Postgres 数据库中的 SQLAlchemy 批量插入语句抛出 AttributeError

Question

我正在尝试使用insert语句将 Python SQLAlchemy 中的行批量插入 Postgres 数据库。 我需要使用 insert 语句而不是bulk_insert_mappings ，因为我想默默地忽略重复条目的失败插入。 这在以前并不明显，但我现在添加了它。

该表已按应有的方式创建。 但是，即使是通过语句 API 进行的非常简单的插入操作也会引发此错误：

AttributeError: '_NoResultMetaData' object has no attribute '_indexes_for_keys'

最小可验证示例：

import os

import sqlalchemy
from sqlalchemy import (
    Column,
    INTEGER,
    TEXT
)
from sqlalchemy.dialects.postgresql import insert
from sqlalchemy.ext.declarative import declarative_base
from sqlalchemy.orm import sessionmaker

Base = declarative_base()


class Test(Base):
    __tablename__ = 'test'
    id = Column(INTEGER, primary_key=True)
    data = Column(TEXT)


engine = sqlalchemy.create_engine(os.environ['DATABASE_CONNECTION'])
Session = sessionmaker(engine)

Base.metadata.create_all(engine, Base.metadata.tables.values(), checkfirst=True)

connection = engine.connect()
buffer = [
    {
        'data': "First test"
    },
    {
        'data': "Second test"
    }
]

insert_statement = insert(Test).values(buffer)
# Using insert statement instead of bulk_insert_mappings so I can do nothing when adding duplicate entries
insert_or_do_nothing = insert_statement.on_conflict_do_nothing(index_elements=[Company.local_id])
orm_statement = sqlalchemy.select(Test).from_statement(insert_or_do_nothing)

with Session() as session:
    session.execute(orm_statement).scalars()

connection.close()

完整的堆栈跟踪：

Traceback (most recent call last):
  File "/project/path/test.py", line 41, in <module>
    session.execute(orm_statement).scalars()
  File "/venv/path/sqlalchemy/orm/session.py", line 1715, in execute
    result = compile_state_cls.orm_setup_cursor_result(
  File "/venv/path/sqlalchemy/orm/context.py", line 354, in orm_setup_cursor_result
    return loading.instances(result, querycontext)
  File "/venv/path/sqlalchemy/orm/loading.py", line 89, in instances
    cursor.close()
  File "/venv/path/sqlalchemy/util/langhelpers.py", line 70, in __exit__
    compat.raise_(
  File "/venv/path/sqlalchemy/util/compat.py", line 208, in raise_
    raise exception
  File "/venv/path/sqlalchemy/orm/loading.py", line 69, in instances
    *[
  File "/venv/path/sqlalchemy/orm/loading.py", line 70, in <listcomp>
    query_entity.row_processor(context, cursor)
  File "/venv/path/sqlalchemy/orm/context.py", line 2627, in row_processor
    _instance = loading._instance_processor(
  File "/venv/path/sqlalchemy/orm/loading.py", line 715, in _instance_processor
    primary_key_getter = result._tuple_getter(pk_cols)
  File "/venv/path/sqlalchemy/engine/result.py", line 934, in _tuple_getter
    return self._metadata._row_as_tuple_getter(keys)
  File "/venv/path/sqlalchemy/engine/result.py", line 106, in _row_as_tuple_getter
    indexes = self._indexes_for_keys(keys)
AttributeError: '_NoResultMetaData' object has no attribute '_indexes_for_keys'

我是否滥用了语句界面？ ORM 语句看起来不错：

INSERT INTO test (data) VALUES (:data_m0), (:data_m1)

我在用

• PostgreSQL 14.4
• psycopg2-二进制 2.9.3
• SQLAlchemy 1.4.39

Answer 1

查看文档，您可以尝试使用session.bulk_insert_mappings() 。

buffer = [
    {
        'data': "First test"
    },
    {
        'data': "Second test"
    }
]


with Session() as session:
    session.bulk_insert_mappings(Test, buffer)

Answer 2

我找到了一个使用插入语句的解决方案：避免使用 ORM 语句。 出于某种原因，使用普通语句似乎可以完成这项工作，而 ORM 语句则抛出AttributeError 。

这令人困惑，因为官方文档要求 ORM 语句：

# THIS APPROACH DID NOT WORK FOR ME

stmt = stmt.on_conflict_do_update(
    index_elements=[User.name], set_=dict(fullname=stmt.excluded.fullname)
).returning(User)

orm_stmt = (
    select(User)
    .from_statement(stmt)
    .execution_options(populate_existing=True)
)
for user in session.execute(
    orm_stmt,
).scalars():
    print("inserted or updated: %s" % user)

但是如果你省略 ORM 语句部分，一切都很好

# THIS WORKS

insert_statement = insert(Test).values(buffer)
insert_or_do_nothing = insert_statement.on_conflict_do_nothing(index_elements=[Test.id])

    with Session() as session:
        session.execute(insert_or_do_nothing)
        session.commit()