DSXchange

The 0x0a characters are the line separators used by your files, so they wouldn't show up in the data.

single-thread a transform stage (sequential as opposed to parallel, or a 1-node configuration) and then put a constraint "@INROWNUM < 6" in.

You are lucky in that you are still at Version 7, recovery at version 8 is much more time-consuming and complex and in certain cases you cannot recover, i.e. the project name is forever unusable unless you do a complete re-install.

What character set is your source XML file in? What is your DataStage NLS character set? Once you know those 2 answers you will know how to fix your problem.

IF "FIELD_NAME" can have a null value, then the derivation should be

I just noticed something from the first post. Stage Variables cannot be null, so you need to ensure that you work around that limitation with the null-handling functions (i.e. IsNull(), NullToValue() )

Note that 4000 is bytes. If you are using NLS then the number of characters displayable in 4000 bytes is less and will cause the job to abort.

What are the datatypes used and can you compute a rough value for Mb/Second written to the file? Also, change the output to a sequential file on /dev/null or a dummy copy stage and find out what speed the rest of the job could work at - that will ensure that the bottleneck really is the dataset.

Waht database and which stage are you using?

The ODBC connector at Version 8 has a "transaction -> record count" value that you can set.

The Enterprise stage, set the "array size" to 1, then you will get an option to set the "row commit interval".

Also remember that the working directory of the job is in the project directory, so you should always specify a full and absolute path to your filename.

Sort the data by descending value, then limit the output to the first 5 rows.

Imperator Yollson,

this means that you have a null value in a row in column "F" and that the column is not nullable. This is data driven and cannot be solved unless you make the column nullable - but you need to know if this null value is valid.

I am afraid that I still don't quite understand, would it be possible to give a short simplified example of what you wish to do?

Have your SQL split the CLOB into n columns of VarChar(2000) each and then use either a transform stage or Column Import stage to concatenate the list of VarChar columns into one large one. This method is being used on my current project because we have CLOBs and BLOBs of up to 60Kb (very wordy XML)...