[
https://issues.apache.org/jira/browse/DERBY-2114?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Dyre Tjeldvoll reopened DERBY-2114:
-----------------------------------
Reopening this issue since I now have an idea about how to solve the VTI
problem.
The real problem was not really related to VTIs, but to the implementation of
StatementCache. Its constructor looks up the LanguageConnectionContext's
statementCache which is of type CacheManager, and implemented by Clock. Since
the CacheManager interface doesn't support iteration over the elements of the
cache, a hack has been introduced here. The knowledge that a CacheManager
really is a Clock AND that Clock extends Hashtable is used to cast
lcc.statementCache to Hashtable. This Hashtable is then used to do the
iteration. However, this hack defeats the purpose of having the CacheManager
interface since it creates a solid unwanted dependency between the
implementation of StatementCache and the implementation of Clock.
If iteration is to be allowed it is better to extend the CacheManager interface
with the necessary methods (They can't simply be added to Clock, since
StatementCache isn't allowed to downcast to Clock which isn't public in its
package.)
> Let Clock embed a HashMap rather than inherit from Hashtable
> ------------------------------------------------------------
>
> Key: DERBY-2114
> URL: https://issues.apache.org/jira/browse/DERBY-2114
> Project: Derby
> Issue Type: Improvement
> Components: Performance
> Affects Versions: 10.2.1.6
> Reporter: Dyre Tjeldvoll
> Assigned To: Dyre Tjeldvoll
> Priority: Trivial
> Fix For: 10.3.0.0
>
>
> Clock currently inherits from Hashtable, but the use of Hashtable is really
> an implementation detail that would benefit from being hidden as private
> member. All access to the hashtable happens inside sychronized blocks so it
> is safe to substitute a HashMap. This change appears to trigger a small
> increase in throughput, as measured by the average TPS number obtained by
> running the select client from DERBY-1961 repeatedly.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
