I'm just about to embark on the upgrade from 0.1x to 0.2x and given that there is a fairly significant amount of work to be done to make it happen, I'm trying to work out the best longer term strategy. For me, the most significant change is that queries, etc are now on a session not the mapper and I have a lot of code that will need to change (as background, no domain objects have SA specific knowledge, they call helper functions that are injected at start-up and these are SA-specific and know about mappers, tables, etc).
So the first question is what is the long term future of the threadlocal plugin - is it safe/prudent to base a major rewrite on this approach? (This looks to have the least impact on the existing code, replace any obj.mapper.select() with objectstore.get_session().query(Class).select(), etc). The other option of changing the interfaces to pass around Session instances means that all my domain objects have to have knowledge of SA sessions which is not good. It's been a long couple of days (fighting with FreeBSD upgrades), so I'm probably not covering all the bases, are there options I'm missing? Robert _______________________________________________ Sqlalchemy-users mailing list Sqlalchemy-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/sqlalchemy-users
