Optimize LOBStateTracker for non-locator servers

• Key: DERBY-3601
• Type: Improvement
• Status: Closed
• Resolution: Fixed
• Priority: Trivial
• Assignee: Kristian Waagan
• Reporter: Kristian Waagan
• Component/s: Network Client
• Affects Version/s: 10.4.1.3, 10.5.1.1
• Fix Version/s: 10.4.2.1, 10.5.1.1

Description

The following comments posted for DERBY-3571 should be addressed if we choose to allow multiple calls to the various getter methods on LOB columns, except for the getter methods returning streams:
----- Knut Anders wrote:
've tried out the 2a patch and it seems to work as intended. My only nits are:

  - LOBStateTracker.checkCurrentRow(): couldn't Arrays.fill() be moved inside the if block?
  - should discardState() and markAccessed() check the release flag?
  - should ResultSet.createLOBColumnTracker() use LOBStateTracker.NO_OP_TRACKER instead of allocating a new when serverSupportsLocators() returns false?
-----

Note that it is a requirement that we allow multiple calls to the getter methods, otherwise we need to maintain the state for non-locator LOB columns as well to track the number of accesses. Another options is to rely on a different mechanism to track accesses.

Comments

[Kristian Waagan]

'derby-3601-1a-comments_and_renaming.diff' changes some of the comments in LOBStateTracker and renames variables and methods to better reflect reality.
I think when it was first written it was supposed to take care of more than it does today. Currently, the class is only used to track whether a LOB has been published to the end-user or not (happens through ResultSet.get[BC]lob).
There shouldn't be any functional changes in patch 1a.

The next step is to implement the improvements suggested under this issue.
Patch ready for review.

[Dag H. Wanvik]

Changes looks good and understandable to me.

+1

[Kristian Waagan]

Thank you for the review, Dag!

I committed 'derby-3601-1a-comments_and_renaming.diff' to trunk with revision 690133.
I expect another patch to be ready soon.

[Kristian Waagan]

'derby-3601-2a-non_locator_optimization.diff' implements the suggested optimizations.
suites.All ran without failures on Solaris 10 with JDK 1.6.0.

Patch ready for review.

[Kristian Waagan]

Committed patch 2a to trunk with revision 701156.
I'll leave the issue open for a while and backport it if no problems are detected.

[Kristian Waagan]

Backported fixes to 10.4 with revision 705502.
Closing issue.