Hello,

1. Sometimes I have got cx_Oracle.DatabaseError: ORA-24422 I want to use FORCE parameter of SessionPool.close(force=False) for to make it true. https://cx-oracle.readthedocs.io/en/latest/api_manual/session_pool.html Could you please add this option.

2. cx_Oracle.SessionPool defaults is encoding=None, nencoding=None Could you please make it as original values.

Could you please add create_pool function with dsn parameter. And it looks better to use cx_Oracle.makedsn to build dsn string instead of your variant "f"{host}:{port}/{db}"". Example from documentation: dsn = cx_Oracle.makedsn("dbhost.example.com", 1521, service_name="orclpdb1") connection = cx_Oracle.connect("hr", userpwd, dsn, encoding="UTF-8")

Could you please support https://cx-oracle.readthedocs.io/en/latest/api_manual/connection.html Connection.ping() and Connection.rollback() methods. This is must have.

Connection.module Connection.action Connection.client_identifier Connection.clientinfo Would be very nice too.

Thank you.

Seems like the Close method of Oracle pool is missed for async. I'm not shure is it just

    async def close_connection_async(self) -> None:
        if self._connection:
            await self._connection.close()

or should be like

    async def close_connection_async(self) -> None:
        if self._connection:
            self._connection.close()
            await self._connection.wait_closed()

When I add cx-Oracle-async by Poetry and call "poetry show" command I got the next error: mydir>poetry show

SolverProblemError

Because myproject depends on cx-Oracle-async (^0.1.3) which doesn't match any versions, version solving failed.

at c:\users\myuser\appdata\local\programs\python\python39-32\lib\site-packages\poetry\puzzle\solver.py:241 in _solve 237│ packages = result.packages 238│ except OverrideNeeded as e: 239│ return self.solve_in_compatibility_mode(e.overrides, use_latest=use_latest) 240│ except SolveFailure as e: → 241│ raise SolverProblemError(e) 242│ 243│ results = dict( 244│ depth_first_search( 245│ PackageNode(self._package, packages), aggregate_package_nodes

Please add class AsyncCursorWarper: ... async def var(self, args): return await self._loop.run_in_executor(self._thread_pool , self._cursor.var, args)

And why "Warper", may be "Wrapper"?

Hi, while I was wrangling with cx_Oracle and my M1, I found out that Oracle released a new way to connect to their database systems: https://oracle.github.io/python-oracledb/

Maybe that will make it easier to write the async functionality due to better integration with Python and the THIN approach.

To prevent this error:

File /path/to/cx_Oracle_async/cursors.py:45, in AsyncCursorWrapper.callproc(self, *args, **kwargs)
     44 async def callproc(self, *args , **kwargs):
---> 45     return await self._loop.run_in_executor(self._thread_pool , self._cursor.callproc, *args , **kwargs)

TypeError: BaseEventLoop.run_in_executor() got an unexpected keyword argument 'keyword_parameters'

Other python database packages (including cx_Oracle) include a method to get the column names which involves using cursor.description. It looks like the object 'description' is not available for cx_Oracle_async for the AsyncCursorWrapper. Is there a different method for returning the column names when using this package?

Describe the bug When the number of async tasks awaiting acquire reaches ThreadPoolExecutorPlus max_workers, this program hangs.

To Reproduce Run the following with the sixth line of code commented out (as-is) to hang. Run the following with the sixth line of code uncommented to run successfully. (sixth line has the # toggle commenting... comment on it) (ThreadPoolExecutorPlus DEFAULT_MAXIMUM_WORKER_NUM is 32 * num-cores for Linux)

import cx_Oracle_async
import asyncio

dbspecs = dict(dsn="WFX742D", user="ars", password="ars")

max_pool_cnxs = 5
num_acquire_tasks = 32 * os.cpu_count() + max_pool_cnxs

#num_acquire_tasks -= 1 # toggle commenting of this line to hang/run
queue_len = 0

async def acquire_conn(pool, i):
   global queue_len
   queue_len += 1
   print_pool_info(f't-{i} acquire...', pool)
   conn = await pool.acquire()
   print_pool_info(f't-{i} postacquire', pool)
   queue_len -= 1
   conn._conn.begin()
   cursor = await conn.cursor()
   await cursor.execute('select current_timestamp from dual', [])
   rows = await cursor.fetchall()
   print(f"t-{i} rows: {len(rows)}")
   await conn.commit()
   await pool.release(conn)

async def test_many_threads():
   pool = await cx_Oracle_async.create_pool(**dbspecs, max=max_pool_cnxs)
   print_pool_info('init', pool)
   tasks = [];
   for i in range(0, num_acquire_tasks):
      tasks.append(asyncio.create_task(acquire_conn(pool, i), name=f'task-{i}'))
   await asyncio.wait(tasks)
   print_pool_info('done', pool)

def print_pool_info(action, pool):
   ipool = pool._pool
   print(f'{action} (queued {queue_len}, busy {ipool.busy}/{ipool.max})')

asyncio.run(test_many_threads())

Expected behavior I expected the queries to be queued and run when connections became available.

Platform:

• OS: Red Hat Enterprise Linux Server release 7.9 (Maipo)
• Oracle version: 11.2.0.4
• Python version 3.8.11

Additional context I worked around this by added a semaphore in my code to limit the number of acquires to the max cnxs. What is the benefit of allowing more threads than cnxs?

Describe the bug Potential memory leak in cursors context manager.

To Reproduce Call many times cursor.fetchone().

Gist with sample code: https://gist.github.com/ilosamart/272a7dadb639f19fd62b7947ae12ab5f It contains three versions of the same method: one async (this lib), one async (this lib) which explicitly closes the cursor, one with pure cx_oracle.

Expected behavior I expected that the cursor context manager closed the cursor.

Platform:

• OS: Ubuntu 20.04
• Oracle version: 12.2
• Python version 3.8

Additional context I used the instructions on https://oracle.github.io/odpi/doc/user_guide/debugging.html to debug memory alloc inside OCI.