fix(runner): 调整导入资源必填规则并优化选择逻辑

- 修改 records2 歌曲基础字段要求，封面和歌词允许为空
- 根据导入类型动态设定必需资源，音频始终必需，歌词封面视情况而定
- 调整各平台录音选择函数，支持可选的必需资源参数
- 修改 _require_primary_assets 函数，支持传入必需资源参数校验
- 更新导入流程以使用动态资源需求，避免无必要的拒绝
- 添加单元测试覆盖不同导入表的资源要求和录音候选选择逻辑
- 代码注释及文档更新以反映新的资源校验规则

@@ -56,7 +56,7 @@ if YINYAN_IMPORT_TABLE not in _ALLOWED_YINYAN_IMPORT_TABLES:
         f"got {YINYAN_IMPORT_TABLE!r}"
         f"got {YINYAN_IMPORT_TABLE!r}"
     )
     )
 
 
-BATCH_SIZE = 1000
+BATCH_SIZE = 10
 BACKFILL_BATCH_SIZE = int(os.environ.get('BACKFILL_BATCH_SIZE', '5000'))
 BACKFILL_BATCH_SIZE = int(os.environ.get('BACKFILL_BATCH_SIZE', '5000'))
 HTTP_POOL_MAXSIZE = int(os.environ.get('HTTP_POOL_MAXSIZE', '128'))
 HTTP_POOL_MAXSIZE = int(os.environ.get('HTTP_POOL_MAXSIZE', '128'))
 HTTP_TRANSFER_RETRIES = int(os.environ.get('HTTP_TRANSFER_RETRIES', '3'))
 HTTP_TRANSFER_RETRIES = int(os.environ.get('HTTP_TRANSFER_RETRIES', '3'))

@@ -28,8 +28,7 @@ WHERE deleted = '0'
 ORDER BY id
 ORDER BY id
 """
 """
 
 
-# 用于补全 yinyan_song_records2 的歌曲必须具备所有下游所需字段。
-# 当前 hk_songs_test 中的 lyrics_url 即歌曲链接字段。
+# records2 初始化不要求封面或歌词；正式导入时这两个字段允许为空。
 _HK_SONGS_RECORDS2_QUERY = """
 _HK_SONGS_RECORDS2_QUERY = """
 SELECT id, source_song_id
 SELECT id, source_song_id
 FROM hk_songs_test
 FROM hk_songs_test
@@ -39,7 +38,6 @@ WHERE deleted = '0'
   AND name IS NOT NULL AND TRIM(name) != ''
   AND name IS NOT NULL AND TRIM(name) != ''
   AND song_time IS NOT NULL AND song_time > 0
   AND song_time IS NOT NULL AND song_time > 0
   AND singer IS NOT NULL AND TRIM(singer) != ''
   AND singer IS NOT NULL AND TRIM(singer) != ''
-  AND lyrics_url IS NOT NULL AND TRIM(lyrics_url) != ''
   AND audio_url IS NOT NULL AND TRIM(audio_url) != ''
   AND audio_url IS NOT NULL AND TRIM(audio_url) != ''
   AND issue_time IS NOT NULL
   AND issue_time IS NOT NULL
 ORDER BY id
 ORDER BY id
@@ -150,7 +148,7 @@ def iter_hk_songs_records2_batches(
     batch_size: int,
     batch_size: int,
     start_after_id: int = 0,
     start_after_id: int = 0,
 ) -> Iterator[list[dict]]:
 ) -> Iterator[list[dict]]:
-    """遍历具备 records2 入库必填字段的歌曲。"""
+    """遍历具备 records2 基础入库字段的歌曲，封面和歌词可为空。"""
     last_id = start_after_id
     last_id = start_after_id
     with conn.cursor() as cur:
     with conn.cursor() as cur:
         while True:
         while True:

@@ -4,7 +4,10 @@ import logging
 from concurrent.futures import ThreadPoolExecutor, as_completed
 from concurrent.futures import ThreadPoolExecutor, as_completed
 from tqdm import tqdm
 from tqdm import tqdm
 
