[Chris Withers, about getting at the collection of modified objects]
> How hard would it be to expose?
The coding would be trivial in ZODB 3.6 (earliest version a new feature
could land) if limited solely to Connection objects. More work to specify
the interface and write tests.
> Would there be any performance implications?
None from just from exposing the current Connection._registered_objects
implementation detail via an official API.
Whether that would be suitable for all intended purposes I don't know -- I
haven't seen detailed use cases.
Doubt it's a good idea regardless. As Jim wrote, "things like this"
_should_ be approached not by adding an endless number of new APIs, but by
introducing an event system. I don't know what the performance implications
of that may be, since ZODB has nothing like it at present.
For more information about ZODB, see the ZODB Wiki:
ZODB-Dev mailing list - ZODB-Dev@zope.org