DSXchange

Yes... yes it is. D'oh. We're not talking about the aggregator?

How did you import the metadata, using which tool/option? Do you have NLS enabled? I've seen this before and I'm sure it has been posted here as well but I don't recall off the top of my head why that happens but I'm sure it's NLS related. These 'certain Oracle instances' - how does their NLS_LANG s...

It does (according to the documentation) convert everything to a double under the covers by default. There is a optional property in there somewhere to "retain type" from what I recall, that should have helped here as well.

By 'known relational database' I meant Oracle, SQL Server, etc. In that case use the corresponding 'native' connector.

Odd... I don't recall anyone posting that issue before. Usually it's what we've been discussing - it 'works' without the variable but returns the pattern as the filename. Add the variable and you get the full filename, easy peasy. With that behaviour I'd contact your official support provider, ask t...

Oh heck, I thought we were talking about the trailing semi-colon, not the one in your data... sorry. What happens if you escape it? Definitely involve your support provider as it certainly does seem like bad behaviour and perhaps there is a fix/patch for it out there.

And as noted and you've asked, if it is a known relational database with an associated Connector, you should be able to use that as well.

It's more of a DataStage thing it would seem. Found out long ago that (for Oracle in my case) it didn't want the trailing semi-colon that Toad would add and didn't mind if it was there. People would "user define" something that wouldn't run simply because of the semi-colon. I don't see the...

Seems like the blurb in the second italicized statement is the issue, yes?

What happens when you leave off the semi-colon?

Add the environment variable.

Of course, there's always the Expand stage. :wink: A clarification, Mr. Wurlod. The docs say "The Expand stage uses the UNIX uncompress or GZIP utility to expand a data set." - is that a generic usage of the term "data set"? For some reason I thought it literally needed to be a c...

That sounds right... but again, long time.

Couple of questions. I'm assuming this is a multi-node job and that running it on a single node "solves" the problem. If true then you have a partitioning problem. If you haven't, try setting it to Hash and use the appropriate keys for the target in the hashing algorithm.

Look into the gunzip -c option which unzips the file to std out which the the Sequential File stage can leverage via the Filter option. If you really need to unzip to files, pipe it to the split -b command which lets you control the size in bytes of each chunk. -b n: Split a file into pieces n bytes...