[ZODB-Dev] Re: [Zope-Checkins] CVS: StandaloneZODB/ZODB - cPersistence.c:1.58 cPersistence.h:1.25 cPickleCache.c:1.55

[Toby Dickenson]

On Thu, 4 Apr 2002 20:12:48 -0500, Jeremy Hylton <jeremy@zope.com>
wrote:

On the new cache implementation....

>+/* XXX Does it make sense for full_sweep() and reallyfull_sweep() to
>+   empty the cache completely?  I agree that it would if dt is 0, but
>+   don't think it should for other times.  Perhaps it should just call
>+   incrgc() if dt > 2; the new cache may be efficient enough that
>+   incrgc() would suffice.
>+*/

I think the problem here is that this part of the cache API was
defined in terms of the limitations of the original implemetation.

Checking through the Zope source I only see either of these functions
used once, in the Catalog indexing loop, where a call to incrgc()
should now work equally well. Ill change that.

The 'reduce to zero' behavior is handy for debugging memory and
refcount problems. but I struggle to think of aplausible use for it
outside of debugging.

Is anyone else making use of them?

Is a tidy up justified? what about proving a new method
reduce_to_zero() and eventually remove fullsweep() and
reallyfullsweep()?





Toby Dickenson
tdickenson@geminidataloggers.com