Cannot import name 'async_sessionmaker' from 'sqlalchemy.ext.asyncio' in 1.0.0a1

[lmocsi]

Checks

• I have checked that this issue has not already been reported.
• I have confirmed this bug exists on the latest version of Polars.

Reproducible example

create table my_test nologging pctfree 0 as
select
   'aaa' prod,
   date'2000-01-01' start_date,
   date'9999-12-31' end_date
from dual

import cx_Oracle
import polars as pl

host1 = '***'
host2 = '***'
service = '***'
username = '***'
password = '***'
dsn = f"(DESCRIPTION=(ADDRESS_LIST=(FAILOVER=yes)(ADDRESS=(PROTOCOL=TCP)(HOST={host1})(PORT=1521))(ADDRESS=(PROTOCOL=TCP)(HOST={host2})(PORT=1521)))(CONNECT_DATA=(SERVICE_NAME={service})))"

v_sql = "select * from my_test"

with cx_Oracle.connect(f"{username}/{password}@{dsn}") as con:
    df2 = pl.read_database(
        query=v_sql,
        connection=con
    )

Log output

ImportError                               Traceback (most recent call last)
/tmp/1000780000/ipykernel_1121/1054175297.py in <module>
     13 
     14 with cx_Oracle.connect(f"{username}/{password}@{dsn}") as con:
---> 15     df2 = pl.read_database(
     16         query=v_sql,
     17         connection=con

/opt/conda/envs/Python-3.9-Premium/lib/python3.9/site-packages/polars/io/database/functions.py in read_database(query, connection, iter_batches, batch_size, schema_overrides, infer_schema_length, execute_options)
    251     # return frame from arbitrary connections using the executor abstraction
    252     with ConnectionExecutor(connection) as cx:
--> 253         return cx.execute(
    254             query=query,
    255             options=execute_options,

/opt/conda/envs/Python-3.9-Premium/lib/python3.9/site-packages/polars/io/database/_executor.py in __exit__(self, exc_type, exc_val, exc_tb)
    106         # if we created it and are finished with it, we can
    107         # close the cursor (but NOT the connection)
--> 108         if self._is_alchemy_async(self.cursor):
    109             from sqlalchemy.ext.asyncio import AsyncConnection
    110 

/opt/conda/envs/Python-3.9-Premium/lib/python3.9/site-packages/polars/io/database/_executor.py in _is_alchemy_async(conn)
    319     def _is_alchemy_async(conn: Any) -> bool:
    320         """Check if the cursor/connection/session object is async."""
--> 321         from sqlalchemy.ext.asyncio import (
    322             AsyncConnection,
    323             AsyncSession,

ImportError: cannot import name 'async_sessionmaker' from 'sqlalchemy.ext.asyncio' (/opt/conda/envs/Python-3.9-Premium/lib/python3.9/site-packages/sqlalchemy/ext/asyncio/__init__.py)

Issue description

The above code works fine in polars-0.20.31, fails in polars-1.0.0a1.

Expected behavior

Read the table without errors.

Installed versions

--------Version info---------
Polars:               1.0.0-alpha.1
Index type:           UInt32
Platform:             Linux-4.18.0-372.76.1.el8_6.x86_64-x86_64-with-glibc2.28
Python:               3.9.13 (main, Oct 13 2022, 21:15:33) 
[GCC 11.2.0]

----Optional dependencies----
adbc_driver_manager:  <not installed>
cloudpickle:          2.0.0
connectorx:           0.3.4a1
deltalake:            <not installed>
fastexcel:            <not installed>
fsspec:               2022.02.0
gevent:               <not installed>
hvplot:               <not installed>
matplotlib:           3.8.4
nest_asyncio:         1.5.5
numpy:                1.23.5
openpyxl:             3.0.9
pandas:               2.2.2
pyarrow:              15.0.2
pydantic:             <not installed>
pyiceberg:            <not installed>
sqlalchemy:           1.4.27
torch:                1.10.2
xlsx2csv:             <not installed>
xlsxwriter:           3.2.0

The results are the same with:
connectorx==0.3.3

[alexander-beedie]

Thanks for the report; you seem to have a very old version of sqlalchemy installed (1.4.27 is a little over 2½ years out of date). If you can upgrade it to the latest in the 1.4.x (which would be 1.4.52) the error will resolve. I will harden the async check so it doesn't cause this error, but I wouldn't recommend being on quite such an old version if it can be avoided 👌

[lmocsi]

How about the newest sqlalchemy-2.0.30? It is good, as well?

[hendrix-bates]

Hi @alexander-beedie
As of 1.0.0 this is still a problem as the hardening was not applied to _is_alchemy_session().
async_sessionmaker is also unforunately not available as late as 1.4.49 of sqlalchemy which is the latest 1.4.x version on conda-forge