 
-from .config import PLATFORM_QQ, PLATFORM_KUGOU, PLATFORM_NETEASE, BATCH_SIZE, BACKFILL_BATCH_SIZE, OSS_CONFIG
+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_pg_conn, get_oss_bucket, refresh_conn, close_all_pools
 from .reader import (
 from .reader import (
     iter_hk_songs_batches,
     iter_hk_songs_batches,
@@ -121,14 +124,19 @@ def _safe_upload_lyric(
         return ''
         return ''
 
 
 
 
-def _require_primary_assets(assets: dict) -> None:
-    """音频、歌曲封面、歌词均为入库必需资源，任一失败即拒绝整条记录。"""
+def _required_import_assets() -> tuple[str, ...]:
+    """按导入状态表返回硬性资源要求；音频在两条支线中始终必需。"""
+    if YINYAN_IMPORT_TABLE == 'yinyan_song_records2':
+        return ('audio',)
+    return ('audio', 'lyric')
+
+
+def _require_primary_assets(assets: dict, required: tuple[str, ...] | None = None) -> None:
+    """校验当前导入支线要求的资源，允许非必需资源以空值入库。"""
     audio_url, _ = assets.get('audio', ('', ''))
     audio_url, _ = assets.get('audio', ('', ''))
-    missing = [
-        name for name, value in (
-            ('audio', audio_url), ('cover', assets.get('cover')), ('lyric', assets.get('lyric')),
-        ) if not value
-    ]
+    values = {'audio': audio_url, 'cover': assets.get('cover'), 'lyric': assets.get('lyric')}
+    required = required or ('audio', 'cover', 'lyric')
+    missing = [name for name in required if not values.get(name)]
     if missing:
     if missing:
         raise RequiredAssetTransferError(missing)
         raise RequiredAssetTransferError(missing)
 
 
@@ -206,14 +214,17 @@ def _select_qq_import_record(
     current_record: dict,
     current_record: dict,
     candidate_records: list[dict],
     candidate_records: list[dict],
     songs_by_mid: dict[str, dict],
     songs_by_mid: dict[str, dict],
+    required_assets: tuple[str, ...] = ('lyric', 'cover'),
 ) -> tuple[dict | None, str]:
 ) -> tuple[dict | None, str]:
     """为导入选择 QQ 录音；当前录音的歌词或封面不可用时，改选同源的合格候选。
     """为导入选择 QQ 录音；当前录音的歌词或封面不可用时，改选同源的合格候选。
     返回 (record, reason)，record 为 None 时 reason 说明失败原因。"""
     返回 (record, reason)，record 为 None 时 reason 说明失败原因。"""
     current_song = songs_by_mid.get(current_record['platform_unique_key'])
     current_song = songs_by_mid.get(current_record['platform_unique_key'])
     if not current_song:
     if not current_song:
         return None, 'spider data missing'
         return None, 'spider data missing'
-    needs_lyric = not _has_usable_qq_lyric(current_song, hk_row)
-    needs_cover = not _is_usable_qq_cover(_qq_cover_source(hk_row, current_song)) or not current_song.get('album_id')
+    needs_lyric = 'lyric' in required_assets and not _has_usable_qq_lyric(current_song, hk_row)
+    needs_cover = 'cover' in required_assets and (
+        not _is_usable_qq_cover(_qq_cover_source(hk_row, current_song)) or not current_song.get('album_id')
+    )
     if not needs_lyric and not needs_cover:
     if not needs_lyric and not needs_cover:
         return current_record, 'ok'
         return current_record, 'ok'
 
