asked on

Problem with decode in the where clause

I have a problem with the following query:

SELECT
TOL.product_heading,
TOL.tenor_category,
TOL.tolerance_pct,
RE.product_heading_order,
decode(TOL.tenor_category,'SHORT',1000000,'MED',2000000,'LONG',3000000) AS FROM_TENOR,
RE.trade_id,
RE.dealer,
RE.profit_centre,
RE.counterparty_code,
TO_CHAR(DMO_FUNCTIONS.TO_ORACLE_DATE(value_date,'YYYYMMDD'),'DD-Mon-YYYY') AS VALUE_DATE,
RE.instrument,
RE.instrument_description,
RE.rate,
TEMP.avg_rate,
'Tolerances used in this report' AS TOLERANCE_TEXT
FROM
dmo_days_prodheading_rate RE,
dmo_rate_tolerance_lookup TOL,
(SELECT
AVG(RE.rate) AS AVG_RATE,
RE.product_heading,
DECODE (GREATEST (2555, daysbucket), 2555,'SHORT',
DECODE (GREATEST (5475, daysbucket), 5475,'MED','LONG')) AS TENOR_CATEGORY
FROM
dmo_days_prodheading_rate RE
WHERE
RE.product_heading_order >19
GROUP BY
RE.product_heading,
DECODE (GREATEST (2555, daysbucket), 2555,'SHORT',
DECODE (GREATEST (5475, daysbucket), 5475,'MED','LONG'))
)TEMP
WHERE
TEMP.product_heading = TOL.product_heading AND
TEMP.tenor_category = TOL.tenor_category AND
TOL.product_heading = RE.product_heading AND
TOL.tenor_category = DECODE (GREATEST (2555, daysbucket), 2555,'SHORT',
DECODE (GREATEST (5475, daysbucket), 5475,'MED','LONG')) AND
product_heading_order >19 AND
rate NOT BETWEEN
(AVG_RATE - (AVG_RATE*TOL.Tolerance_pct/100)) AND
(AVG_RATE + (AVG_RATE*TOL.Tolerance_pct/100))

I get the following error:

ORA - 01843: Not a valid month

The problem appears to lie with the
TOL.tenor_category = DECODE (GREATEST (2555, daysbucket), 2555,'SHORT',
DECODE (GREATEST (5475, daysbucket), 5475,'MED','LONG'))

line as commenting this out allows the query to run.

Why do I get the 01843 error when this line is included? The daysbucket field is a number generated by subtracting 2 dates. The TOL.tenor_category field is a varchar2(20) field.

Thanks,

Rich.

seazodiac

How do you generate daysbucket?

the reason why you get this error is one of the 2 dates you are using to derive daysbucket does not have a valid date format

bvanderveen

Looks like this error is from:
>> TO_CHAR(DMO_FUNCTIONS.TO_ORACLE_DATE(value_date,'YYYYMMDD'),'DD-Mon-YYYY') AS VALUE_DATE,

Don't know what the function you are calling is, but "not a valid month" is from a character-to-date conversion error. Look at what is being generated by:
DMO_FUNCTIONS.TO_ORACLE_DATE(value_date,'YYYYMMDD')
and see if it is an Oracle date.

andrewst

My money would be on this line:

TO_CHAR(DMO_FUNCTIONS.TO_ORACLE_DATE(value_date,'YYYYMMDD'),'DD-Mon-YYYY') AS VALUE_DATE,

What does DMO_FUNCTIONS.TO_ORACLE_DATE look like? If it returns a CHAR value like '20040825' then the line above is equivalent to:

TO_CHAR(TO_CHAR(value_date,'YYYYMMDD'),'DD-Mon-YYYY') AS VALUE_DATE,

... which doesn't make sense.

OTOH, if it takes a char parameter and returns a DATE then is value_date a character string in the required 'YYYYMMDD' format?

bvanderveen

Seazodiac may be right. I was assuming that daysbucket was stored as an integer value, not derived. If it is derived, you have a problem with the date conversions in this computation, as he said.

RichardKnight

ASKER

The value date was the first place I looked as it was the obvious place to try. Unfortunately commenting this out doesn't change things. The whole query runs correctly until you put in the decode join:

TOL.tenor_category = DECODE (GREATEST (2555, daysbucket), 2555,'SHORT',
DECODE (GREATEST (5475, daysbucket), 5475,'MED','LONG'))

and I can't see why this gives a problem.

seazodiac

Rich, as I said in the very first post, how do you derive daysbucket? that's the key, which most often is ignored by people..

RichardKnight

ASKER

Daysbucket comes from the dmo_days_prodheading_rate view:

decode (FU.product_type, 'RPO',TO_DATE (FU.effective_end_date,'YYYYMMDD') -
TO_DATE (FU.effective_start_date,'YYYYMMDD'),
TO_DATE (FU.expir_matur_century,'YYYYMMDD') -
decode(least(FU.deliv_century, expir_matur_century),
expir_matur_century, TO_DATE(expir_matur_century,'YYYYMMDD'),
TO_DATE(FU.deliv_century,'YYYYMMDD'))) AS DAYSBUCKET,

The code used to generate the view works fine when run as a standalone query.

bvanderveen

>>The code used to generate the view works fine when run as a standalone query.

Try scrolling through all the records in the view. I'll bet you get an error. But you won't see it until you scroll through, and the bad formatting is encountered.

seazodiac

is FU a table or view?

can you try this and post back if you have any errors?

select TO_DATE (FU.effective_end_date,'YYYYMMDD'),
TO_DATE (FU.effective_start_date,'YYYYMMDD'),
TO_DATE (FU.expir_matur_century,'YYYYMMDD'),
TO_DATE(FU.deliv_century,'YYYYMMDD')
from FU; --you need to replace FU with the real table or view name here.

kamar_k

Try to use the function TO_NUMBER(daysbucket) instead of daysbucket.

RichardKnight

ASKER

seazodiac - your suggested code does produce an error but I knew it would. There are various checks and clauses in the where clause (I didn't show the where clause when showing my daysbucket code)that prevent any dates getting through that may cause an error. Daysbucket is included in a view that runs perfectly and all of the daysbucket values are as expected with no errors.

RichardKnight

ASKER

It turns out that the problem was related to the way that Oracle 'optimises' queries.
I was using views and inline views and the predicates were getting mixed up when Oracle was combining the code into one uber-query. A simple 'order by' in the offending view has fixed the problem. If there is a better way of dealing with this not using order by (or rownum or by aggregating) I'd be grateful to hear it. Oracle suggest using decode around any columns that may cause problems and setting the offending values in those columns to NULL. This was too fiddly for my needs - I chose the easy route!

ASKER CERTIFIED SOLUTION

modulo

membership

This solution is only available to members.

To access this solution, you must be a member of Experts Exchange.

Start Free Trial