DSXchange

No.

Runtime propagation adds columns that are in the schema but not explicitly in your design. If anything, one could argue that this degrades performance (the extra processing to make that decision). In practice, though, it's negligible.

As I said, you can customize your Palette and have it wherever you want it. It's up to you what effect the stage will have. For example, if you're reading from a delimited text file (an operation that is necessarily sequential), then running that result into a BASIC Transformer stage does not contri...

For staging, where you don't need individual key-based access, a sequential file will always be substantially faster both to write and to read than a hashed file. The main reason for this is you can't sort the data by hashed file group number, since the hashing algorithms are not in the public domai...

If you want to clear them, just change from Append to Overwrite. If you don't want to clear them, leave the setting at Append. If you only want to clear them occasionally (such as at the beginning of the month) make a copy of your job, set the rule to Overwrite, but make sure it sends no rows (use a...

Training class DataStage Best Practices for one.

Freezing in an Australian winter (32C in Darwin!)

There's not really enough information to diagnose this one. I presume you're using an External Filter stage. Can you post the exact error or warning message that was generated? DataStage will normally insist on supplying a column because that column appears in the Columns grid. Have you tried removi...

Your approach only works if the hashed files have identical metadata in terms of how data are stored in them.

DataStage is absolutely metadata driven. If you need different metadata (table definition) then you need a new job design.

Welcome aboard! :D When there is one process writing to a buffer and another process draining the buffer downstream, the downstream process needs a mechanism for detecting whether something has gone wrong with the upstream process. No new rows for 10 seconds (which is a long time if you're processin...

One or two "sandpit" projects in which developers can experiment with ideas is often useful.

You can capture them in the same job, but only reliably in an after-job subroutine, which executes only after everything else has finished.

Augment the documentation on the General tab of your routine.
Add idiot-proofing to the routine's code.
Stop paying VAT.

Update relies on a key match, not a complete row match. Given that your SQL Server table does not have a key, then you're up for some pretty clever detection strategies, or simply you allow duplicate rows to be inserted.

Some versions back there was a problem where the exit status was only reporting one byte of an integer. But I believe this was fixed. Could it have been reintroduced? What do "they" say?

Please post your reasons for believing that there are bugs. Be specific about stage types and nature of bug in each case. I usually advocate not using RCP but not because of any bug. My reason is that tools such as MetaStage can't provide usage/lineage analysis of columns that are generated "on the ...

The point about using the BASIC Transformer stage is that the regular parallel Transformer stage does not have Iconv and Oconv available. Without these it's a really ugly problem; you will have to write your own function. As I said earlier, you need to read the timestamp as a string, pick the pieces...