DSXchange

A recent DSX newsletter contained a white paper on hashed files. Check it out.

It's cleaner to just create a new project for each major release. Just make the project name a variable in any scripts that initiate jobstreams, so that you can "version" releases in new projects. Best practice tip

Did you write the inserts to a file? Did you write the updates to a file? I bet you directly wrote to the database. If your hash file only contains the necessary rows for processing, you'll find the hash file is more performant. Trust Craig and I on this.

The whole row is a single text string value, there's only one column defined. Simple design.

Use a Sequential stage with 1 column defined to read the entire row as a single column and then use a combination of FIELD and concatenation statements to assemble the changed row to output to another Sequential file again as a single column.

Here, of course! There's no central document that lists all error codes across all APIs, Clients, and Server jobs. This site is a great tool. All of your issues can be found here with solutions, so use the Search tool.

Consider add an OR condition to evaluate to TRUE if there is no date in the hash reference lookup. Something like:

If you take the natural keys in your source data, load them into a work table in your target database, then inner join those keys to your target table and spool those rows to the hashed lookup file, your hashed file will only contain those rows found in the database. During transformation, reference...

Your question takes a book to answer. Consider reading Ralph Kimballs ETL Toolkit for starters. Ultimately, it all depends on budget, time to deliver, and complexity of solution. Best Practices for one company are different than the next. A best recommendation would be to engage a consultant who has...

During transformation split your data into rows for insertion and rows for updating. Bulk load your inserts, then perform the updates by either bulk loading into a work table and "merge" the data, or simply use update existing rows only. This method will be the fastest.

DSExecute is an API. Read your DS BASIC manual for more details. The command line interface program dsjob is well documented in its own manual under your Start menu.

Check here for an example script:

http://www.dsxchange.com/viewtopic.php?t=85578

The error message indicates 8192 bytes is the limit.

Arnd: The problem is a long constraint expression (See other posts regarding this).

The transformer can only call this API when the stage finishes, which means the after-transformer routine call. To do it for every row is expensive, and since there's no mechanism to tell you the last row being processed you can't keep count by yourself in a stage variable. Therefore, to be able to ...

Break your long constraint into sections using stage variables, and then compare a set of stage variables in the constraint. Your text is too long.