So, if I understand you correctly this is more some kind of bug in
sqlalchemy than a problem with my code, am I right?
Regards,
Christopher.
On Jul 26, 8:11 pm, Michael Bayer <mike...@zzzcomputing.com> wrote:
> unit of work logging can be enabled:
>
> import logging
> logging.basicConfig()
> logging.getLogger('sqlalchemy.orm.unitofwork').setLevel(logging.DEBUG)
>
> when a class has a relation to itself, a dependency is created during
> flush:
>
> DEBUG:sqlalchemy.orm.unitofwork.UOWTransaction:Dependent tuples:
> (Forum->Forum)
>
> it means, we can no longer insert all Forum rows in the order they
> were received - we have to check each object for dependencies on a
> parent Forum object and modify the insert order in that way.
>
> A sort is then performed among all three Forum objects, where "sort"
> is a topological dependency sort. The ordering of this sort is not
> deterministic. in this case, there are no dependencies at all so the
> sort turns up an ordering that is mostly random. but topological
> doesn't take "insert order" into account at this level.
>
> the resulting list is then grouped into a tree-like structure, which
> is intended to create "batches" of objects, where within each batch,
> the uow can again fall back upon the insert ordering among those
> rows. the "tree" step is total craziness which I wrote at some point
> and was rewritten by ants aasma, our resident uber genius, and would
> be great to get rid of.
>
> Here's what the code has to say:
>
> # rationale for "tree" sort as opposed to a straight
> # dependency - keep non-dependent objects
> # grouped together, so that insert ordering as determined
> # by session.add() is maintained.
> # An alternative might be to represent the "insert order"
> # as part of the topological sort itself, which would
> # eliminate the need for this step (but may make the original
> # topological sort more expensive)
>
> so in theory topological could be improved to take "insert order" into
> account, we can lose the unnecessary tree stage which doesn't really
> solve the issue completely anyway, and that would probably improve
> this test.
--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups
"sqlalchemy" group.
To post to this group, send email to sqlalchemy@googlegroups.com
To unsubscribe from this group, send email to
sqlalchemy+unsubscr...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/sqlalchemy?hl=en
-~----------~----~----~----~------~----~------~--~---
