fix(database): harden fresh migration chain compatibility (#6278)

database/versions/93f8cb6a4d1e_2_2_4.py

@@ -145,6 +145,13 @@ DOWNGRADE_RESTORE_INDEXES = {
 
 def _load_schema_state(inspector: sa.Inspector):
     tables = set(inspector.get_table_names())
+    table_columns = {
+        table_name: {
+            column["name"]
+            for column in inspector.get_columns(table_name)
+        }
+        for table_name in tables
+    }
     table_indexes = {
         table_name: {
             index["name"]: {
@@ -155,7 +162,7 @@ def _load_schema_state(inspector: sa.Inspector):
         }
         for table_name in tables
     }
-    return tables, table_indexes
+    return tables, table_columns, table_indexes
 
 
 def _drop_index(
@@ -215,10 +222,13 @@ def _create_index(
     index_name: str,
     columns: list[str],
     tables: set[str],
+    table_columns: dict[str, set[str]],
     table_indexes: dict[str, dict[str, dict[str, object]]],
 ) -> None:
     if table_name not in tables:
         return
+    if not set(columns).issubset(table_columns[table_name]):
+        return
     if index_name in table_indexes[table_name]:
         return
     if _has_index_signature(table_name, columns, tables, table_indexes, unique=False):
@@ -231,8 +241,9 @@ def _create_index(
 
 
 def upgrade() -> None:
+    """以字段签名幂等替换 2.2.4 高频查询索引。"""
     inspector = sa.inspect(op.get_bind())
-    tables, table_indexes = _load_schema_state(inspector)
+    tables, table_columns, table_indexes = _load_schema_state(inspector)
 
     for table_name, index_specs in REDUNDANT_ID_INDEXES.items():
         for index_name, columns in index_specs:
@@ -258,12 +269,20 @@ def upgrade() -> None:
 
     for table_name, index_specs in CREATE_INDEXES.items():
         for index_name, columns in index_specs:
-            _create_index(table_name, index_name, columns, tables, table_indexes)
+            _create_index(
+                table_name,
+                index_name,
+                columns,
+                tables,
+                table_columns,
+                table_indexes,
+            )
 
 
 def downgrade() -> None:
+    """移除组合索引并恢复适用于当前表结构的旧索引。"""
     inspector = sa.inspect(op.get_bind())
-    tables, table_indexes = _load_schema_state(inspector)
+    tables, table_columns, table_indexes = _load_schema_state(inspector)
 
     for table_name, index_specs in CREATE_INDEXES.items():
         for index_name, _ in index_specs:
@@ -271,8 +290,22 @@ def downgrade() -> None:
 
     for table_name, index_specs in DOWNGRADE_RESTORE_INDEXES.items():
         for index_name, columns in index_specs:
-            _create_index(table_name, index_name, columns, tables, table_indexes)
+            _create_index(
+                table_name,
+                index_name,
+                columns,
+                tables,
+                table_columns,
+                table_indexes,
+            )
 
     for table_name, index_specs in REDUNDANT_ID_INDEXES.items():
         for index_name, columns in index_specs:
-            _create_index(table_name, index_name, columns, tables, table_indexes)
+            _create_index(
+                table_name,
+                index_name,
+                columns,
+                tables,
+                table_columns,
+                table_indexes,
+            )

database/versions/e6a1c4b8d2f0_2_2_13.py

@@ -36,55 +36,87 @@ def _has_column(
     )
 
