feat(env): 更新目标库配置及添加 PostgreSQL 环境变量

- 目标库名称由 hk_songs 修改为 hk_songs_test，新增中间表 hk_songs_import_staging - 添加测试环境 crawler_dev PostgreSQL 连接配置变量 - 添加正式环境 archive_crawler PostgreSQL 连接配置变量 refactor(check_record_titles): 调整平台原始标题核对逻辑 - 变更核对数据库源为 archive_crawler 与 hikoon-data-spider - 按平台分表批量查询 spider 原始标题替代原录音名查询 - 合并结果中使用 spider_title 与 crawler_title 精确比对 - 输出 Excel 标题及字段相应更新，匹配时整行高亮 - 修改脚本参数说明和默认批处理数量描述 - 日志及信息输出文本调整以反映新数据源 feat(fix_record_titles): 新增基于 spider 标题的归一化批量修复工具 - 支持从 archive_crawler 读取待修复记录的推送关系 - 结合 spider 标题，计算归一化标题和版本号 - 支持对不同平台数据批量生成更新 SQL 预览或应用 - 支持备份更新前数据，支持安全回退操作 - 导出更新报告 Excel，突出显示变化字段 - 支持长文本分批过滤，避免数据库字段长度异常 - 增加命令行参数控制应用范围、批量大小及排除标题列表 - 使用严格的 SQL 语句构造与变更检测优化更新效率 - 实现基于 run_id 的多表回滚机制，保证数据一致性

feat(env): 更新目标库配置及添加 PostgreSQL 环境变量
- 目标库名称由 hk_songs 修改为 hk_songs_test，新增中间表 hk_songs_import_staging - 添加测试环境 crawler_dev PostgreSQL 连接配置变量 - 添加正式环境 archive_crawler PostgreSQL 连接配置变量 refactor(check_record_titles): 调整平台原始标题核对逻辑 - 变更核对数据库源为 archive_crawler 与 hikoon-data-spider - 按平台分表批量查询 spider 原始标题替代原录音名查询 - 合并结果中使用 spider_title 与 crawler_title 精确比对 - 输出 Excel 标题及字段相应更新，匹配时整行高亮 - 修改脚本参数说明和默认批处理数量描述 - 日志及信息输出文本调整以反映新数据源 feat(fix_record_titles): 新增基于 spider 标题的归一化批量修复工具 - 支持从 archive_crawler 读取待修复记录的推送关系 - 结合 spider 标题，计算归一化标题和版本号 - 支持对不同平台数据批量生成更新 SQL 预览或应用 - 支持备份更新前数据，支持安全回退操作 - 导出更新报告 Excel，突出显示变化字段 - 支持长文本分批过滤，避免数据库字段长度异常 - 增加命令行参数控制应用范围、批量大小及排除标题列表 - 使用严格的 SQL 语句构造与变更检测优化更新效率 - 实现基于 run_id 的多表回滚机制，保证数据一致性
沈秋雨
Commit ae633f56 ... ae633f56f1aa24cc8606645f550c60decbb50c25 authored 2026-07-17 13:07:34 +0800 by 沈秋雨
Showing 14 changed files with 579 additions and 118 deletions
.env.example
check_record_titles.py
etl_to_crawler/config.py
etl_to_crawler/connections.py
etl_to_crawler/runner.py
fix_record_titles.py
fix_title_version.py
gen_gap_xlsx.py
python/title-norm/tests/test_title_norm.py
python/title-norm/title_norm/__init__.py
sync_singer_hot_songs.py
test_check_record_titles.py
test_fix_record_titles.py
tests/test_runner.py
--- a/.env.example
View file @ae633f5
+++ b/.env.example
View file @ae633f5
@@ -5,13 +5,14 @@ SOURCE_DB_USER=
 SOURCE_DB_PASSWORD=
 SOURCE_DB_NAME=
-# 目标库 - 词曲库
+# 目标库 - 词曲测试
 TARGET_DB_HOST=
 TARGET_DB_PORT=3306
 TARGET_DB_USER=
 TARGET_DB_PASSWORD=
 TARGET_DB_NAME=
-TARGET_TABLE_NAME=hk_songs
+TARGET_TABLE_NAME=hk_songs_test
+TARGET_TABLE_NAME_TMP=hk_songs_import_staging
 # run_etl.py 默认读取并回写的导入状态表：yinyan_song_records 或 yinyan_song_records2
 YINYAN_IMPORT_TABLE=yinyan_song_records
@@ -22,3 +23,21 @@ OSS_ACCESS_KEY_SECRET=
 OSS_ENDPOINT=
 OSS_BUCKET_NAME=
 OSS_FILE_BASE_NAME=
+# 测试环境 - crawler_dev PostgreSQL
+TEST_CRAWLER_DB_HOST=
+TEST_CRAWLER_DB_PORT=5432
+TEST_CRAWLER_DB_USER=
+TEST_CRAWLER_DB_PASSWORD=
+TEST_CRAWLER_DB_NAME=crawler_dev
+TEST_CRAWLER_DB_SSL=false
+TEST_ARCHIVE_DATA_DB_NAME=data_dev
+# 正式环境 - archive_crawler PostgreSQL
+ARCHIVE_CRAWLER_DB_HOST=
+ARCHIVE_CRAWLER_DB_PORT=5432
+ARCHIVE_CRAWLER_DB_USER=
+ARCHIVE_CRAWLER_DB_PASSWORD=
+ARCHIVE_CRAWLER_DB_NAME=archive_crawler
+ARCHIVE_CRAWLER_DB_SSL=false
+ARCHIVE_DATA_DB_NAME=archive_data
--- a/check_record_titles.py
View file @ae633f5
+++ b/check_record_titles.py
View file @ae633f5
 #!/usr/bin/env python3
-"""核对音眼录音名与 crawler 平台歌曲标题，并输出 Excel。
+"""核对 spider 平台原始标题与正式 crawler 歌曲标题，并输出 Excel。
 数据链路：
-    CRAWLER_DB.yinyan_song_records.record_id
+    archive_crawler.yinyan_song_records.(platform, platform_song_id)
-      -> SOURCE_DB.hk_music_record.id -> record_name
+      -> hikoon-data-spider.media_*_songs.id -> spider_title
-    CRAWLER_DB.yinyan_song_records.(platform, platform_song_id)
+      -> archive_crawler.crawler_*_songs.platform_song_id -> crawler_title
-      -> 对应 crawler_*_songs.platform_song_id -> title
 用法：
    .venv/bin/python check_record_titles.py
@@ -21,7 +20,7 @@ from openpyxl import Workbook
 from openpyxl.styles import Alignment, Font, PatternFill
 from openpyxl.utils import get_column_letter
-from etl_to_crawler.connections import close_all_pools, get_pg_conn, get_source_conn
+from etl_to_crawler.connections import close_all_pools, get_archive_crawler_conn, get_spider_conn
 log = logging.getLogger(__name__)
@@ -29,6 +28,11 @@ log = logging.getLogger(__name__)
 DEFAULT_OUTPUT = "record_title_check.xlsx"
 DEFAULT_BATCH_SIZE = 2000
 ALLOWED_IMPORT_TABLES = {"yinyan_song_records", "yinyan_song_records2"}
+SPIDER_TABLES = {
+    "1": "media_tencent_songs",
+    "2": "media_ku_gou_songs",
+    "4": "media_netease_songs",
+}
 HEADER_FILL = PatternFill(fill_type="solid", fgColor="4472C4")
 MATCH_FILL = PatternFill(fill_type="solid", fgColor="C6EFCE")
@@ -81,67 +85,77 @@ def fetch_crawler_rows(pg_conn, import_table: str) -> list[dict]:
            "platform": row[1],
            "platform_song_id": row[2],
            "recording_id": row[3],
-            "title": row[4],
+            "crawler_title": row[4],
        }
        for row in rows
    ]
-def fetch_record_names(mysql_conn, record_ids: Iterable[int], batch_size: int) -> dict[int, str | None]:
+def fetch_spider_titles(mysql_conn, crawler_rows: list[dict], batch_size: int) -> dict[tuple[str, int], str]:
-    """按 record_id 分批读取 hk_music_record.record_name。"""
+    """按 platform_song_id 分平台批量读取 spider 原始 title。"""
-    unique_ids = sorted({int(record_id) for record_id in record_ids if record_id is not None})
+    titles: dict[tuple[str, int], str] = {}
-    names: dict[int, str | None] = {}
+    for platform, table in SPIDER_TABLES.items():
