DSXchange

You get this error when you perform a SELECT, UPDATE, or DELETE with a clause such as WHERE ROWID = '....' , and the ROWID listed not only does not exist in the table, it could NEVER exist in the table. A ROWID is a combination of the Object ID, File Number, Block Number (within the file), and Row N...

The OVER clause is attached to the COALESCE() function, but COALESCE is not an analytic function.

I think you want the OVER clause attached to the LEAD() function. I'm surprised this worked in your SQL editor.

What is the UPDATE ACTION on the OCI stage? If it is "Insert New Rows Only" - no problem. If it is "Insert new rows or Update existing" or "Update existing rows or insert new" then you will have terrible performance problems - you need to split your inserts and updates ...

By "flat files", so you mean the Zip Files? If the Zip files are on Unix, try to source a WinZip compatible version of Zip compatible for your flavour of Unix - use Google.

Long story, but the architecture of my current DW is such that the entire Staging layer is in Oracle Externally Organised Tables, and the Enterprise (3NF) layer is in regular Oracle tables. Instead of doing lookups with Hashed Files, this enabled us to use SQL joins on Oracle. Of course it would the...

Ray once posted the URL for the UniVerse Administrator's guide here . As for a specific DS Administrator guide - I once asked Ascential the same thing - there was nothing beyond the supplied manauls on your install CD. There's also no Ascential training course that I could track down, although it is...

Mmmm, yes . I really should read through to the end of the OP's post. "client tools are on Windows" is a bit inconclusive, but I think you're right to bet that the OP wants to wait on a remote file.

This query has a certain academic interest, but surely coding your own lookups in BASIC defeats the purpose of spending AUD$100K on an ETL tool.

Apols to Ray if he covered it lower in his reply, but you can get Zip and Unzip for Unix. I have used them on both Tru64 and Linux, I'd be surprised if they're not available for (some) other platforms as well - google it.

Question for the gurus: the code looks OK, so could something have happened to the underlying hashed file?

As a test (without destroying your keys file), you could create copies of the KeyMgt routines, edit them to change the underlying file name, and then try your test using the new routines.

Make sure that the LC Inputs and Outputs have the same KEY columns checked. I've had similar probs when they are not matched.

Unless you are using Oracle v7.3, leave the indexes where they are and use SQL*Loader in Direct Path mode. Index maintenance is deferred until after the load. This is usually much better than dropping them or marking them as unusable. The exception is when the loaded data is large compared to the ex...

An alternate idea (I've never tried it, so it may not work): Say you had a Hashed file containing the calendar with no missing dates. Take your source file and write it to either another Hashed File or a UV stage (UniVerse table). Then you should be able to use the UV stage to perform a SQL join on ...

Apols. I read your earlier report of 14000/s as being a simple read/write from flat file to benchmark your disk - not as a throughput figure on the job in question. Re the quoted figure of 1000/s using sqlldr: are you running sqlldr across the network? Or locally on the Oracle server after an FTP? I...

So Auto mode (with IPC) writes at 600/s, and Manual at 1000/s. This tells me that the maximum throughput of the transformation is 1000/s, but that the maximum write-rate slows that down to 600/s. Since 1000/s writes to a local flat file, that eliminates Oracle and the network. Your initial post ment...