DSXchange

The dsjob executable is your gateway to everything you need to do. It is a command line executable, so you can use the switch to list all jobs in a project. Then, you can cycle thru all jobs and issue the reset command. All of this is available from the system command prompt in both Unix and Windoze...

I could do this with multiple Odbc stages and use a constraint in a transformer stage and decide the target. You supplied your own answer. An alternative is to derive the target tablename as an output column and stream to a sequential file. Then, use an after-job cutter script to separate the singl...

I assume you mean a DS hash file, and not a Universe/Unidata hash file. There are two ways: 1. Go look at the job that creates and populates the hash file. As a courtesy, save the definition into the DS Manager library. 2. Import hash file metadata using DS Manager to get it into the library if you ...

No. If you find that you need historical log information you should extract it and move it into a separate DB. You can find techniques if you search this forum. As for the reasons why the log is cleared, it's actually removed. Anytime you import a job, all of its existing support tables are dropped ...

Are you using this function and then applying a constraint to throw away rows in a subsequent transformer stage? All I see is that if this function is called for the first time with a new argument value that it starts numbering from 1000. Otherwise, it constantly returns the current value in the has...

Please post the logic for your routine. We can't help you otherwise.

Optimize your spool of data out of DB2. Write a job that is just DB2 --> xfm --> seq and see how fast that runs. Then, pick a numeric column, something like the primary key, and instantiate that job passing in the parameters to divide the data in the SQL. You'll find that the degree of instantiantio...

Unless you have obscene amounts of source rows, why not use a relational database to do the work for you? Load the primarykey, startdate, and endate into a work table. You hopefully have a table of dates somewhere (TIME dimension in Kimball speak). You could select your work table joining to the TIM...

I haven't done a CDR (Call detail record?) warehouse, but the principles still apply. If you have a VLD (Very Large Dimension), then you typically have a high volume of fact records processing daily, this only makes sense. In a high volume fact processing solution, you MUST employ job instantiation ...

A hard crash that you are describing is tricky to programmatically recover. Who watches the watcher? If the main controlling job itself crashes, corrupting its log, status, and config files, then how does that get automatically rectified? I think in the event of a catastraphic failure, such as a reb...

You did not answer my first point about verifying that the first row listed is the same in both hash file viewings.

If you have a hash file and two jobs have the same exact definition, then viewing data in both jobs should show the EXACT SAME ROWS. It will read the hash file the same way/order, so the first row listed in both jobs should be the same row. Is it? If not, you're looking at different files. If it is,...

You have a metadata problem. If you swear that the metadata in the reference stage is the same as the metadata in the job that creates/populates the hash file, then recheck the name of the hash file and/or the path.

This might be a little interesting: m However, to directly answer your question, the differences between the two from a hardware perspective makes it a difficult comparison. Intel CPUs are really fast, whereas Unix boxes tend to be slower. So, it depends on how you design your jobs. If your jobs are...

How to check if a variable is a NULL: If ISNULL(variable) Then dothis Else dothat How to get the first letter of a variable: LEFT(variable,1) or variable[1,1] In a derivation, you would do something like this to choose between two link values: If LEFT(link.column_A,1) = 'X' T...