+        ids = sorted({
-    for batch in chunked(unique_ids, batch_size):
+            int(row["platform_song_id"])
-        placeholders = ", ".join(["%s"] * len(batch))
+            for row in crawler_rows
-        sql = f"""
+            if str(row["platform"]) == platform and row["platform_song_id"] is not None
-            SELECT id, record_name
+        })
-            FROM hk_music_record
+        for batch in chunked(ids, batch_size):
-            WHERE id IN ({placeholders})
+            placeholders = ", ".join(["%s"] * len(batch))
-        """
+            with mysql_conn.cursor() as cursor:
-        with mysql_conn.cursor() as cursor:
+                cursor.execute(
-            cursor.execute(sql, tuple(batch))
+                    f"SELECT id, title FROM {table} WHERE id IN ({placeholders})",
-            for row in cursor.fetchall():
+                    tuple(batch),
-                names[int(row["id"])] = row["record_name"]
+                )
+                for row in cursor.fetchall():
-    return names
+                    if row["title"] is not None:
+                        titles[(platform, int(row["id"]))] = str(row["title"])
+    return titles
-def combine_rows(crawler_rows: list[dict], record_names: dict[int, str | None]) -> list[dict]:
+def combine_rows(crawler_rows: list[dict], spider_titles: dict[tuple[str, int], str]) -> list[dict]:
    """合并两个数据库的查询结果，并计算是否完全相同。"""
    result = []
    for row in crawler_rows:
-        record_id = row["record_id"]
+        platform = str(row["platform"])
-        record_name = record_names.get(int(record_id)) if record_id is not None else None
+        platform_song_id = row["platform_song_id"]
-        title = row["title"]
+        spider_title = (
+            spider_titles.get((platform, int(platform_song_id)))
+            if platform_song_id is not None else None
+        )
+        crawler_title = row["crawler_title"]
        result.append(
            {
-                "record_id": record_id,
+                "record_id": row["record_id"],
-                "record_name": record_name,
+                "spider_title": spider_title,
-                "platform": row["platform"],
+                "crawler_title": crawler_title,
-                "platform_song_id": row["platform_song_id"],
+                "platform": platform,
+                "platform_song_id": platform_song_id,
                "recording_id": row["recording_id"],
-                "title": title,
                # 两边都必须有值；不做 trim、大小写或繁简转换，按库中原值精确比较。
-                "matched": record_name is not None and title is not None and record_name == title,
+                "matched": (
+                    spider_title is not None
+                    and crawler_title is not None
+                    and spider_title == crawler_title
+                ),
            }
        )
    return result
 def write_excel(rows: list[dict], output_path: Path) -> None:
-    """写出三列表格；record_name 与 title 相同时整行标绿。"""
+    """写出核对表；spider_title 与 crawler_title 相同时整行标绿。"""
    output_path.parent.mkdir(parents=True, exist_ok=True)
    workbook = Workbook()
    worksheet = workbook.active
-    worksheet.title = "录音名与标题核对"
+    worksheet.title = "Spider与Crawler标题核对"
    worksheet.freeze_panes = "A2"
    headers = (
        "record_id",
-        "record_name",
+        "spider_title",
-        "title",
+        "crawler_title",
        "platform",
        "platform_song_id",
        "recording_id",
@@ -155,8 +169,8 @@ def write_excel(rows: list[dict], output_path: Path) -> None:
    for row_number, row in enumerate(rows, start=2):
        values = (
            row["record_id"],
-            row["record_name"],
+            row["spider_title"],
-            row["title"],
+            row["crawler_title"],
            row["platform"],
            row["platform_song_id"],
            row["recording_id"],
@@ -175,15 +189,15 @@ def write_excel(rows: list[dict], output_path: Path) -> None:
 def parse_args() -> argparse.Namespace:
-    parser = argparse.ArgumentParser(description="核对 hk_music_record.record_name 与 crawler title")
+    parser = argparse.ArgumentParser(description="核对 spider title 与正式 crawler title")
    parser.add_argument("--output", default=DEFAULT_OUTPUT, help=f"Excel 输出路径（默认: {DEFAULT_OUTPUT}）")
    parser.add_argument(
        "--table",
        choices=sorted(ALLOWED_IMPORT_TABLES),
        default="yinyan_song_records",
-        help="待核对的 CRAWLER_DB 关联表",
+        help="待核对的 ARCHIVE_CRAWLER_DB 关联表",
    )
-    parser.add_argument("--batch-size", type=int, default=DEFAULT_BATCH_SIZE, help="MySQL 单批查询数量")
+    parser.add_argument("--batch-size", type=int, default=DEFAULT_BATCH_SIZE, help="spider MySQL 单批查询数量")
    args = parser.parse_args()
    if args.batch_size <= 0:
        parser.error("--batch-size 必须大于 0")
@@ -195,28 +209,24 @@ def main() -> None:
    output_path = Path(args.output).expanduser().resolve()
    try:
-        pg_conn = get_pg_conn()
+        pg_conn = get_archive_crawler_conn()
-        mysql_conn = get_source_conn()
+        mysql_conn = get_spider_conn()
        crawler_rows = fetch_crawler_rows(pg_conn, args.table)
        log.info("从 %s 读取 %d 条关联记录", args.table, len(crawler_rows))
-        record_names = fetch_record_names(
+        spider_titles = fetch_spider_titles(mysql_conn, crawler_rows, args.batch_size)
-            mysql_conn,
