DSXchange

You might want to post some of these 'weird characters'. You might be seeing zoned decimal or something of that ilk.

Best answer is Ray's answer - get the provider of the file to cut you a break and build you a straight non-funky ascii file.

Sure. Don't exceed your limits, or increase them if they are being set for you. Out of curiosity, what hardware / OS is your server running on?

Or write smaller files. Can you break up your process so it writes into multiple (smaller) files rather than one large file?

I don't recall having to do that.

Or possibly exceeded your ulimit.

If you want to delete single entries, you have to turn that option on as it is disabled by default.

If you literally want to delete 'all entry from version control', drop and recreate the appropriate VERSION project.

Cool. Are you using ODBC? From what I recall it doesn't do deletes for some reason. You never said what database you were dealing with, but most of the native stages (like OCI for example) have that option from what I recall. The 'downside' is the fact that you'll have no idea how many rows were act...

I understand. But they intend to switch it back and forth between Oracle and SQL Server, so the solution needs to be applicable to both.

Makes it kind of hard to provide good advice when you leave out details like that, Andal. Don't worry, people here can read 'complex' sql. My only point was using user defined sql in a DRS stage removes any benefit of the 'D'ynamic part of it, unless you can stick with extremely generic sql. At wors...

Something like Parameter Manager from the fine folks who run this site could easily get your parameter into all of the jobs. However, if they all need different default values, you're still stuck with the monkey work of changing them all one by one.

Arnd - does the PX transformer not have the 'Find and Replace' functionality of the Server transformer?

Not a cut-and-paste solution, per se, but you can use regular expressions of a sort to do all of the changes in one swell foop.

Or there's always the export-edit-and-reload approach.

Why does this need to be 'user defined' sql? Let the stage generate it.

ray.wurlod wrote:A search on this Forum is likely to prove productive.

Especially if you Search by Author when looking for pointers on which manuals to read. I find that 'ray*' works best for me, then I don't have to remember how to spell 'Wurlod' or if there's a dot in it or not.

There's always more than one way to solve problems here. If you do take this route and 'latest record' means 'max' date or something to that effect, make sure you sort the file first if there is any doubt about the order they will come in. Otherwise you will literally get the last record for that so...

Yup, that would be a simple approach that leverages the 'destructive overwrite' of hash files - last one in wins for each combination of the key fields. It would require that your data be sorted properly. Another option would be to sort the file descending on the key fields and then take the first r...

You could create a DataStage job to do this. Or three. Me, I'd stick to one. What you need is a transformer and your RDBMS stage of choice, one link for each of these processes. Define a stage variable in the transformer so the compiler doesn't complain, but you won't have to use it. Have your thre...