Mailing-List: contact derby-dev-help@db.apache.org; run by ezmlm
Precedence: bulk
Reply-To: <derby-dev@db.apache.org>
Date: Thu, 18 Sep 2014 03:53:34 +0000 (UTC)
From: "ASF subversion and git services (JIRA)" <jira@apache.org>
To: derby-dev@db.apache.org
Message-ID: <JIRA.12742354.1411011570000.63024.1411012414411@Atlassian.JIRA>
In-Reply-To: <JIRA.12742354.1411011570000@Atlassian.JIRA>
References: <JIRA.12742354.1411011570000@Atlassian.JIRA>
 <JIRA.12742354.1411011570736@arcas>
Subject: [jira] [Commented] (DERBY-6743) MERGE code needs changes since now
 UPDATE statement can change the identity column value using DEFAULT
 keyword.
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: quoted-printable


    [ https://issues.apache.org/jira/browse/DERBY-6743?page=3Dcom.atlassian=
.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=3D1413=
8456#comment-14138456 ]=20

ASF subversion and git services commented on DERBY-6743:
--------------------------------------------------------

Commit 1625884 from [~mamtas] in branch 'code/trunk'
[ https://svn.apache.org/r1625884 ]

CERBY-6414(Incorrect handling when using an UPDATE to SET an identity colum=
n to DEFAULT)

I have created DERBY-6742 for JDBC part of this feature so we can generate =
auto generated resultset for an update statement updating generated columns=
. This implementation is supported through Statement.RETURN_GENERATED_KEYS =
flag.

Junit tests for this feature are in GeneratedColumnsTest.java. Upgrade test=
s are in Changes10_12.java. Upgrade test shows that a soft upgrade from pre=
-10.11 will not support update of generated columns using DEFAULT but 10.11=
 to trunk soft upgrade will allow it. This is because starting 10.11, we ha=
ve started using sequence generator to create unique ids. If we want to sup=
port updating of idetity columns for pre-10.11 releases, we will need to ma=
intain the code for old way of generating unique ids. In order to avoid tha=
t, this feature is available to only 10.11 db during soft upgrade.

For UPDATE of identity column to work in the MERGE sql, we need to make cha=
nges to MERGE code(DERBY-6743). MERGE already has required code for insert =
putting DEFAULT in generated columns. Implementation needs to be added for =
update putting DEFAULT in generated columns in case of MERGE. In the mean t=
ime, I have renamed existing junit test test_015_bug_6414 to atest_015_bug_=
6414 in MergeStatementTest.java so we do not run into failures.

Some detail about the part of the implementation that was tricky for DERBY-=
6414.
The existing code was originally written to handle inserting values in iden=
tity columns using DEFAULT. In case of insert, InsertResultSet uses ColumnD=
escriptors in resultDescription to find the type of the generated columns. =
This data structure holds the column descriptors of all the columns in the =
table. All the columns are in this data structure because even though INSER=
T statement may not explicitly assign a value to each and every column in t=
he table, all the columns end up getting some value in them through an INSE=
RT statement. The code in InsertResultSet.initializeAIcache method relies o=
n availability of all the columns type information. But in case of Update, =
resultDescription does not include all the columns in the table. It only ha=
s the columns being touched by the Update statement(the rest of the columns=
 in the table will retain their original values), and for each of those tou=
ched columns, it has a duplicate entry in resultDescription in order to hav=
e before and after values for the changed column values. This difference in=
 array content of resultDescription requires us to have separate implementa=
tion of initializeAIcache method for insert and update.

> MERGE code needs changes since now UPDATE statement can change the identi=
ty column value using DEFAULT keyword.
> -------------------------------------------------------------------------=
--------------------------------------
>
>                 Key: DERBY-6743
>                 URL: https://issues.apache.org/jira/browse/DERBY-6743
>             Project: Derby
>          Issue Type: Improvement
>          Components: SQL
>    Affects Versions: 10.12.0.0
>            Reporter: Mamta A. Satoor
>
> DERBY-6414(Incorrect handling when using an UPDATE to SET an identity col=
umn to DEFAULT) has been fixed in Derby 10.12. This will require changes in=
 MERGE code in order for UPDATE of identity column to work in the MERGE sql=
.


--
This message was sent by Atlassian JIRA
(v6.3.4#6332)