+        log.info("从 hikoon-data-spider 读取 %d 个平台标题", len(spider_titles))
-            (row["record_id"] for row in crawler_rows),
-            args.batch_size,
-        )
-        log.info("从 hk_music_record 读取 %d 个录音名", len(record_names))
-        rows = combine_rows(crawler_rows, record_names)
+        rows = combine_rows(crawler_rows, spider_titles)
        write_excel(rows, output_path)
        matched = sum(row["matched"] for row in rows)
-        missing_name = sum(row["record_name"] is None for row in rows)
+        missing_spider = sum(row["spider_title"] is None for row in rows)
-        missing_title = sum(row["title"] is None for row in rows)
+        missing_crawler = sum(row["crawler_title"] is None for row in rows)
        log.info(
-            "核对完成：总数=%d，相同=%d，不同=%d，缺少 record_name=%d，缺少 title=%d",
+            "核对完成：总数=%d，相同=%d，不同=%d，缺少 spider_title=%d，缺少 crawler_title=%d",
-            len(rows), matched, len(rows) - matched, missing_name, missing_title,
+            len(rows), matched, len(rows) - matched, missing_spider, missing_crawler,
        )
        log.info("Excel 已输出至 %s", output_path)
    finally:
--- a/etl_to_crawler/config.py
View file @ae633f5
+++ b/etl_to_crawler/config.py
View file @ae633f5
@@ -21,13 +21,32 @@ HK_SONGS_DB = {
    'charset': 'utf8mb4',
 }
-CRAWLER_DB = {
+TEST_CRAWLER_DB = {
-    'host': os.environ['CRAWLER_DB_HOST'],
+    'host': os.environ['TEST_CRAWLER_DB_HOST'],
-    'port': int(os.environ.get('CRAWLER_DB_PORT', '5432')),
+    'port': int(os.environ.get('TEST_CRAWLER_DB_PORT', '5432')),
-    'user': os.environ['CRAWLER_DB_USER'],
+    'user': os.environ['TEST_CRAWLER_DB_USER'],
-    'password': os.environ['CRAWLER_DB_PASSWORD'],
+    'password': os.environ['TEST_CRAWLER_DB_PASSWORD'],
-    'database': os.environ['CRAWLER_DB_NAME'],
+    'database': os.environ['TEST_CRAWLER_DB_NAME'],
-    'ssl_context': None if os.environ.get('CRAWLER_DB_SSL', 'false').lower() != 'true' else True,
+    'ssl_context': None if os.environ.get('TEST_CRAWLER_DB_SSL', 'false').lower() != 'true' else True,
+}
+TEST_ARCHIVE_DATA_DB = {
+    **TEST_CRAWLER_DB,
+    'database': os.environ.get('TEST_ARCHIVE_DATA_DB_NAME', 'data_dev'),
+}
+ARCHIVE_CRAWLER_DB = {
+    'host': os.environ['ARCHIVE_CRAWLER_DB_HOST'],
+    'port': int(os.environ.get('ARCHIVE_CRAWLER_DB_PORT', '5432')),
+    'user': os.environ['ARCHIVE_CRAWLER_DB_USER'],
+    'password': os.environ['ARCHIVE_CRAWLER_DB_PASSWORD'],
+    'database': os.environ.get('ARCHIVE_CRAWLER_DB_NAME', 'archive_crawler'),
+    'ssl_context': None if os.environ.get('ARCHIVE_CRAWLER_DB_SSL', 'false').lower() != 'true' else True,
+}
+ARCHIVE_DATA_DB = {
+    **ARCHIVE_CRAWLER_DB,
+    'database': os.environ.get('ARCHIVE_DATA_DB_NAME', 'archive_data'),
 }
 OSS_CONFIG = {
--- a/etl_to_crawler/connections.py
View file @ae633f5
+++ b/etl_to_crawler/connections.py
View file @ae633f5
@@ -3,7 +3,16 @@ import pymysql
 import pymysql.cursors
 import pg8000
 import oss2
-from .config import SOURCE_DB, HK_SONGS_DB, CRAWLER_DB, OSS_CONFIG, OSS_CONNECTION_POOL_SIZE
+from .config import (
+    ARCHIVE_CRAWLER_DB,
+    ARCHIVE_DATA_DB,
+    SOURCE_DB,
+    HK_SONGS_DB,
+    TEST_CRAWLER_DB,
+    TEST_ARCHIVE_DATA_DB,
+    OSS_CONFIG,
+    OSS_CONNECTION_POOL_SIZE,
+)
 CONN_POOL_SIZE = 4
@@ -238,7 +247,10 @@ class _PgConnPool:
 _source_pool = None
 _hk_songs_pool = None
 _spider_pool = None
-_pg_pool = None
+_test_crawler_pool = None
+_test_archive_pool = None
+_archive_pool = None
+_archive_crawler_pool = None
 def _get_source_pool() -> _MySqlConnPool:
@@ -264,11 +276,32 @@ def _get_spider_pool() -> _MySqlConnPool:
    return _spider_pool
-def _get_pg_pool() -> _PgConnPool:
+def _get_test_crawler_pool() -> _PgConnPool:
-    global _pg_pool
+    global _test_crawler_pool
-    if _pg_pool is None:
+    if _test_crawler_pool is None:
-        _pg_pool = _PgConnPool(CRAWLER_DB)
+        _test_crawler_pool = _PgConnPool(TEST_CRAWLER_DB)
-    return _pg_pool
+    return _test_crawler_pool
+def _get_test_archive_pool() -> _PgConnPool:
+    global _test_archive_pool
+    if _test_archive_pool is None:
+        _test_archive_pool = _PgConnPool(TEST_ARCHIVE_DATA_DB)
+    return _test_archive_pool
+def _get_archive_pool() -> _PgConnPool:
+    global _archive_pool
+    if _archive_pool is None:
+        _archive_pool = _PgConnPool(ARCHIVE_DATA_DB)
+    return _archive_pool
+def _get_archive_crawler_pool() -> _PgConnPool:
+    global _archive_crawler_pool
+    if _archive_crawler_pool is None:
+        _archive_crawler_pool = _PgConnPool(ARCHIVE_CRAWLER_DB)
+    return _archive_crawler_pool
 def get_source_conn() -> pymysql.Connection:
@@ -283,8 +316,20 @@ def get_hk_songs_conn() -> pymysql.Connection:
    return _get_hk_songs_pool().get()
-def get_pg_conn() -> _NulSafePgConnection:
+def get_test_crawler_conn() -> _NulSafePgConnection:
-    return _get_pg_pool().get()
+    return _get_test_crawler_pool().get()
+def get_test_archive_conn() -> _NulSafePgConnection:
+    return _get_test_archive_pool().get()
+def get_archive_conn() -> _NulSafePgConnection:
+    return _get_archive_pool().get()
+def get_archive_crawler_conn() -> _NulSafePgConnection:
+    return _get_archive_crawler_pool().get()
 def get_oss_bucket() -> oss2.Bucket:
@@ -298,18 +343,24 @@ def close_all_pools():
    for pool in (_source_pool, _hk_songs_pool, _spider_pool):
        if pool is not None:
            pool.close_all()
