SQLAlchemy（无ORM）：更新语句未提交

Question

I have this (anonymized) function in my database module:

def fix_publishing_dates(row_id, last_time=None, next_time=None,
    next_index=1, user="python_script"):
  sql = """
  UPDATE
    schema.table
  SET
    last_time = :last_time
  , next_time = :next_time
  , next_index = :next_index
  , col4 = SYSDATE
  , col5 = :user_id
  , is_active = 1
  WHERE
    id = :row_id
  """
  with closing(Session()) as s:
    with s.begin_nested():
      user_id = get_userid_by_name(user)
      args = dict(
          last_time=last_time,
          next_time=next_time,
          next_index=next_index,
          row_id=row_id,
          user_id=user_id,
          )
      s.execute(sql, args)
      s.flush()
    s.commit()

For some reason, this is not working. I query the above table for is_active=1, and I get zero rows. Am I doing something obviously wrong here?

Note

I don't want to use SQLAlchemy ORM and add lots of boilerplate Table classes* for this; I just like using Session() with text queries for the transaction support.

*: nor slow down my startup time with introspection; the network pipe to this database is slow.

Edit 1

• I'm using an Oracle 11 database through cx_oracle.
• In case it matters, one of the bound values is sometimes None/

• This code (anonymized differently) also isn't working:

   def fix_publishing_dates(**kwargs): sql = insert_query_here user_id = get_userid_by_name(user) args = dict(kwargs) print "*" * 50 print "* About to update database with values: {}".format(args) print "*" * 50 result = engine.execute(sql, args) print "Row count is:", result.rowcount #import ipdb;ipdb.set_trace() #s.commit() 

Answer 1

I don't think the Sessions will manage transactions without using ORM objects. But you can just use the transactions on the engine directly:

with engine.begin() as conn:
    conn.execute(...)