On Fri, 28 Jun 2002, Jeremy Hylton wrote:
Your suggestion seems to be that we should treat references from older generations to newer generations as external roots. So a cycle that spans generations will not get collected until everything is in the same generation. Indeed, that does not seem harmful.
Not really -- I'm saying that certain types of containers tend to hold references to other, much larger, containers. These small containers tend to be ephemoral -- they appear and disappear quickly -- but sometimes are unlucky enough to be around when a collection is triggered. In my example, the small containers were bound-method objects, which store back-references to their class instance, a huge list, which will live in generation 2 very quickly.
I do not advocate making objects store which generation they belong to, but rather to delay the traversal of certain containers until after generation 0. This means that they've been around the block a few times, and may have fallen in with a bad cyclical crowd.
This annotation should be added to objects that tend to shadow other containers, like bound-methods, iterators, generators, descriptors, etc.
In some tests using real workloads, I've found that upwards of 99% of these ephemoral objects never make it to a generation 1 collection anyway.
Haulin' garbage, -Kevin
-- Kevin Jacobs The OPAL Group - Enterprise Systems Architect Voice: (216) 986-0710 x 19 E-mail: email@example.com Fax: (216) 986-0714 WWW: http://www.theopalgroup.com