-    if _pg_pool is not None:
+    if _test_crawler_pool is not None:
-        _pg_pool.close_all()
+        _test_crawler_pool.close_all()
+    if _test_archive_pool is not None:
+        _test_archive_pool.close_all()
+    if _archive_pool is not None:
+        _archive_pool.close_all()
+    if _archive_crawler_pool is not None:
+        _archive_crawler_pool.close_all()
 def refresh_conn(conn, pool_type: str):
    """验证连接健康度，断开时返回新连接，正常时原样返回。
-    用法：在批次循环开头调用，如 ``conn = refresh_conn(conn, 'pg')``。
+    用法：在批次循环开头调用，如 ``conn = refresh_conn(conn, 'test_crawler')``。
-    pool_type 取 'source' | 'hk_songs' | 'spider' | 'pg'。
+    pool_type 取 'source' | 'hk_songs' | 'spider' | 'test_crawler'。
    """
    try:
-        if pool_type == 'pg':
+        if pool_type == 'test_crawler':
            conn.run("SELECT 1")
            return conn
        # MySQL
@@ -322,7 +373,7 @@ def refresh_conn(conn, pool_type: str):
        'source': _get_source_pool,
        'hk_songs': _get_hk_songs_pool,
        'spider': _get_spider_pool,
-        'pg': _get_pg_pool,
+        'test_crawler': _get_test_crawler_pool,
    }
    pool = pools.get(pool_type)
    if pool is None:
--- a/etl_to_crawler/runner.py
View file @ae633f5
+++ b/etl_to_crawler/runner.py
View file @ae633f5
@@ -8,7 +8,7 @@ from .config import (
    PLATFORM_QQ, PLATFORM_KUGOU, PLATFORM_NETEASE, BATCH_SIZE, BACKFILL_BATCH_SIZE,
    OSS_CONFIG, YINYAN_IMPORT_TABLE,
 )
-from .connections import get_hk_songs_conn, get_source_conn, get_spider_conn, get_pg_conn, get_oss_bucket, refresh_conn, close_all_pools
+from .connections import get_hk_songs_conn, get_source_conn, get_spider_conn, get_test_crawler_conn, get_oss_bucket, refresh_conn, close_all_pools
 from .reader import (
    iter_hk_songs_batches,
    fetch_hk_songs_by_source_ids,
@@ -1337,7 +1337,7 @@ def _pick_record_with_singer(records: list[dict], spider_conn) -> dict | None:
 def initialize_yinyan_song_records(platforms: list[str], max_batches: int | None = None) -> None:
    hk_conn = get_hk_songs_conn()
    src_conn = get_source_conn()
-    pg_conn = get_pg_conn()
+    pg_conn = get_test_crawler_conn()
    # 查询已存在的 song_id，用于跳过已处理的记录
    with pg_conn.cursor() as pg_cur:
@@ -1351,7 +1351,7 @@ def initialize_yinyan_song_records(platforms: list[str], max_batches: int | None
            # 长任务连接可能断开，每批次开始前刷新
            hk_conn = refresh_conn(hk_conn, 'hk_songs')
            src_conn = refresh_conn(src_conn, 'source')
-            pg_conn = refresh_conn(pg_conn, 'pg')
+            pg_conn = refresh_conn(pg_conn, 'test_crawler')
            if max_batches is not None and i >= max_batches:
                break
@@ -1416,7 +1416,7 @@ def initialize_yinyan_song_records2(platforms: list[str], max_batches: int | Non
    后续导入允许歌曲的 singers 为空。
    """
    src_conn = get_source_conn()
-    pg_conn = get_pg_conn()
+    pg_conn = get_test_crawler_conn()
    total_inserted = 0
    try:
@@ -1427,7 +1427,7 @@ def initialize_yinyan_song_records2(platforms: list[str], max_batches: int | Non
        for index, start in enumerate(tqdm(range(0, len(song_ids), BATCH_SIZE), desc='init-yinyan-records2')):
            # 长任务连接可能断开，每批次开始前刷新
            src_conn = refresh_conn(src_conn, 'source')
-            pg_conn = refresh_conn(pg_conn, 'pg')
+            pg_conn = refresh_conn(pg_conn, 'test_crawler')
            if max_batches is not None and index >= max_batches:
                break
@@ -1467,7 +1467,7 @@ def initialize_yinyan_song_records2(platforms: list[str], max_batches: int | Non
 def backfill_yinyan_record_platforms(max_batches: int | None = None) -> None:
    src_conn = get_source_conn()
-    pg_conn = get_pg_conn()
+    pg_conn = get_test_crawler_conn()
    total = 0
    skipped = 0
@@ -1477,7 +1477,7 @@ def backfill_yinyan_record_platforms(max_batches: int | None = None) -> None:
        while max_batches is None or batch_index < max_batches:
            # 长任务连接可能断开，每批次开始前刷新
            src_conn = refresh_conn(src_conn, 'source')
-            pg_conn = refresh_conn(pg_conn, 'pg')
+            pg_conn = refresh_conn(pg_conn, 'test_crawler')
            with pg_conn.cursor() as pg_cur:
                rows = fetch_yinyan_records_missing_platform(pg_cur, BACKFILL_BATCH_SIZE)
@@ -1523,7 +1523,7 @@ def _backfill_qq_singers(spider_conn, pg_conn, bucket, base_url, max_batches):
    while max_batches is None or batch_index < max_batches:
        # 长任务连接可能断开，每批次开始前刷新
        spider_conn = refresh_conn(spider_conn, 'spider')
-        pg_conn = refresh_conn(pg_conn, 'pg')
+        pg_conn = refresh_conn(pg_conn, 'test_crawler')
        with pg_conn.cursor() as pg_cur:
            rows = fetch_qq_songs_missing_singers(pg_cur, BATCH_SIZE)
@@ -1603,7 +1603,7 @@ def _backfill_kugou_singers(spider_conn, pg_conn, bucket, base_url, max_batches)
    while max_batches is None or batch_index < max_batches:
        # 长任务连接可能断开，每批次开始前刷新
        spider_conn = refresh_conn(spider_conn, 'spider')
-        pg_conn = refresh_conn(pg_conn, 'pg')
+        pg_conn = refresh_conn(pg_conn, 'test_crawler')
        with pg_conn.cursor() as pg_cur:
            rows = fetch_kugou_songs_missing_singers(pg_cur, BATCH_SIZE)
@@ -1677,7 +1677,7 @@ def _backfill_netease_singers(spider_conn, pg_conn, bucket, base_url, max_batche
    while max_batches is None or batch_index < max_batches:
        # 长任务连接可能断开，每批次开始前刷新
        spider_conn = refresh_conn(spider_conn, 'spider')
