DSXchange

I have a job which is reading a MQ source which is generating waves. For reason I won't delve into, I would like to write this data to a sequential file and later read this file while keeping the wave related information. I would like to be able to write an "end-of-wave" record to the sequ...

What method are you using to load to Oracle? See about setting "APT_ORACLE_PRESERVE_BLANKS" to change the load behaviour on spaces.

You can generate a sequence from a file or from a database sequence in the transform stage. In addition, there is a whole stage dedicated to this called the "Surrogate Key Generator" stage.

Is your key order indeed CLASS_TYPE_CDE then CLASS_CDE everywhere in the job? Typically when a lookup like this fails it is due to some key issue.

miwinter wrote:...Is this a reference to using the 'Position' attribute?

Yes. Key should be 0.
Could you do a "LIST DICT <yourfilename>" from a DS shell and post the I-Type listing for the combined key?

I still think your root cause is either the compound key or the column ordering. Do you see column numbers?

Assuming you are keeping string format: This assume the year 10 as the century break year.

Can you order the keys before the data columns

Are you using a single key or compound key? Also, have you kept the same column names and ordering throughout the job?

Is your output column to the sequential defined as Decimal(10,2) or as a string? How about doing an explicit conversion in your transform stage using DecimaltoString() and writing a varchar to the sequential file (assuming you are currently writing a decimal field)? That explicit conversion is bette...

That could explain it - I know the code contains a plethora of ifdefs depending upon OS when it cmoes to internal dates and times.

The lookup stage should work correctly, regardless of the source data partitioning. The join does depend upon correct partitioning; you should partition on your join key to ensure that all rows are in the correct partition.

It would seem that the library should work. The "file" command only checks the initial bytes, so it could be corrupted - yet the "ldd" command doesn't show an error. Could it be a 64bit library in the 32bit directory or is the date-time modifed different from the other objects in...

If you do a "view data" on the source, are the numeric values displayed correctly there?

Are you using custom SQL , and if so, please post it. If you are getting duplicate "key" columns, then most likely your table DDL is not correct. You might look at that and/post it here as well.