 
@@ -239,6 +250,7 @@ def _select_kugou_import_record(
     current_record: dict,
     current_record: dict,
     candidate_records: list[dict],
     candidate_records: list[dict],
     songs_by_id: dict[int, dict],
     songs_by_id: dict[int, dict],
+    required_assets: tuple[str, ...] = ('lyric', 'cover'),
 ) -> tuple[dict | None, str]:
 ) -> tuple[dict | None, str]:
     """为导入选择酷狗录音；当前录音的歌词或封面不可用时，改选同源的合格候选。
     """为导入选择酷狗录音；当前录音的歌词或封面不可用时，改选同源的合格候选。
     返回 (record, reason)，record 为 None 时 reason 说明失败原因。"""
     返回 (record, reason)，record 为 None 时 reason 说明失败原因。"""
@@ -246,8 +258,8 @@ def _select_kugou_import_record(
     current_song = songs_by_id.get(song_id)
     current_song = songs_by_id.get(song_id)
     if not current_song:
     if not current_song:
         return None, 'spider data missing'
         return None, 'spider data missing'
-    needs_lyric = not _has_kugou_lyric(current_song, hk_row)
-    needs_cover = not _has_kugou_cover(hk_row, current_song)
+    needs_lyric = 'lyric' in required_assets and not _has_kugou_lyric(current_song, hk_row)
+    needs_cover = 'cover' in required_assets and not _has_kugou_cover(hk_row, current_song)
     if not needs_lyric and not needs_cover:
     if not needs_lyric and not needs_cover:
         return current_record, 'ok'
         return current_record, 'ok'
 
 
@@ -274,6 +286,7 @@ def _select_netease_import_record(
     current_record: dict,
     current_record: dict,
     candidate_records: list[dict],
     candidate_records: list[dict],
     songs_by_id: dict[int, dict],
     songs_by_id: dict[int, dict],
+    required_assets: tuple[str, ...] = ('lyric', 'cover'),
 ) -> tuple[dict | None, str]:
 ) -> tuple[dict | None, str]:
     """为导入选择网易录音；当前录音的歌词或封面不可用时，改选同源的合格候选。
     """为导入选择网易录音；当前录音的歌词或封面不可用时，改选同源的合格候选。
     返回 (record, reason)，record 为 None 时 reason 说明失败原因。"""
     返回 (record, reason)，record 为 None 时 reason 说明失败原因。"""
@@ -281,8 +294,8 @@ def _select_netease_import_record(
     current_song = songs_by_id.get(song_id)
     current_song = songs_by_id.get(song_id)
     if not current_song:
     if not current_song:
         return None, 'spider data missing'
         return None, 'spider data missing'
-    needs_lyric = not _has_netease_lyric(current_song, hk_row)
-    needs_cover = not _has_netease_cover(hk_row, current_song)
+    needs_lyric = 'lyric' in required_assets and not _has_netease_lyric(current_song, hk_row)
+    needs_cover = 'cover' in required_assets and not _has_netease_cover(hk_row, current_song)
     if not needs_lyric and not needs_cover:
     if not needs_lyric and not needs_cover:
         return current_record, 'ok'
         return current_record, 'ok'
 
 
@@ -332,7 +345,7 @@ def _prepare_qq_payload(hk_row: dict, pr: dict, bucket, base_url, sp: dict, sing
             )
             )
         )
         )
     assets = _run_io_tasks(tasks)
     assets = _run_io_tasks(tasks)
-    _require_primary_assets(assets)
+    _require_primary_assets(assets, _required_import_assets())
     audio_url, audio_md5 = assets['audio']
     audio_url, audio_md5 = assets['audio']
     cover_url = assets['cover']
     cover_url = assets['cover']
     lyric_url = assets['lyric']
     lyric_url = assets['lyric']
@@ -461,7 +474,7 @@ def _prepare_kugou_payload(hk_row: dict, pr: dict, bucket, base_url, sp: dict, s
             )
             )
         )
         )
     assets = _run_io_tasks(tasks)
     assets = _run_io_tasks(tasks)
-    _require_primary_assets(assets)
+    _require_primary_assets(assets, _required_import_assets())
     audio_url, audio_md5 = assets['audio']
     audio_url, audio_md5 = assets['audio']
     cover_url = assets['cover']
     cover_url = assets['cover']
     lyric_url = assets['lyric']
     lyric_url = assets['lyric']
@@ -573,7 +586,7 @@ def _prepare_netease_payload(hk_row: dict, pr: dict, bucket, base_url, sp: dict,
             )
             )
         )
         )
     assets = _run_io_tasks(tasks)
     assets = _run_io_tasks(tasks)
-    _require_primary_assets(assets)
+    _require_primary_assets(assets, _required_import_assets())
     audio_url, audio_md5 = assets['audio']
     audio_url, audio_md5 = assets['audio']
     cover_url = assets['cover']
     cover_url = assets['cover']
     lyric_url = assets['lyric']
     lyric_url = assets['lyric']
@@ -1846,6 +1859,7 @@ def run(
 