-        pg_conn = refresh_conn(pg_conn, 'pg')
+        pg_conn = refresh_conn(pg_conn, 'test_crawler')
        with pg_conn.cursor() as pg_cur:
            rows = fetch_netease_songs_missing_singers(pg_cur, BATCH_SIZE)
@@ -1746,7 +1746,7 @@ def _backfill_netease_singers(spider_conn, pg_conn, bucket, base_url, max_batche
 def backfill_empty_singers(platforms: list[str], max_batches: int | None = None) -> None:
    spider_conn = get_spider_conn()
-    pg_conn = get_pg_conn()
+    pg_conn = get_test_crawler_conn()
    bucket = get_oss_bucket()
    base_url = OSS_CONFIG['base_url']
    try:
@@ -1769,7 +1769,7 @@ def backfill_qq_invalid_covers(max_batches: int | None = None) -> None:
    hk_conn = get_hk_songs_conn()
    src_conn = get_source_conn()
    spider_conn = get_spider_conn()
-    pg_conn = get_pg_conn()
+    pg_conn = get_test_crawler_conn()
    bucket = get_oss_bucket()
    base_url = OSS_CONFIG['base_url']
    batch_index = replaced = filtered = retry_later = 0
@@ -1780,7 +1780,7 @@ def backfill_qq_invalid_covers(max_batches: int | None = None) -> None:
            hk_conn = refresh_conn(hk_conn, 'hk_songs')
            src_conn = refresh_conn(src_conn, 'source')
            spider_conn = refresh_conn(spider_conn, 'spider')
-            pg_conn = refresh_conn(pg_conn, 'pg')
+            pg_conn = refresh_conn(pg_conn, 'test_crawler')
            with pg_conn.cursor() as pg_cur:
                invalid_rows = fetch_qq_songs_with_invalid_covers(
@@ -1879,7 +1879,7 @@ def run(
    hk_conn = get_hk_songs_conn()
    src_conn = get_source_conn()
    spider_conn = get_spider_conn()
-    pg_conn = get_pg_conn()
+    pg_conn = get_test_crawler_conn()
    bucket = get_oss_bucket()
    base_url = OSS_CONFIG['base_url']
    total_ok = total_err = 0
@@ -1905,7 +1905,7 @@ def run(
            hk_conn = refresh_conn(hk_conn, 'hk_songs')
            src_conn = refresh_conn(src_conn, 'source')
            spider_conn = refresh_conn(spider_conn, 'spider')
-            pg_conn = refresh_conn(pg_conn, 'pg')
+            pg_conn = refresh_conn(pg_conn, 'test_crawler')
            with pg_conn.cursor() as pg_cur:
                pending_records = fetch_pending_yinyan_song_records(pg_cur, BATCH_SIZE, platforms)
--- a/fix_record_titles.py 0 → 100644
View file @ae633f5
+++ b/fix_record_titles.py 0 → 100644
View file @ae633f5
--- a/fix_title_version.py
View file @ae633f5
+++ b/fix_title_version.py
View file @ae633f5
@@ -15,7 +15,7 @@ import argparse
 import logging
 from etl_to_crawler.config import PLATFORM_QQ, PLATFORM_KUGOU, PLATFORM_NETEASE
-from etl_to_crawler.connections import get_source_conn, get_pg_conn, close_all_pools
+from etl_to_crawler.connections import get_source_conn, get_test_crawler_conn, close_all_pools
 from etl_to_crawler.utils import split_title_version
 logging.basicConfig(level=logging.INFO, format='%(asctime)s %(levelname)s %(message)s')
@@ -134,7 +134,7 @@ def main():
    args = parser.parse_args()
    mysql_conn = get_source_conn()
-    pg_conn = get_pg_conn()
+    pg_conn = get_test_crawler_conn()
    total_updated = 0
    total_diff = 0
--- a/gen_gap_xlsx.py 0 → 100644
View file @ae633f5
+++ b/gen_gap_xlsx.py 0 → 100644
View file @ae633f5
--- a/python/title-norm/tests/test_title_norm.py 0 → 100644
View file @ae633f5
+++ b/python/title-norm/tests/test_title_norm.py 0 → 100644
View file @ae633f5
+import title_norm as tn
+def test_normalize_name():
+    cases = [
+        ("周杰伦", "周杰伦"),
+        (" 周杰伦 ", "周杰伦"),
+        ("Ｇｌａｓｓ　Ａｎｉｍａｌｓ", "glass animals"),
+        ("Jay  Chou", "jay chou"),
+        ("", ""),
+    ]
+    for raw, want in cases:
+        assert tn.normalize_name(raw) == want, raw
+def test_singer_identity_key():
+    assert tn.singer_identity_key("crawler_qqmusic_singers", 162918) == "qqmusic:162918"
+def test_author_identity_key():
+    assert tn.author_identity_key(" 周杰伦 ") == "name:周杰伦"
+def test_platform_from_singer_table():
+    cases = [
+        ("crawler_qqmusic_singers", "qqmusic"),
+        ("crawler_netease_singers", "netease"),
+        ("crawler_kugou_songs", "kugou"),
+    ]
+    for raw, want in cases:
+        assert tn.platform_from_singer_table(raw) == want
+def test_split_names():
+    got = tn.split_names("周杰伦/方文山")
+    assert got == ["周杰伦", "方文山"]
+    assert tn.split_names("") is None
+def test_normalize_title():
+    cases = [
+        ("七里香", "七里香", ""),
+        ("  七里香  ", "七里香", ""),
+        ("七里香(Live)", "七里香", "live"),
+        ("七里香（Live版）", "七里香", "live"),
+        ("七里香 [Remix]", "七里香", "remix"),
+        ("七里香（混音版）", "七里香", "remix"),
+        ("七里香（现场版）", "七里香", "live"),
+        ("七里香(DJ版)", "七里香", "dj版"),
+        ("七里香(ass version)", "七里香", "ass version"),
+        ("ANGEL (天使)", "angel (天使)", ""),
+        ("Radio(Dum-Dum)", "radio(dum-dum)", ""),
+        ("七里香(完整版)", "七里香", ""),
+        ("七里香(原版)", "七里香", ""),
+        ("《七里香》", "七里香", ""),
+        ("《七裡香》（Live）", "七里香", "live"),
+        ("七裡香", "七里香", ""),
+        ("七里香 - DJ版", "七里香", "dj版"),
+        ("七里香 - 周杰伦", "七里香 - 周杰伦", ""),
+        ("七里香—Live", "七里香", "live"),
+        ("七里香(天使)(Live)", "七里香(天使)", "live"),
+        ("七里香 现场版", "七里香", "live"),
+    ]
+    for raw, want_title, want_ver in cases:
+        got_title, got_ver = tn.normalize_title(raw)
+        assert (got_title, got_ver) == (want_title, want_ver), raw
+def test_normalize_name_set():
+    a = tn.normalize_name_set("周杰伦,方文山")
