[sqlalchemy] Re: SQLAlchemy FlushError

2007-10-09 Thread jepr
I tested my hypothesis, and inactivated the trigger on the Project table, then added a Sequence('project_id_seq') to the project_id column in the project table definition, and the updates work like a charm. So, is it not permissible to have triggers set on tables if one is using SQLAlchemy?

[sqlalchemy] Re: SQLAlchemy FlushError

2007-10-09 Thread Michael Bayer
are you using Postgres ? if so, youd have to put a primary-key generating expression into your PassiveDefault. Psycopg2 gives us no good way to get back the newly generated ID from your trigger (only giving back OIDs, which according to PG docs are off by default and are deprecated:

[sqlalchemy] Re: SQLAlchemy FlushError

2007-10-09 Thread jepr
Hi Michael, No, I'm actually using Oracle. But, perhaps I should adopt the same methodology for it as well? Thanks! Joann On Oct 9, 1:52 pm, Michael Bayer [EMAIL PROTECTED] wrote: are you using Postgres ? if so, youd have to put a primary-key generating expression into your

[sqlalchemy] Re: SQLAlchemy FlushError

2007-10-09 Thread Michael Bayer
On Oct 9, 2007, at 2:01 PM, jepr wrote: Hi Michael, No, I'm actually using Oracle. But, perhaps I should adopt the same methodology for it as well? cx_oracle doesnt even have a lastrowid attribute on its cursor, so yes the same idea applies. note this limitation only applies to the