On 5/26/05, Oleg Broytmann <phd@...> wrote:
> On Thu, May 26, 2005 at 09:12:51AM -0400, Kevin Dangoor wrote:
> > I'm working on a patch to support pysqlite2, which exposes more of
> > sqlite's functionality and is designed to better handle multiple
> > threads.
> Nice to hear. I thought I need to look there... but now I just can
> wait! ;)
As much as I'd enjoy someone more knowledgeable about sqlobject doing
this, my timeline is likely tighter than yours :)
Seriously, though, I'm happy to help out.
> > if a database has good unicode
> > support, and the db interface returns unicode objects appropriately,
> > it's better to take advantage of that.
> Where? There are StringCol that are ascii strings, and UnicodeCol...
I think I jumped the gun on this message. I've just been looking more
closely at the failures and while there *is* a problem with PickleCol
(that can be addressed as I mentioned), the issues that the tests are
running into do not appear to be unicode related as I had thought.
Curiously, when using an on-disk database 82 tests pass, 38 tests
fail. Using :memory:, 108 pass, 12 fail. That kind of thing is always
a little disturbing.
There is at least one possible autocommit related issue (the
autocommit flag has gone away in pysqlite2, replaced by
isolation_level, which is more granular). And, then there's stuff like
def _executeRetry(self, conn, cursor, query):
E return cursor.execute(query)
> OperationalError: SQL logic error or missing database
Fun fun fun!
> > One other thing: why do the tests use a database file on disk rather
> > than :memory:?
> They don't. E.g. I declare
Aha. I just took the brute force approach and changed conftest.py...