+    b = tn.normalize_name_set("方文山、周杰伦")
+    assert a == b
+    assert tn.normalize_name_set("") is None
+def test_strip_clip_markers():
+    cases = [
+        ("父亲写的散文诗-剪辑版", "父亲写的散文诗"),
+        ("人生路漫漫主歌", "人生路漫漫"),
+        ("Stay with me（氛围beat）-全曲", "Stay with me（氛围beat）"),
+        ("咏春（副歌）", "咏春（副歌）"),
+        ("小气鬼（加速版）", "小气鬼"),
+        ("枪火（剪辑一）", "枪火"),
+        ("恋人", "恋人"),
+    ]
+    for raw, want in cases:
+        assert tn.strip_clip_markers(raw) == want, raw
+def test_is_version_marker():
+    cases = [
+        ("DJ版", True),
+        ("钢琴版", True),
+        ("女生版", True),
+        ("完整版", False),
+        ("原版", False),
+        ("ass version", True),
+        ("TV Version", True),
+        ("Extended Version", True),
+        ("live", True),
+        ("Live", True),
+        ("伴奏", True),
+        ("对唱", True),
+        ("dj", True),
+        ("remix", True),
+        ("0.8x", True),
+        ("1.1X", True),
+        ("×0.93", True),
+        ("0.8倍", True),
+        ("1.2倍速", True),
+        ("天使", False),
+        ("Dum-Dum", False),
+        ("周杰伦", False),
+        ("", False),
+    ]
+    for raw, want in cases:
+        assert tn.is_version_marker(raw) == want, raw
+def test_split_names_clip_authors():
+    cases = [
+        ("唐伯虎Annie、伯爵Johnny", 2),
+        ("梨香JZH & 口古口古", 2),
+        ("派偉俊/mac ova sea", 2),
+        ("李荣浩", 1),
+    ]
+    for raw, want in cases:
+        got = tn.split_names(raw)
+        assert len(got or []) == want, raw
+def test_normalize_set():
+    a = tn.normalize_set(["方文山", "周杰伦"])
+    b = tn.normalize_set(["周杰伦", "方文山"])
+    assert a == b
+    assert tn.normalize_set(None) is None
--- a/python/title-norm/title_norm/__init__.py 0 → 100644
View file @ae633f5
+++ b/python/title-norm/title_norm/__init__.py 0 → 100644
View file @ae633f5
--- a/sync_singer_hot_songs.py 0 → 100644
View file @ae633f5
+++ b/sync_singer_hot_songs.py 0 → 100644
View file @ae633f5
--- a/test_check_record_titles.py
View file @ae633f5
+++ b/test_check_record_titles.py
View file @ae633f5
@@ -2,7 +2,7 @@ from pathlib import Path
 from openpyxl import load_workbook
-from check_record_titles import combine_rows, fetch_crawler_rows, fetch_record_names, write_excel
+from check_record_titles import combine_rows, fetch_crawler_rows, fetch_spider_titles, write_excel
 class Cursor:
@@ -42,7 +42,7 @@ def test_fetch_crawler_rows_joins_all_platform_tables():
        "platform": "1",
        "platform_song_id": 101,
        "recording_id": "recording-1",
-        "title": "同名",
+        "crawler_title": "同名",
    }]
    assert "WHERE ysr.is_archive_push = TRUE" in cursor.calls[0][0]
@@ -58,31 +58,35 @@ def test_fetch_crawler_rows_rejects_unknown_table():
        raise AssertionError("应拒绝非白名单表")
-def test_fetch_record_names_batches_queries():
+def test_fetch_spider_titles_batches_queries_by_platform():
    connection = Connection([
-        [{"id": 1, "record_name": "甲"}, {"id": 2, "record_name": "乙"}],
+        [{"id": 101, "title": "甲"}],
-        [{"id": 3, "record_name": "丙"}],
+        [{"id": 102, "title": "乙"}],
    ])
+    crawler_rows = [
+        {"platform": "1", "platform_song_id": 101},
+        {"platform": "2", "platform_song_id": 102},
+    ]
-    names = fetch_record_names(connection, [3, 1, 2, 2], batch_size=2)
+    titles = fetch_spider_titles(connection, crawler_rows, batch_size=2)
-    assert names == {1: "甲", 2: "乙", 3: "丙"}
+    assert titles == {("1", 101): "甲", ("2", 102): "乙"}
 def test_excel_marks_only_exact_non_null_matches_green(tmp_path: Path):
    crawler_rows = [
-        {"record_id": 1, "platform": "1", "platform_song_id": 101, "recording_id": "r1", "title": "相同"},
+        {"record_id": 1, "platform": "1", "platform_song_id": 101, "recording_id": "r1", "crawler_title": "相同"},
-        {"record_id": 2, "platform": "2", "platform_song_id": 102, "recording_id": "r2", "title": " 名称 "},
+        {"record_id": 2, "platform": "2", "platform_song_id": 102, "recording_id": "r2", "crawler_title": " 名称 "},
-        {"record_id": 3, "platform": "4", "platform_song_id": 103, "recording_id": None, "title": None},
+        {"record_id": 3, "platform": "4", "platform_song_id": 103, "recording_id": None, "crawler_title": None},
    ]
-    rows = combine_rows(crawler_rows, {1: "相同", 2: "名称", 3: None})
+    rows = combine_rows(crawler_rows, {("1", 101): "相同", ("2", 102): "名称"})
    output = tmp_path / "check.xlsx"
    write_excel(rows, output)
    worksheet = load_workbook(output).active
    assert [worksheet.cell(1, column).value for column in range(1, 7)] == [
-        "record_id", "record_name", "title", "platform", "platform_song_id", "recording_id",
+        "record_id", "spider_title", "crawler_title", "platform", "platform_song_id", "recording_id",
    ]
    assert worksheet["A2"].fill.fgColor.rgb == "00C6EFCE"
    assert worksheet["A3"].fill.fill_type is None
--- a/test_fix_record_titles.py 0 → 100644
View file @ae633f5
+++ b/test_fix_record_titles.py 0 → 100644
View file @ae633f5
+from openpyxl import load_workbook
+from datetime import datetime, timezone
+from fix_record_titles import (
+    InputRow,
+    _archive_recording_sql,
+    _archive_source_sql,
+    _crawler_sql,
+    _rollback_sql,
+    _row_value,
+    build_backup_records,
+    build_report_rows,
+    build_fix_rows,
+    fetch_input_rows,
+    load_backup,
+    write_backup_atomic,