 
+def _has_index(
+    inspector: sa.Inspector,
+    table_name: str,
+    index_name: str,
+) -> bool:
+    """检查数据表是否已存在指定索引。"""
+    if table_name not in inspector.get_table_names():
+        return False
+    return any(
+        index["name"] == index_name
+        for index in inspector.get_indexes(table_name)
+    )
+
+
+def _column_names(inspector: sa.Inspector, table_name: str) -> set[str]:
+    """读取数据表当前全部字段名。"""
+    if table_name not in inspector.get_table_names():
+        return set()
+    return {
+        column["name"]
+        for column in inspector.get_columns(table_name)
+    }
+
+
+def _ensure_column_and_index(column_name: str) -> None:
+    """独立补齐整理历史的规范身份字段及其索引。"""
+    table_name = "transferhistory"
+    index_name = f"ix_{table_name}_{column_name}"
+    inspector = sa.inspect(op.get_bind())
+    if not _has_column(inspector, table_name, column_name):
+        op.add_column(
+            table_name,
+            sa.Column(column_name, sa.String(), nullable=True),
+        )
+
+    inspector = sa.inspect(op.get_bind())
+    if not _has_index(inspector, table_name, index_name):
+        op.create_index(index_name, table_name, [column_name])
+
+
 def upgrade() -> None:
     """升级整理历史数据源字段。"""
-    inspector = sa.inspect(op.get_bind())
-    if not _has_column(inspector, "transferhistory", "media_source"):
-        op.add_column(
-            "transferhistory",
-            sa.Column("media_source", sa.String(), nullable=True),
-        )
-        op.create_index(
-            "ix_transferhistory_media_source",
-            "transferhistory",
-            ["media_source"],
-        )
+    _ensure_column_and_index("media_source")
+    _ensure_column_and_index("media_id")
 
-    inspector = sa.inspect(op.get_bind())
-    if not _has_column(inspector, "transferhistory", "media_id"):
-        op.add_column(
-            "transferhistory",
-            sa.Column("media_id", sa.String(), nullable=True),
-        )
-        op.create_index(
-            "ix_transferhistory_media_id",
-            "transferhistory",
-            ["media_id"],
-        )
-
-    transfer_history = sa.table(
-        "transferhistory",
-        sa.column("tmdbid", sa.Integer()),
-        sa.column("doubanid", sa.String()),
+    columns = _column_names(sa.inspect(op.get_bind()), "transferhistory")
+    table_columns = [
         sa.column("media_source", sa.String()),
         sa.column("media_id", sa.String()),
-    )
+    ]
+    if "tmdbid" in columns:
+        table_columns.append(sa.column("tmdbid", sa.Integer()))
+    if "doubanid" in columns:
+        table_columns.append(sa.column("doubanid", sa.String()))
+    transfer_history = sa.table("transferhistory", *table_columns)
     connection = op.get_bind()
-    connection.execute(
-        transfer_history.update()
-        .where(transfer_history.c.tmdbid.is_not(None))
-        .where(transfer_history.c.media_id.is_(None))
-        .values(
-            media_source="themoviedb",
-            media_id=sa.cast(transfer_history.c.tmdbid, sa.String()),
+
+    if "tmdbid" in columns:
+        connection.execute(
+            transfer_history.update()
+            .where(transfer_history.c.tmdbid.is_not(None))
+            .where(transfer_history.c.media_id.is_(None))
+            .values(
+                media_source="themoviedb",
+                media_id=sa.cast(transfer_history.c.tmdbid, sa.String()),
+            )
         )
-    )
-    connection.execute(
+
+    if "doubanid" not in columns:
+        return
+    douban_update = (
         transfer_history.update()
-        .where(transfer_history.c.tmdbid.is_(None))
         .where(transfer_history.c.doubanid.is_not(None))
         .where(transfer_history.c.media_id.is_(None))
-        .values(
+    )
+    if "tmdbid" in columns:
+        douban_update = douban_update.where(
+            transfer_history.c.tmdbid.is_(None)
+        )
+    connection.execute(
+        douban_update.values(
             media_source="douban",
             media_id=transfer_history.c.doubanid,
         )
@@ -93,15 +125,12 @@ def upgrade() -> None:
 
 def downgrade() -> None:
     """回滚整理历史数据源字段。"""
-    inspector = sa.inspect(op.get_bind())
-    if _has_column(inspector, "transferhistory", "media_id"):
-        op.drop_index("ix_transferhistory_media_id", table_name="transferhistory")
-        op.drop_column("transferhistory", "media_id")
+    for column_name in ("media_id", "media_source"):
+        index_name = f"ix_transferhistory_{column_name}"
+        inspector = sa.inspect(op.get_bind())
+        if _has_index(inspector, "transferhistory", index_name):
+            op.drop_index(index_name, table_name="transferhistory")
 
