DSXchange

Go for the Object File option while creating the parallel routine in DataStage. Provide the path of your object file(.o). The external name will be the name of your function as its in the code.

You can use either one. You might have more control in a routine as your can do error handling and produce appropriate messages in logs. If you dont need all that, then go for execute command stage.

Well I wrote a small piece of code in the morning and was stuck in between how to return a value with all rows. I was getting some junk data. Therefore my misinterepretation and hence my assumption that I made earlier. But your example Ray, was like a wake up call as I use that all the time too. I l...

My apologies. I was trying to build the same but was messing up at a place. I take my words back. If I get this done in a while Ill post the code.
My apologies once again.

Actually C wont work. Come to think of it, C variables wont hold values with line feeds. The workaround could have been writing to a file but that would require the execution to be sequential. Parallel writes wont work. I guess its either a Basic transformer or an awk script. C will work but it has ...

ray.wurlod wrote:At row #1 it's unlikely to be a space issue. Code -100 is often used to signal "end of data". Are you trying to sort an empty file?

Didnt pay attention to that, opsi

Me thinks the same

Is this table definition of some database table? If yes then you can query the system tables and get all the column names.

Search for it, it has been covered time and again.

You dont need to add it. It will appear on its own. Test it out.

First get your small program compiled. Its fine the way it is. It will compile and run. But when your ready to convert this into a parallel routine, you need to get rid of main() and give it a function name like int AddNum(). So if i have to translate your code to a parallel routine it will be #incl...

Its hard to give figures without knowing the system and performing some tests. It depends upon how much is your data size, how many nodes are utilized etc etc etc.

Welcome Aboard
I think you are running into some sort of limitation, temp space maybe. How many records are you trying to sort? Cut that in half, do you still get this error, if yes, cut it further down and see if the problem still persists.

ArndW wrote:The DataStage read of sequential files is pretty efficient; and DSParams is small file (if the file had millions of records I might reconsider, but this is going to take milliseconds to process).

Thanks for the clarification ArndW.

I believe you can write a C routine that includes a header file which has these functions defined to be called from a px transformer. I vaguely remember going through them in Advanced parallel developers guide. And am not at work today so cannot access the pdf files to confirm that.