DSXchange

Why don't you use the reject link and handling builtin to the Oracle Enterprise stage?

If you troll through the DSXchange posts and look for "DS_JOBS" and job numbers you will see many examples of how to associate a job number with the appropriate job name; then you can search the sub-directories and look for your keyowords in the files.

Hello and welcome to DSXchange, Pierre. IYou posted your question in the "Site/Forum" section, where it does not belong and it most likely won't get answered until it is in the correct forum. It would help if you could also state why you think the files are held in memory - are you getting error mes...

In the project directory for each job number there will be a subdirectory named RT_SC{nnn} which will contain osh scripts.

Ok, your Japanese text deals with text attachments that aren't supported... It works in your SQL Server because you told it that you have LATIN-1 data. When you inserted your cut-and-pasted data it did no conversion, assuming that your screen text is also LATIN-1. The same happens when you read from...

I could try to explain, but could not come close to the numerous detailed descriptions in the Parallel Job Developer's Guide. The documentation details reject handling for all stages that support this output.

Where are you calling the shell script from?

That is a likely source of the error. You hadn't mentioned 2 ODBC stages before, have you written so that you have one insert and one update stage?

It depends where you want to set the environment variables. It won't work within jobs, since the shell in which the variables are set is lost when returning to the job.

In a PX job that has more than one node, you need to use @INROWNUM with @PARTNUM to get a unique row number.

In PX you have the "normal", or PX transform stage. This is compiled to c++ code and then bound in at runtime. You can also call a BASIC transform stage which uses an instance of the DS-Engine at runtime to execute the code. Due to several factors the use of the BASIC transform stage in PX job is di...

Ok - sounds like you did try a search (there's nothing that gets frequent posters angry faster than seeing someone post a question whose answer is available using the search function). Repartitioning and sorting unnecessarily will kill performance and use lots of space as so many implementations dis...

Sure, many of us occasionally read the manuals but make sure we are not watched or disturbed in the process. I will usually only admit under duress that I even own a manual, and only the utmost coercion will force me to admit having actually read from it :D

I can see how a Transform stage can act like a filter, but the sort portion of the question is unclear to me as a transform stage doesn't have anything like that functionality. Any job without a sort is going to be faster than one with.

I'm not a site where I can test this, but recall that you need a record-id column that is used, in a manner similar to a switch stage, to identify a column position that differentiates the different record types. I know I've gotten it working, but I did struggle with it.