-    inspector = sa.inspect(op.get_bind())
-    if _has_column(inspector, "transferhistory", "media_source"):
-        op.drop_index(
-            "ix_transferhistory_media_source",
-            table_name="transferhistory",
-        )
-        op.drop_column("transferhistory", "media_source")
+        inspector = sa.inspect(op.get_bind())
+        if _has_column(inspector, "transferhistory", column_name):
+            op.drop_column("transferhistory", column_name)

database/versions/f7b2d5c9a301_2_2_14.py

@@ -46,6 +46,17 @@ def _has_index(
     )
 
 
+def _column_names(table_name: str) -> set[str]:
+    """读取数据表当前全部字段名。"""
+    inspector = sa.inspect(op.get_bind())
+    if table_name not in inspector.get_table_names():
+        return set()
+    return {
+        column["name"]
+        for column in inspector.get_columns(table_name)
+    }
+
+
 def _add_columns(table_name: str, columns: Iterable[sa.Column]) -> None:
     """为指定表补充尚不存在的字段。"""
     for column in columns:
@@ -61,22 +72,29 @@ def _create_index(table_name: str, index_name: str, columns: list[str]) -> None:
         op.create_index(index_name, table_name, columns)
 
 
-def _backfill_media_identity(table_name: str, has_mediaid: bool = False) -> None:
-    """使用兼容 ID 幂等回填统一媒体身份。"""
+def _backfill_media_identity(table_name: str) -> None:
+    """使用表中实际存在的兼容 ID 幂等回填统一媒体身份。"""
+    existing_columns = _column_names(table_name)
     columns = [
-        sa.column("tmdbid", sa.Integer()),
-        sa.column("doubanid", sa.String()),
-        sa.column("bangumiid", sa.Integer()),
-        sa.column("anilistid", sa.Integer()),
         sa.column("media_source", sa.String()),
         sa.column("media_id", sa.String()),
     ]
-    if has_mediaid:
-        columns.append(sa.column("mediaid", sa.String()))
+    identity_columns = {
+        "mediaid": sa.String,
+        "tmdbid": sa.Integer,
+        "doubanid": sa.String,
+        "bangumiid": sa.Integer,
+        "anilistid": sa.Integer,
+    }
+    columns.extend(
+        sa.column(column_name, column_type())
+        for column_name, column_type in identity_columns.items()
+        if column_name in existing_columns
+    )
     table = sa.table(table_name, *columns)
     connection = op.get_bind()
 
-    if has_mediaid:
+    if "mediaid" in existing_columns:
         for prefix, source in (
                 ("tmdb", "themoviedb"),
                 ("themoviedb", "themoviedb"),
@@ -100,6 +118,8 @@ def _backfill_media_identity(table_name: str, has_mediaid: bool = False) -> None
             ("bangumi", "bangumiid"),
             ("anilist", "anilistid"),
     ):
+        if field not in existing_columns:
+            continue
         identity_column = table.c[field]
         connection.execute(
             table.update()
@@ -162,8 +182,8 @@ def upgrade() -> None:
         ["type", "media_source", "media_id", "site"],
     )
 
-    _backfill_media_identity("subscribe", has_mediaid=True)
-    _backfill_media_identity("subscribehistory", has_mediaid=True)
+    _backfill_media_identity("subscribe")
+    _backfill_media_identity("subscribehistory")
     _backfill_media_identity("downloadhistory")
     _backfill_media_identity("transferhistory")
     _backfill_media_identity("downloadfailure")