On 05/06/2019 17:15, Mike Bayer wrote:

How come close() doesn't rollback the SessionTransaction if it throws it
away?

that's currently what .close() does, it discards the connection.   this is safe because the connection pool ensures transactions are rolled back.   This might have to change with some of the 2.0 things I'm thinking about but it's not clear yet.

Anyway, you don't get a rollback here because the session is bound to an external connection so the connection pool is not involved.

Ah, right, so under normal conditions, if I just close a session, any transaction and subtransactions in flight will be rolled back when the connection is returned to the pool?


if you want to roll back the work that the application did, that would be your sub_transaction.rollback():

     sub_transaction = conn.begin_nested()
     try:

         session = Session()

         # code under test:
         event = Event(text='some stuff got done')
         session.add(event)
         session.flush()
         session.close()

     finally:
         sub_transaction.rollback()
         assert session.query(Event).count() == 0

Sure, but the test then fails when the code is correct:

try:
    Base.metadata.create_all(bind=conn, checkfirst=False)
    Session.configure(bind=conn)

    sub_transaction = conn.begin_nested()
    try:

        session = Session()

        # code under test:
        event = Event(text='some stuff got done')
        session.add(event)
        session.flush()
        session.commit()
        session.close()

        # test:
        sub_transaction.rollback()

        assert session.query(Event).count() == 1

    finally:
        sub_transaction.rollback()

finally:
    transaction.rollback()

The panacea I'm after is to be able to run the DDL in a transaction, run each test in a subtransaction off that which is rolled back at the end of each test, but also be able to check that the code under test is doing session.commit() where it should. Where the pattern we're discussing, I certainly have the first two, but as you can see from what I've just pasted above, the third one isn't there, but is it possible?

the test harness is giving you two choices.  you can look at the state of the DB after your program has done some things and *before* your harness has reversed its work, or you can look at the state of the DB *after* your harness has reversed its work.

Unless I'm missing something, neither of these let the test confirm that the code under test has called commit() when it should.

cheers,

Chris

--
SQLAlchemy - The Python SQL Toolkit and Object Relational Mapper

http://www.sqlalchemy.org/

To post example code, please provide an MCVE: Minimal, Complete, and Verifiable 
Example.  See  http://stackoverflow.com/help/mcve for a full description.
--- You received this message because you are subscribed to the Google Groups "sqlalchemy" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to sqlalchemy+unsubscr...@googlegroups.com.
To post to this group, send email to sqlalchemy@googlegroups.com.
Visit this group at https://groups.google.com/group/sqlalchemy.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sqlalchemy/d7f7c9ac-2d32-6e60-efac-f86b501683f7%40withers.org.
For more options, visit https://groups.google.com/d/optout.

Reply via email to