DSXchange

If your target is a Sequential File, change the definition of NULL in the Format tab. The default is to replace NULL with "".

Either sort data to be loaded into the hashed file in reverse order, as Craig suggested, or de-duplicate the data by other means before loading them into the hashed file. All writes to hashed files via the Hashed File stage are destructive overwrites. If you use a UV stage to insert rows you will ac...

This error can also occur if the file is fixed-width format and you either neglect to check the Fixed Width check box on the format tab or get the display widths (= column widths) wrong in the Columns grid.

I don't have an answer to your particular question. However, why not post on DeveloperNet - there are vendor experts who are permitted to respond there.

Yeah. Weird. Nothing overwriting it subsequently? Do let us know what "they" have to say.

In the Parallel Job Developer's Guide is specifically states (page 29-3) that the second argument with LIKE must be a regular expression (that is, a UNIX regular expression). It also states that IS NULL and IS NOT NULL are supported as operators.

There is a Chapter on the Modify stage in Parallel Job Developer's Guide. One of the things it describes is how to change the nullability of a field. It also describes the available null-handling functions for that stage type.

Do you have a limit on a link handling rejected rows?

In Director, in Tools > Options (Limits tab), what is the default limit?

Look in the folder RT_BP2038 within your project (on the server) for a file called JOB.1364871999.DT.1368147597.TRANS2. This is the routine created by compiling the Transformer stage. Look around line 444 and tell us what's happening there. Please also post your job design; what stage types and how ...

It won't work in the next ("Hawk") release.

ODBC or OLEDB is the most effective way to work with Excel. No-one seems to have implemented OLEDB on UNIX, so you're stuck with ODBC.

The ODBC driver for Excel presents a worksheet as a system table.

If you are really struggling with this, why not hire the services of an expert in the field? For example, MetaView360 (one of the supporters of this site) offer a free HIPAA data quality assessmentand lots of other (meta)data services. Maybe just a one-day engagement would do it for you.

If using NLS, try using a map of NONE to see what happens. Otherwise there seems to be a difficulty in handling NChar or NVarChar; try using Char or VarChar instead (and, perhaps, using CAST in your SELECT statement (column definitions)).

You missed my point. Have the DBA log in as a Directory Administrator and create the directory. The software looks after table structures and so on. If the DBA insists on knowing the metadata in advance, you can find the scripts (I've forgotten where exactly). Look for files with a suffix of ".sql".

Maybe just luck. Data in hashed files are stored a characters, so (for example), an Integer may be anywhere from 0 to 10 characters. Further, each data record is padded to a whole "word" boundary, where a "word" is either a four-byte word (32-bit addressing) or an eight-byte word (64-bit addressing).