By using this site, you agree to our updated Privacy Policy and our Terms of Use. Manage your Cookies Settings.
438,856 Members | 2,163 Online
Bytes IT Community
+ Ask a Question
Need help? Post your question and get tips & solutions from a community of 438,856 IT Pros & Developers. It's quick & easy.

Case Statement Within A Select Where 2 or More Instances Of The Record Exist.

P: n/a
Ok,
I have a data warehouse that I am pulling records from using Oracle
SQL. I have a select statement that looks like the one below. Now what
I need to do is where the astrics are **** create a case statement or
whatever it is in Oracle to say that for this record if a 1/19/2005
record exists then End_Date needs to be=1/19/2005 else get
End_Date=12/31/9999. Keep in mind that a record could have both a
1/19/2005 and 12/31/9999 instance of that account record. If 1/19
exists that takes presedent if it doesnt then 12/31/9999. The problem
is that the fields I pull from the table where the end_date is in
question change based on which date I pull(12/31/9999 being the most
recient which in some cases as you see I dont want.) so they are not
identical. This is tricky.
Please let me know if you can help.

SELECT
COLLECTOR_RESULTS.USER_ID,
COLLECTOR_RESULTS.LETTER_CODE,
COLLECTOR_RESULTS.ACCT_NUM AS ACCT_NUM,
COLLECTOR_RESULTS.ACTIVITY_DATE,
COLLECTOR_RESULTS.BEGIN_DATE,
COLLECTOR_RESULTS.COLLECTION_ACTIVITY_CODE,
COLLECTOR_RESULTS.PLACE_CALLED,
COLLECTOR_RESULTS.PARTY_CONTACTED_CODE,
COLLECTOR_RESULTS.ORIG_FUNC_AREA,
COLLECTOR_RESULTS.ORIG_STATE_NUMBER,
COLLECTOR_RESULTS.CACS_FUNCTION_CODE,
COLLECTOR_RESULTS.CACS_STATE_NUMBER,
COLLECTOR_RESULTS.STATE_POSITION,
COLLECTOR_RESULTS.TIME_OBTAINED,
COLLECTOR_RESULTS.TIME_RELEASED,
COLLECT_ACCT_SYS_DATA.DAYS_DELINQUENT_NUM,
sum(WMB.COLLECT_ACCT_SYS_DATA.PRINCIPAL_AMT)As PBal,
FROM
COLLECTOR_RESULTS,
COLLECT_ACCT_SYS_DATA,
COLLECT_ACCOUNT
WHERE
COLLECT_ACCOUNT.ACCT_NUM=COLLECT_ACCT_SYS_DATA.ACC T_NUM(+)
AND
COLLECT_ACCOUNT.LOCATION_CODE=COLLECT_ACCT_SYS_DAT A.LOCATION_CODE(+)
AND COLLECT_ACCOUNT.ACCT_NUM=COLLECTOR_RESULTS.ACCT_NU M(+)
AND COLLECT_ACCOUNT.LOCATION_CODE=COLLECTOR_RESULTS.LO CATION_CODE(+)
AND COLLECTOR_RESULTS.ACTIVITY_DATE =
to_date(''01/19/2005'',''mm/dd/yyyy'')
AND COLLECT_ACCOUNT.END_DATE = to_date(''12/31/9999'',''mm/dd/yyyy'')
AND COLLECT_ACCT_SYS_DATA.END_DATE = *****************

Jul 23 '05 #1
Share this Question
Share on Google+
1 Reply


P: n/a
On 20 Jan 2005 11:13:31 -0800, ph******@msn.com wrote:
Ok,
I have a data warehouse that I am pulling records from using Oracle
SQL.
Hi philipdm,

You posted this question in a newsgroup for MS SQL Server. I doubt you'll
get any specific Oracle help here.

But if I understand your requirements correctly, I think you can solve it
using only ANSI-standard SQL: correlated subquery, group by and aggregate
functions. I assume Oracle will have no trouble running those! I'm not
sure if the NULL handling requires sppecial attention (see notes down
below).

First, let me check if I correctly understand your requirements:
Keep in mind that a record could have both a
1/19/2005 and 12/31/9999 instance of that account record. If 1/19
exists that takes presedent if it doesnt then 12/31/9999.


The way I read this is: check collect_acct_sys_data for a particular
acct_num / location_code combinations. If there's a row for 1/19/2005, use
that. If there's a row for 12/31/9999 but no row for 1/19/2005, use the
12/31/9999 row. If there's no row for 1/19/2005 and no row for 12/31/9999,
use no row at all - the join will fail and the rows for the other tables
that use this acct_num / location_code combination won't be included in
the result set either.

I'd use something like the following. Note: I've used table aliasses and
converted the table and column names to lower case to improve readability;
I kept the (+) symbols you included and didn't change the format of the
to_date function calls - neither of these are known in MS SQL Server, so I
have no idea if they're right or wrong.

SELECT CR.User_ID,
.... (lots of other columns)
FROM Collector_Results AS CR,
Collector_Acct_Sys_Date AS CASD,
Collect_Account AS CA
WHERE CA.Acct_Num = CASD.Acct_Num(+)
AND CA.Location_Code = CASD.Location_Code(+)
AND CA.Acct_Num = CR.Acct_Num(+)
AND CA.Location_Code = CR.Location_Code(+)
AND CR.Activity_Date = to_date(''01/19/2005'',''mm/dd/yyyy'')
AND CA.End_Date = to_date(''12/31/9999'',''mm/dd/yyyy'')
AND CASD.End_Date =
(SELECT MIN(CASD2.End_Date)
FROM Collector_Acct_Sys_Date AS CASD2
WHERE CASD2.Acct_Num = CASD.Acct_Num(+) -- Do you need the
(+) here?
AND CASD2.Location_Code = CASD.Location_Code(+) -- Do you
need the (+) here?
AND ( CASD2.End_Date = to_date(''01/19/2005'',''mm/dd/yyyy'')
OR CASD2.End_Date = to_date(''12/31/9999'',''mm/dd/yyyy'')))

Final note: if the possibility exists that no row in CASD for a given
acct_num / location_code with either of the two dates, the subquery should
return NULL; the clause "AND CASD.End_Date = (subquery)" will evaluate to
"AND CASD.End_Date = NULL", which should not be true for any value of
CASD.End_Date (not even if CASD.End_Date is NULL!!). This is how NULLS
should be treated according to ANSI standard. If Oracle treats the result
of an empty subquery or comparison to NULL differently, then you should
tweak the query to get the correct results.

Best, Hugo
--

(Remove _NO_ and _SPAM_ to get my e-mail address)
Jul 23 '05 #2

This discussion thread is closed

Replies have been disabled for this discussion.