+    write_report,
+)
+class Cursor:
+    def __init__(self, rows):
+        self.rows = rows
+        self.calls = []
+    def __enter__(self):
+        return self
+    def __exit__(self, exc_type, exc_value, traceback):
+        return False
+    def execute(self, sql, params=None):
+        self.calls.append((sql, params))
+    def fetchall(self):
+        return self.rows
+class Connection:
+    def __init__(self, rows):
+        self.raw_cursor = Cursor(rows)
+    def cursor(self):
+        return self.raw_cursor
+def test_fetch_input_rows_reads_archive_push_relations_directly():
+    connection = Connection([
+        (1, "1", 101, "REC-1"),
+        (2, "2", 102, "REC-2"),
+    ])
+    rows, stats = fetch_input_rows(connection)
+    assert rows == [
+        InputRow(1, "1", 101, "REC-1"),
+        InputRow(2, "2", 102, "REC-2"),
+    ]
+    assert stats == {"total": 2, "missing_key": 0, "unsupported_platform": 0}
+    sql = connection.raw_cursor.calls[0][0]
+    assert "FROM yinyan_song_records" in sql
+    assert "WHERE is_archive_push = TRUE" in sql
+def test_fetch_input_rows_applies_limit():
+    connection = Connection([(1, "1", 101, "REC-1")])
+    fetch_input_rows(connection, limit=10)
+    sql, params = connection.raw_cursor.calls[0]
+    assert "LIMIT %s" in sql
+    assert params == (10,)
+def test_build_fix_rows_uses_spider_title_for_normalization():
+    input_row = InputRow(1, "1", 101, "REC-1")
+    rows, fix_stats = build_fix_rows(
+        [input_row],
+        {("1", 101): "七里香（Live版）"},
+    )
+    assert fix_stats["missing_spider"] == 0
+    assert rows[0].spider_title == "七里香（Live版）"
+    assert rows[0].title_norm == "七里香"
+    assert rows[0].version == "live"
+    assert _row_value(rows[0], "archive_platform") == "qqmusic"
+    assert _row_value(rows[0], "archive_version") == "live"
+def test_build_fix_rows_skips_missing_spider_title():
+    input_row = InputRow(1, "4", 101, "REC-1")
+    rows, fix_stats = build_fix_rows([input_row], {})
+    assert rows == []
+    assert fix_stats["missing_spider"] == 1
+def test_build_fix_rows_temporarily_excludes_exact_spider_title():
+    excluded = 'Waiting On A Wish (From "Disney\'s Snow White"/Soundtrack Version|Reprise)'
+    input_rows = [
+        InputRow(1, "1", 101, "REC-1"),
+        InputRow(2, "1", 102, "REC-2"),
+    ]
+    spider_titles = {
+        ("1", 101): excluded,
+        ("1", 102): "Waiting On A Wish",
+    }
+    rows, fix_stats = build_fix_rows(
+        input_rows,
+        spider_titles,
+        exclude_titles={excluded},
+    )
+    assert [row.platform_song_id for row in rows] == [102]
+    assert fix_stats["excluded_title"] == 1
+def test_update_sql_uses_spider_values_and_only_updates_differences():
+    recording_sql, recording_params = _archive_recording_sql(1, True)
+    source_sql, source_params = _archive_source_sql(1, True)
+    crawler_sql, crawler_params = _crawler_sql("crawler_qqmusic_songs", 1, True)
+    assert recording_params == ["recording_id", "spider_title", "title_norm", "archive_version"]
+    assert "title_norm = source.new_title_norm" in recording_sql
+    assert "IS DISTINCT FROM source.new_title" in recording_sql
+    assert "target.platform_song_id = source.platform_song_id" in source_sql
+    assert "spider_title" in source_params
+    assert crawler_params == ["platform_song_id", "spider_title", "version"]
+    assert "source.platform_song_id::bigint" in crawler_sql
+    assert "version = source.new_version" in crawler_sql
+    assert "RETURNING target.recording_id" in recording_sql
+    assert "RETURNING target.platform_song_id" in crawler_sql
+    locked_sql, _ = _archive_recording_sql(1, False, lock=True)
+    assert "FOR UPDATE OF target" in locked_sql
+def test_report_lists_only_returned_target_keys(tmp_path):
+    fix = build_fix_rows(
+        [InputRow(1, "1", 101, "REC-1")],
+        {("1", 101): "七里香（Live版）"},
+    )[0][0]
+    matched = {
+        "archive_recording": [
+            ("REC-1", "旧标题", "七里香（Live版）", "旧标题", "七里香", None, "live"),
+        ],
+        "archive_recording_platform_source": [
+            ("REC-1", "qqmusic", "101", "旧标题", "七里香（Live版）", None, "七里香", None, "live"),
+        ],
+        "crawler_qqmusic_songs": [
+            (101, "旧标题", "七里香（Live版）", "", "live"),
+        ],
+        "crawler_kugou_songs": [],
+        "crawler_netease_songs": [],
+    }
+    rows = build_report_rows([fix], matched, apply=False)
+    output = tmp_path / "report.xlsx"
+    write_report(output, rows)
+    worksheet = load_workbook(output).active
+    assert len(rows) == 3
+    assert worksheet.max_row == 4
+    assert worksheet["A1"].value == "database"
+    assert worksheet["H1"].value == "old_title"
+    assert worksheet["I1"].value == "new_title"
+    assert worksheet["H2"].value == "旧标题"
+    assert worksheet["I2"].value == "七里香（Live版）"
+    assert worksheet["J3"].value == "<NULL>"
+    assert worksheet["J3"].fill.fgColor.rgb == "00FFC7CE"
+    assert worksheet["K3"].fill.fgColor.rgb == "00C6EFCE"
+    assert worksheet["N2"].value == "would_update"
+def test_backup_round_trip_preserves_null_empty_and_updated_at(tmp_path):
+    fix = build_fix_rows(
+        [InputRow(1, "1", 101, "REC-1")],
+        {("1", 101): "七里香（Live版）"},
+    )[0][0]
+    updated_at = datetime(2026, 7, 15, 10, 0, tzinfo=timezone.utc)
+    preview = {
+        "archive_recording": [
