[
https://issues.apache.org/jira/browse/DERBY-2920?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Knut Anders Hatlen updated DERBY-2920:
--------------------------------------
Attachment: ReadInts.java
Thanks, Mike. I was wondering about that myself, so I had started a test run
with the attached ReadInts test class. The test is similar to what you
suggested, but the table was smaller (3 integer columns, 10000 rows).
After 100 runs with trunk and 100 runs with the patch (each run performs a full
table scan 20000 times for warmup and 10000 times while collecting data), I
found that on average there was a 0.02% difference. That is, no measurable
difference. This was with JDK 7u9 on Solaris 11.
I'll change the size of the test table to 10x1000000 and see if that changes
anything.
> Share code between readExternal() and readExternalFromArray()
> -------------------------------------------------------------
>
> Key: DERBY-2920
> URL: https://issues.apache.org/jira/browse/DERBY-2920
> Project: Derby
> Issue Type: Improvement
> Components: Services, SQL
> Reporter: Knut Anders Hatlen
> Assignee: Knut Anders Hatlen
> Priority: Minor
> Labels: derby_triage10_10
> Attachments: d2920-1a.diff, ReadInts.java
>
>
> Most of the implementations of
> DataValueDescriptor.readExternalFromArray(ArrayInputStream) are identical to
> their corresponding Externalizable.readExternal(ObjectInput) methods. Since
> ArrayInputStream implements ObjectInput, readExternalFromArray() could in
> those cases just have forwarded calls to readExternal() instead of
> duplicating the code. A default forwarding implementation of
> readExternalFromArray() could be placed in
> org.apache.derby.iapi.types.DataType, and all the existing implementations,
> except those with optimizations for ArrayInputStream, could be removed.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
