On Jul 26, 2006, at 3:29 PM, Gary Robinson wrote:
[edited ... question concerning stability ...]
) have been using SQLObject 0.7 bugfix branch (and predecessors) for about a year. Once we sorted out the caching issues and fixed a thread deadlock bug we've had smooth sailing. We do currently do about O( 1 Million ) DB operations / day and growing, using Postgres 8. Probably about 5% of our DB operations involved queries complex enough that we found it easier to go around SQLObject and write directly to psycopg2.
Not surprisingly, once we got to understand SQLObject's behavior well, we became more comfortable in using it efficiently. In general I have no love for ORM's, and IMHO SQLObject can do plenty of magic that can easily lead to very inefficient database use. In development mode we use "debug=True" to print out all of the queries as they happen, and we try to be ruthless at minimizing them. SQLObject makes it very easy to write data manipulation code in Python -- code that might be 10x or more efficient if done in SQL in the database. I don't believe that's the "fault" of SQLObject -- just the ORM pattern.
Anyway -- the short summary is this: it's good stuff.