 
             prepare_inputs = []
             prepare_inputs = []
             rejected_pending: list[dict] = []
             rejected_pending: list[dict] = []
+            required_assets = _required_import_assets()
             for pending in pending_records:
             for pending in pending_records:
                 src_id = int(pending['song_id'])
                 src_id = int(pending['song_id'])
                 platform = str(pending['platform'])
                 platform = str(pending['platform'])
@@ -1866,15 +1880,15 @@ def run(
                     continue
                     continue
                 if platform == PLATFORM_QQ:
                 if platform == PLATFORM_QQ:
                     selected, reason = _select_qq_import_record(
                     selected, reason = _select_qq_import_record(
-                        hk_row, pr, qq_records_by_song.get(src_id, []), qq_songs_map,
+                        hk_row, pr, qq_records_by_song.get(src_id, []), qq_songs_map, required_assets,
                     )
                     )
                 elif platform == PLATFORM_KUGOU:
                 elif platform == PLATFORM_KUGOU:
                     selected, reason = _select_kugou_import_record(
                     selected, reason = _select_kugou_import_record(
-                        hk_row, pr, kugou_records_by_song.get(src_id, []), kugou_songs_map,
+                        hk_row, pr, kugou_records_by_song.get(src_id, []), kugou_songs_map, required_assets,
                     )
                     )
                 elif platform == PLATFORM_NETEASE:
                 elif platform == PLATFORM_NETEASE:
                     selected, reason = _select_netease_import_record(
                     selected, reason = _select_netease_import_record(
-                        hk_row, pr, netease_records_by_song.get(src_id, []), netease_songs_map,
+                        hk_row, pr, netease_records_by_song.get(src_id, []), netease_songs_map, required_assets,
                     )
                     )
                 else:
                 else:
                     selected, reason = pr, 'ok'
                     selected, reason = pr, 'ok'

@@ -78,6 +78,33 @@ def test_required_asset_validation_rejects_partial_transfer():
         })
         })
 
 
 
 
+def test_records_import_requires_audio_and_lyric_but_allows_empty_cover(monkeypatch):
+    monkeypatch.setattr(runner, 'YINYAN_IMPORT_TABLE', 'yinyan_song_records')
+
+    runner._require_primary_assets({
+        'audio': ('https://archive.example/audio.mp3', 'md5'),
+        'cover': '',
+        'lyric': 'https://archive.example/lyric.txt',
+    }, runner._required_import_assets())
+
+    with pytest.raises(runner.RequiredAssetTransferError, match='lyric'):
+        runner._require_primary_assets({
+            'audio': ('https://archive.example/audio.mp3', 'md5'),
+            'cover': '',
+            'lyric': '',
+        }, runner._required_import_assets())
+
+
+def test_records2_import_allows_empty_cover_and_lyric(monkeypatch):
+    monkeypatch.setattr(runner, 'YINYAN_IMPORT_TABLE', 'yinyan_song_records2')
+
+    runner._require_primary_assets({
+        'audio': ('https://archive.example/audio.mp3', 'md5'),
+        'cover': '',
+        'lyric': '',
+    }, runner._required_import_assets())
+
+
 def test_lyric_upload_prefers_hk_url_and_forces_oss_transfer(monkeypatch):
 def test_lyric_upload_prefers_hk_url_and_forces_oss_transfer(monkeypatch):
     download = MagicMock(return_value='[00:01.00]HK lyric')
     download = MagicMock(return_value='[00:01.00]HK lyric')
     monkeypatch.setattr(runner, 'download_text_url', download)
     monkeypatch.setattr(runner, 'download_text_url', download)
@@ -184,6 +211,36 @@ def test_select_qq_import_record_returns_none_when_no_candidate_satisfies_missin
     assert 'lyric' in reason and 'cover' in reason
     assert 'lyric' in reason and 'cover' in reason
 
 
 
