[ http://issues.apache.org/jira/browse/DERBY-424?page=all ]
Mamta A. Satoor updated DERBY-424:
----------------------------------
Attachment: Derby424SessionSchemaCaching.txt
I have attached a review package for this bug, hopefully, in time for 10.1.2
release.
The files affected by this change are
M java\engine\org\apache\derby\impl\sql\GenericStatement.java
M java\engine\org\apache\derby\impl\sql\compile\FromList.java
M java\engine\org\apache\derby\impl\sql\GenericPreparedStatement.java
M
java\testing\org\apache\derbyTesting\functionTests\tests\lang\declareGlobalTempTableJava.java
M
java\testing\org\apache\derbyTesting\functionTests\master\declareGlobalTempTableJava.out
The changes for this fix are very localized, affecting only 3 files in Derby
engine. Basically, the problem is that, during the compile phase of views, the
reference to the view gets replaced by the view definition, which causes us to
loose the information that the view might have belonged in SESSION schema. In
order to fix this, during the bind phase in FromList, before the view gets
replaced by its definition, I find out if the view is from SESSION schema, If
yes, then I save this information in FromList and this gets used by FromList
when it is asked if it has any items referencing SESSION schema objects. This
information is again lost during the optimization and generate phase and hence
I moved the check for SESSION schema reference to right after the bind phase in
GenericStatement. If there is a reference to SESSION schema object,
GenericStatement will remove the statement from the cache.
I have put in quite a big of comments in the code which hopefully will make the
patch easier to understand. I have added a new test for this and have run all
the tests with no failures using Sun's jdk142 on Windows XP machine.
> Queryplan for a query using SESSION schema view is incorrectly put in
> statement cache. This could cause incorrect plan getting executed later if a
> temp. table is created with that name.
> -----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
>
> Key: DERBY-424
> URL: http://issues.apache.org/jira/browse/DERBY-424
> Project: Derby
> Type: Bug
> Components: SQL
> Versions: 10.1.1.0, 10.2.0.0
> Environment: generic
> Reporter: Satheesh Bandaram
> Assignee: Mamta A. Satoor
> Priority: Minor
> Attachments: Derby424SessionSchemaCaching.txt
>
> See DERBY-405 for some discussion related this issue. I suspect this is
> because of statement caching. Derby doesn't seem to recognise we have a
> temporary table that just overloaded an existing physical table.
> It would have been good to avoid permanent tables/views/synonyms in SESSION
> schema. Not sure what we should do now about this, though.
> ij> create view session.st1 as select * from t;
> 0 rows inserted/updated/deleted
> ij> select * from session.st1;
> I |J |K
> -----------------------------------
> 1 |1 |NULL
> 2 |2 |NULL
> 3 |3 |NULL
> 4 |4 |NULL
> 4 rows selected
> ij> select * from t;
> I |J |K
> -----------------------------------
> 1 |1 |NULL
> 2 |2 |NULL
> 3 |3 |NULL
> 4 |4 |NULL
> 4 rows selected
> ij> declare global temporary table st1(c11 int, c12 int) on commit preserve
> rows
> not logged;
> 0 rows inserted/updated/deleted
> ij> select * from session.st1;
> I |J |K
> -----------------------------------
> 1 |1 |NULL
> 2 |2 |NULL
> 3 |3 |NULL
> 4 |4 |NULL
> 4 rows selected
> ij> select * from session.st1; <==== This statement has an extra space
> between FROM and session.st1
> C11 |C12
> -----------------------
> 0 rows selected
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://issues.apache.org/jira/secure/Administrators.jspa
-
For more information on JIRA, see:
http://www.atlassian.com/software/jira
