APT_CombinedOperatorController,1: Null string argument.

rgr · Post by **rgr** » Mon Aug 10, 2009 1:23 pm

Hi,
Iam using following exp IsNull(column_name) Then '' Else If Num(column_name)) Then column_name) Else 'aaa'
the source data type is char. The above expression is working fine in server but in parallel Iam getting APT_CombinedOperatorController,1: Null string argument.warning, even I tried Using NVL function in source query.
It would be helpful some one provide resolution.

Thanks In Adavance

Rgr

ray.wurlod · Post by **ray.wurlod** » Mon Aug 10, 2009 5:20 pm

Start by disabling operator combination so you can see which stage is actually generating the error. It may not be where you think.

rgr · Post by **rgr** » Tue Aug 11, 2009 10:37 am

ray.wurlod wrote:Start by disabling operator combination so you can see which stage is actually generating the error. It may not be where you think. ...

Thanks Ray,
I have disabled operation cobination it is showing in the same transformer
where Iam using above exp.

Thanks,
RGR

ArndW · Post by **ArndW** » Tue Aug 11, 2009 12:10 pm

If you replace the derivation with a simple constant value, does the job run without that warning? The expression itself does not look like it would cause that error message (I assume you didn't cut-and-paste as there is a sytactical error with parenthesis in the statement).

rgr · Post by **rgr** » Wed Aug 12, 2009 12:07 pm

Ray,
still I am getting same warning

ray.wurlod · Post by **ray.wurlod** » Wed Aug 12, 2009 4:39 pm

Is your expression missing If at its beginning?

rgr · Post by **rgr** » Fri Aug 14, 2009 9:41 am

ray.wurlod wrote:Is your expression missing If at its beginning? ...

No

Thanks,
RGR

ArndW · Post by **ArndW** » Fri Aug 14, 2009 9:53 am

If you are still getting the warning after having removed the statement then I would suggest you look elsewhere for the cause.

rcanaran · Post by **rcanaran** » Wed Aug 19, 2009 12:10 pm

I am having a similar problem.

Stage Variable Defintion populated with :
If IsNull(linkname.VARIABLE_NAME)
Then "000"
Else If Len(Trim(linkname.VARIABLE_NAME)) <> 3
Then "000"
Else If (Num(linkname.VARIABLE_NAME))
Then linkname.VARIABLE_NAME
Else "000"

If the incoming variable string is not a 3-digit number, then the sv should contain string "000" else it should contain the value from the incoming variable.

It appears that the NUM function is evaluating the variable even though the IF/ELse structure should have bypassed this. Perhaps DS parallel evaluates all the nested conditions to T/F and then processes the structure acording to the T/F results, isntead of only processing the variables/functions it needs to.

I forced this by having another sv prior to this which evaluated the null value to a string containing '' (an EMPTY string; I guess this differs from a NULL string). The (now 2nd, but really the original) sv that I was working with origianlly now uses the 1st sv instead of the incoming variable.

This solved the situation for me, but I consider this a workaround. I think the real solution would involve DS parallel evaluating only what it needs to and doing this in a single sv.

chulett · Post by **chulett** » Wed Aug 19, 2009 12:17 pm

"(an EMPTY string; I guess this differs from a NULL string)"

It most definitely does. Differ, that is.

rcanaran · Post by **rcanaran** » Wed Aug 19, 2009 12:29 pm

Thanks Craig. And I love the tag line.
But any insight into how DS is processing the derivation?

chulett · Post by **chulett** » Wed Aug 19, 2009 1:52 pm

To the best of my knowledge, it basically goes "left to right" and stops evaluating as soon as it can. Your problem may have strictly been a null versus empty string issue but I'm going to have to defer that judgement to those that have a more intimate knowledge of How Things Work, like Ray or Arnd.

ray.wurlod · Post by **ray.wurlod** » Wed Aug 19, 2009 6:40 pm

You can inspect the generated C++ code in the RT_SCnnn directory, where nnn is the job number.

In theory it's left to right, as Craig believes, but the optimizer may make "decisions" that change that. Inspecting the generated code is the only way you can be sure.

slosa · Post by **slosa** » Sat Aug 22, 2009 2:06 pm

I seen this before, i found out that this behaviour occurs only if

slosa · Post by **slosa** » Sat Aug 22, 2009 2:41 pm

sorry, forget my last post

It seems to be related to the type of the input coumn. If I set it to Vachar (unbounded) the warning disapears - if i use a bound (VarChar[10]) i see this error.