+            ("REC-1", "旧标题", "七里香（Live版）", None, "七里香", None, "live", updated_at),
+        ],
+        "archive_recording_platform_source": [],
+        "crawler_qqmusic_songs": [
+            (101, "旧标题", "七里香（Live版）", "", "live", updated_at.replace(tzinfo=None)),
+        ],
+        "crawler_kugou_songs": [],
+        "crawler_netease_songs": [],
+    }
+    records = build_backup_records([fix], preview, "run-1")
+    path = tmp_path / "backup.jsonl"
+    write_backup_atomic(path, records, "run-1")
+    metadata, loaded = load_backup(path)
+    assert metadata["change_count"] == 2
+    assert loaded[0]["old"]["title_norm"] is None
+    assert loaded[1]["old"]["version"] == ""
+    assert loaded[0]["old"]["updated_at"] == updated_at.isoformat()
+    try:
+        write_backup_atomic(path, records, "run-2")
+    except FileExistsError:
+        pass
+    else:
+        raise AssertionError("已有备份不应被覆盖")
+def test_rollback_sql_restores_old_values_with_new_value_guard():
+    archive_sql, _ = _rollback_sql("archive_recording", 1)
+    crawler_sql, _ = _rollback_sql("crawler_qqmusic_songs", 1)
+    assert "SET title = source.old_title" in archive_sql
+    assert "updated_at = source.old_updated_at::timestamptz" in archive_sql
+    assert "target.title IS NOT DISTINCT FROM source.new_title" in archive_sql
+    assert "target.title_norm IS NOT DISTINCT FROM source.new_title_norm" in archive_sql
+    assert "updated_at = source.old_updated_at::timestamp" in crawler_sql
+    assert "target.version IS NOT DISTINCT FROM source.new_version" in crawler_sql
--- a/tests/test_runner.py
View file @ae633f5
+++ b/tests/test_runner.py
View file @ae633f5
@@ -375,7 +375,7 @@ def test_run_imports_only_pending_yinyan_platform_record(monkeypatch):
    monkeypatch.setattr(runner, 'get_hk_songs_conn', lambda: _Connection())
    monkeypatch.setattr(runner, 'get_source_conn', lambda: _Connection())
    monkeypatch.setattr(runner, 'get_spider_conn', lambda: _Connection())
-    monkeypatch.setattr(runner, 'get_pg_conn', lambda: pg_conn)
+    monkeypatch.setattr(runner, 'get_test_crawler_conn', lambda: pg_conn)
    monkeypatch.setattr(runner, 'get_oss_bucket', lambda: object())
    monkeypatch.setattr(runner, 'refresh_conn', lambda conn, name: conn)
    monkeypatch.setattr(runner, 'fetch_pending_yinyan_song_records', lambda cur, batch_size, platforms: [
@@ -436,7 +436,7 @@ def test_run_retry_failed_resets_failures_and_restores_hk_sources(monkeypatch):
    monkeypatch.setattr(runner, 'get_hk_songs_conn', lambda: hk_conn)
    monkeypatch.setattr(runner, 'get_source_conn', lambda: _Connection())
    monkeypatch.setattr(runner, 'get_spider_conn', lambda: _Connection())
-    monkeypatch.setattr(runner, 'get_pg_conn', lambda: pg_conn)
+    monkeypatch.setattr(runner, 'get_test_crawler_conn', lambda: pg_conn)
    monkeypatch.setattr(runner, 'get_oss_bucket', lambda: object())
    monkeypatch.setattr(runner, 'refresh_conn', lambda conn, name: conn)
    monkeypatch.setattr(runner, 'reset_failed_yinyan_song_records', reset)
@@ -456,7 +456,7 @@ def test_initialize_yinyan_song_records_inserts_primary_records(monkeypatch):
    monkeypatch.setattr(runner, 'get_hk_songs_conn', lambda: _Connection())
    monkeypatch.setattr(runner, 'get_source_conn', lambda: _Connection())
-    monkeypatch.setattr(runner, 'get_pg_conn', lambda: pg_conn)
+    monkeypatch.setattr(runner, 'get_test_crawler_conn', lambda: pg_conn)
    monkeypatch.setattr(runner, 'refresh_conn', lambda conn, name: conn)
    monkeypatch.setattr(runner, 'fetch_existing_yinyan_song_ids', lambda cur: set())
@@ -501,7 +501,7 @@ def test_initialize_yinyan_song_records2_writes_all_relations_then_deduplicates(
    dedupe_calls = []
    monkeypatch.setattr(runner, 'get_source_conn', lambda: _Connection())
-    monkeypatch.setattr(runner, 'get_pg_conn', lambda: pg_conn)
+    monkeypatch.setattr(runner, 'get_test_crawler_conn', lambda: pg_conn)
    monkeypatch.setattr(runner, 'refresh_conn', lambda conn, name: conn)
    monkeypatch.setattr(runner, 'fetch_yinyan_song_ids', lambda cur: [10])
    monkeypatch.setattr(runner, 'fetch_all_song_record_relations', lambda conn, song_ids: [
@@ -532,7 +532,7 @@ def test_initialize_yinyan_song_records2_keeps_rows_without_valid_singers(monkey
    monkeypatch.setattr(runner, 'get_source_conn', lambda: _Connection())
    get_spider_conn = MagicMock(side_effect=AssertionError('records2 init must not query spider'))
    monkeypatch.setattr(runner, 'get_spider_conn', get_spider_conn)
-    monkeypatch.setattr(runner, 'get_pg_conn', lambda: pg_conn)
+    monkeypatch.setattr(runner, 'get_test_crawler_conn', lambda: pg_conn)
    monkeypatch.setattr(runner, 'refresh_conn', lambda conn, name: conn)
    monkeypatch.setattr(runner, 'fetch_yinyan_song_ids', lambda cur: [10])
    monkeypatch.setattr(runner, 'fetch_all_song_record_relations', lambda conn, song_ids: [
@@ -561,7 +561,7 @@ def test_backfill_yinyan_record_platforms_updates_missing_platform_rows(monkeypa
    updated = []
    monkeypatch.setattr(runner, 'get_source_conn', lambda: _Connection())
-    monkeypatch.setattr(runner, 'get_pg_conn', lambda: pg_conn)
+    monkeypatch.setattr(runner, 'get_test_crawler_conn', lambda: pg_conn)
    monkeypatch.setattr(runner, 'refresh_conn', lambda conn, name: conn)
    monkeypatch.setattr(runner, 'fetch_yinyan_records_missing_platform', lambda cur, batch_size: [
        {'song_id': 10, 'record_id': 100},