+        results['speedup'] = round(results['full_scan_5q_ms'] / results['inverted_scan_5q_ms'], 1)
+    else:
+        results['speedup'] = float('inf')
+
+    return results
+
+
+def bench_cnmap():
+    """CN_MAP 查询基准。"""
+    from common.terms import CN_MAP, BUSINESS_PATH_MAP, get_cn_map_with_auto
+    results = {}
+    results['cn_map_static_count'] = len(CN_MAP)
+    results['business_path_map_count'] = len(BUSINESS_PATH_MAP)
+
+    # auto-expand (含 PRD 扫描)
+    def auto_expand():
+        return get_cn_map_with_auto()
+    results['auto_expand_ms'] = measure(auto_expand)[0]
+    results['cn_map_total_count'] = len(get_cn_map_with_auto())
+
+    # 查询扩展
+    test_terms = ['考勤', '保险', '库存', '采购', '发票']
+    def expand_query():
+        out = []
+        for t in test_terms:
+            if t in CN_MAP:
+                out.extend(CN_MAP[t].split())
+        return out
+    results['query_expand_5q_ms'] = measure(expand_query, repeat=10)[0] / 10
+
+    return results
+
+
+def bench_index_sizes():
+    """索引文件大小。"""
+    results = {}
+    for name in ['code-keyword.json', 'code-api.json', 'ui-style.json',
+                 'prd-index.json', 'code-stats.json', 'prd-code-map.json']:
+        p = os.path.join(INDEX_DIR, name)
+        if os.path.isfile(p):
+            results[name] = os.path.getsize(p)
+    # 缓存文件
+    for name in ['.inverted-cache.json', '.file-keys.json']:
+        p = os.path.join(INDEX_DIR, name)
+        if os.path.isfile(p):
+            results[name] = os.path.getsize(p)
+    return results
+
+
+def main():
+    parser = argparse.ArgumentParser(description='QODER 性能基准')
+    parser.add_argument('--search', action='store_true', help='只测搜索')
+    parser.add_argument('--build', action='store_true', help='测构建 (耗时)')
+    parser.add_argument('--json', action='store_true', help='JSON 输出')
+    args = parser.parse_args()
+
+    report = {
+        'timestamp': datetime.now().isoformat(timespec='seconds'),
+        'python': sys.version.split()[0],
+        'platform': sys.platform,
+        'base': os.path.basename(BASE),
+    }
+
+    if not args.build:
+        print('=' * 50)
+        print('QODER 性能基准')
+        print('=' * 50)
+
+    if not args.build or args.search:
+        if not args.json:
+            print('\n--- 搜索性能 ---')
+        sr = bench_search()
+        report['search'] = sr
+        if not args.json:
+            print('  keyword-index 加载: %.0f ms (%d keys)' % (sr['index_load_ms'], sr['keyword_count']))
+            print('  全扫描 (5 query): %.2f ms' % sr['full_scan_5q_ms'])
+            print('  倒排索引构建: %.0f ms (首次)' % sr['inverted_build_ms'])
+            print('  缓存加载: %.2f ms (热)' % sr['inverted_cache_load_ms'])
+            print('  倒排搜索 (5 query): %.2f ms' % sr['inverted_scan_5q_ms'])
+            print('  提速: %.1fx' % sr['speedup'])
+
+        if not args.json:
+            print('\n--- CN_MAP ---')
+        cr = bench_cnmap()
+        report['cnmap'] = cr
+        if not args.json:
+            print('  静态词: %d, 路径映射: %d' % (cr['cn_map_static_count'], cr['business_path_map_count']))
+            print('  auto-expand: %.0f ms, 总计 %d 词' % (cr['auto_expand_ms'], cr['cn_map_total_count']))
+
+    if not args.build:
+        if not args.json:
+            print('\n--- 索引大小 ---')
+        sz = bench_index_sizes()
+        report['index_sizes'] = sz
+        if not args.json:
+            for name, size in sorted(sz.items(), key=lambda x: -x[1]):
+                print('  %-30s %s' % (name, format_size(size)))
+
+    if args.json:
+        print(json.dumps(report, indent=2, ensure_ascii=False))
+    else:
+        print('\n' + '=' * 50)
+
+
+def format_size(n):
+    for unit in ['B', 'KB', 'MB']:
+        if n < 1024:
+            return '%.1f %s' % (n, unit)
+        n /= 1024
+    return '%.1f GB' % n
+
+
+if __name__ == '__main__':
+    main()