Mailing-List: contact derby-dev-help@db.apache.org; run by ezmlm
Precedence: bulk
Reply-To: <derby-dev@db.apache.org>
Message-ID: <1056522.1173162324356.JavaMail.root@brutus>
Date: Mon, 5 Mar 2007 22:25:24 -0800 (PST)
From: "Mayuresh Nirhali (JIRA)" <jira@apache.org>
To: derby-dev@db.apache.org
Subject: [jira] Commented: (DERBY-2386) timestampdiff function fails when
 using SQL_TSI_FRAC_SECOND for datepart parameter, except for very small
 intervals
In-Reply-To: <5899350.1172769710673.JavaMail.jira@brutus>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit


    [ https://issues.apache.org/jira/browse/DERBY-2386?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12478275 ] 

Mayuresh Nirhali commented on DERBY-2386:
-----------------------------------------

The SQL_TSI_FRAC_SECOND type is used for fractional timestamp differences. IF the expected difference is in seconds then SQL_TSI_SECOND should be used. This is a solution to the error that is being reported in the description. So, I think this is not a bug.

http://mail-archives.apache.org/mod_mbox/db-derby-dev/200511.mbox/%3Cc25576af0511281539te6c1eefw48a9dcc993ad709c@mail.gmail.com%3E

The discussion above on the derby-dev list on the same topic and that talks about returning BIGINT instead of INT. I am inclined towards NOT doing that because there is no gap between the values covered by TSI_FRAC_SECOND and TSI_SECOND. I tried a small experiment as below,

<snip>
ij> select * from t2;
ID         |STARTDATE                 |ENDDATE
-----------------------------------------------------------------
5          |2006-11-20 04:20:00.0     |2006-11-20 04:20:00.2
6          |2006-11-20 04:20:00.0     |2006-11-20 04:20:30.0
7          |2006-11-20 04:20:00.0     |2006-11-20 04:20:00.3
8          |2006-11-20 04:20:00.0     |2006-11-20 04:20:00.9999

4 rows selected
ij> select {fn timestampdiff(SQL_TSI_FRAC_SECOND, startdate, enddate)} as diff from t2 where id =8;
DIFF
-----------
999900000

1 row selected
</snip>

So, for all values less than 1 second the function will not fail if FRAC_SECOND is used and for values in seconds TSI_SECOND should be used. 

The only downside of this existing approach is that for difference in seconds (when TSI_SECOND is used) the fractional difference cannot be identified.

please share your thoughts.

> timestampdiff function fails when using SQL_TSI_FRAC_SECOND for datepart parameter, except for very small intervals
> -------------------------------------------------------------------------------------------------------------------
>
>                 Key: DERBY-2386
>                 URL: https://issues.apache.org/jira/browse/DERBY-2386
>             Project: Derby
>          Issue Type: Bug
>          Components: SQL
>    Affects Versions: 10.2.2.0
>         Environment: SUSE Linux Enterprise Desktop 10
>            Reporter: Don Smith
>         Assigned To: Mayuresh Nirhali
>
> Using the timestampdiff function produces and integer overflow except for very small intervals. Error message is:
> [Error Code: -1, SQL State: 22003]  The resulting value is outside the range for the data type INTEGER.
> I inserted the following row into my test table:
> insert into datetest (ID, 
> startdate, 
> enddate) values (
> 5, 
> '2006-11-20 04:20:00.0', 
> '2006-11-20 04:20:00.2');
> This test row works:
> select {fn timestampdiff(SQL_TSI_FRAC_SECOND, startdate, enddate)} as diff from datetest where id = 5
> DIFF
> 200000000
> The value also looks too large, which could be exacerbating the problem.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.