DSXchange

You must have ordered data so that the string is built in the right order. You can use a hash file to write out the row using your unique identifier. Reference the existing row, if it's not there, you write out the key and the text. If the row is there, write out the key and the reference column con...

You ran out of disk space.

Visit www.asktom.oracle.com and search his archives. This guy is the definitive expert on Oracle, and you can see a lot of discussion and benchmarking concerning using sqlldr. You can get his books, his Expert One-On-One is a must read for Oracle enthusiasts.

Did you search the forum? There's a lot of threads on this issue.

Sorry, didn't notice the topic was QualityStage.

To get a string of spaces, try SPACE(500). Another method is STR(" ", 500). Another method is to format a variable, padding it to the desired width and justification. This is FMT(variable, 'R#500') to right justify in a field of 500 spaces. Look in your DS BASIC manual for more information.

YES
YES
YES
YES

Writing is a whole other topic. But, keep in mind that hash reads have the ability to be LOCKED reads, so avoid that and you're completely safe and performance won't suffer.

There's no way. Every column requires a derivation.

Yes and it's not pretty. In the metadata there is a column for descriptions, but, if you reload the metadata, it gets overwitten. So, choose wisely.

The surrogate key hash file contains a localized set of values to use for assignments. You need to keep these values in sync with the actual target table. Imagine if you manually loaded more rows into the target table, now your hash file has no awareness of the current maximum value assigned in the ...

If you are running the job from Director or the Designer debugger, the invocation dialog box has a tab that allows you to set the limit of rows. If you are using custom job control, there is an API to call after you have attached a job handle. Look in your DS BASIC manual, I think it's called DSSetJ...

If you read the disk caching guide under your Start menu, you'll see that the only size limitation is 1/2 a terabyte, and that the guide actually has a tuning table where up to 2 gigabyte hash files are documented. In addition, it doesn't talk anywhere about a user limitation, in fact, it describes ...

Ray said it better than I have been able to say. If your requirement is to put 30 million rows into a hash file for repetitive lookup, you will probably exceed the read caching capabilities. My technique suggests partitioning the data so that you have a subset in N hash files, where the size of the ...

Are you guys talking about a Control job calling these 4-5 identical jobs (with different selection criteria) or executing them as 5 separate jobs from director? In fact does it make a difference? You can either make physical clones, or use job instances. The affect is the same. An instance is just...

In your suggestions you would have the input file read by a transformer and have 'mypartitionnumber' passes as a parameter, is that correct? Yes. One master job, with N instantiated clones. The job uses a partitioning algorithm to only get a portion of the source data. My question is this....we hav...