 
+def test_records_selection_does_not_reject_or_replace_for_missing_cover():
+    current = {'record_id': 10, 'platform_unique_key': 'current-mid'}
+
+    selected, reason = runner._select_qq_import_record(
+        {'cover_url': '', 'lyrics_url': 'https://example.com/hk-lyric.lrc'},
+        current,
+        [current],
+        {'current-mid': {'album_id': None, 'cover': '', 'lyric': ''}},
+        required_assets=('audio', 'lyric'),
+    )
+
+    assert selected == current
+    assert reason == 'ok'
+
+
+def test_records2_selection_accepts_current_record_with_empty_cover_and_lyric():
+    current = {'record_id': 10, 'platform_unique_key': 'current-mid'}
+
+    selected, reason = runner._select_qq_import_record(
+        {'cover_url': '', 'lyrics_url': ''},
+        current,
+        [current],
+        {'current-mid': {'album_id': None, 'cover': '', 'lyric': ''}},
+        required_assets=('audio',),
+    )
+
+    assert selected == current
+    assert reason == 'ok'
+
+
 def test_run_imports_only_pending_yinyan_platform_record(monkeypatch):
 def test_run_imports_only_pending_yinyan_platform_record(monkeypatch):
     pg_conn = _PgConnection()
     pg_conn = _PgConnection()
     prepare = MagicMock(return_value={
     prepare = MagicMock(return_value={
@@ -206,6 +263,7 @@ def test_run_imports_only_pending_yinyan_platform_record(monkeypatch):
     monkeypatch.setattr(runner, 'get_spider_conn', lambda: _Connection())
     monkeypatch.setattr(runner, 'get_spider_conn', lambda: _Connection())
     monkeypatch.setattr(runner, 'get_pg_conn', lambda: pg_conn)
     monkeypatch.setattr(runner, 'get_pg_conn', lambda: pg_conn)
     monkeypatch.setattr(runner, 'get_oss_bucket', lambda: object())
     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: [
     monkeypatch.setattr(runner, 'fetch_pending_yinyan_song_records', lambda cur, batch_size: [
         {'song_id': 10, 'record_id': 200, 'platform': '2'},
         {'song_id': 10, 'record_id': 200, 'platform': '2'},
     ] if pg_conn.commits == 0 else [])
     ] if pg_conn.commits == 0 else [])
@@ -262,6 +320,7 @@ def test_initialize_yinyan_song_records_inserts_primary_records(monkeypatch):
     monkeypatch.setattr(runner, 'get_hk_songs_conn', lambda: _Connection())
     monkeypatch.setattr(runner, 'get_hk_songs_conn', lambda: _Connection())
     monkeypatch.setattr(runner, 'get_source_conn', lambda: _Connection())
     monkeypatch.setattr(runner, 'get_source_conn', lambda: _Connection())
     monkeypatch.setattr(runner, 'get_pg_conn', lambda: pg_conn)
     monkeypatch.setattr(runner, 'get_pg_conn', lambda: pg_conn)
+    monkeypatch.setattr(runner, 'refresh_conn', lambda conn, name: conn)
     monkeypatch.setattr(runner, 'fetch_existing_yinyan_song_ids', lambda cur: set())
     monkeypatch.setattr(runner, 'fetch_existing_yinyan_song_ids', lambda cur: set())
 
 
     monkeypatch.setattr(runner, 'iter_hk_songs_batches', lambda conn, batch_size: [[
     monkeypatch.setattr(runner, 'iter_hk_songs_batches', lambda conn, batch_size: [[
@@ -335,6 +394,7 @@ def test_backfill_yinyan_record_platforms_updates_missing_platform_rows(monkeypa
 
 
     monkeypatch.setattr(runner, 'get_source_conn', lambda: _Connection())
     monkeypatch.setattr(runner, 'get_source_conn', lambda: _Connection())
     monkeypatch.setattr(runner, 'get_pg_conn', lambda: pg_conn)
     monkeypatch.setattr(runner, 'get_pg_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: [
     monkeypatch.setattr(runner, 'fetch_yinyan_records_missing_platform', lambda cur, batch_size: [
         {'song_id': 10, 'record_id': 100},
         {'song_id': 10, 'record_id': 100},
         {'song_id': 11, 'record_id': 101},
         {'song_id': 11, 'record_id': 101},