DSXchange

I doubt if it is possible. The XML output of a job is very complex with lots of repeating groups. It does sound reasonable though. I would not trust it knowing how complex this data structure is. I would think it would be easier to edit the repository directly. There maybe cases where your idea may ...

We never assume a hashed file retains data. We always clear or delete and load the hashed file every ETL run. If you have large amounts of data in your hashed files then rethink how you are using them.

I would bet that your columns should be nvarchar.

Sorry, but it should not happen. Got to be something we are missing.

Nope.

Try them both and let us know.

Do a search on EtlStats.

Right.

If you want to go against repository tables then do a search on DS_JOBOBJECTS. This is where the column names are stored for each link. The record structure of this hashed file or table varies by the OLETYPE field. This field represents the stage type and or link type as to direction either input or...

You are correct, all of these are in there. It seems to me there used to be missing but I just checked.

You are correct, all of these are in there. It seems to me there used to be missing but I just checked.

Does not make sense. An update only should fail or specific records. These should not work the same. The order is important on upserts whether update and then insert or insert then update based on which is more common.

Does this table have a primary key?

Are you getting blocked? Look at your database lock table. Make sure you have nolock on your source queries.

I forgot left(), right() and trim(). Trim() can delete any character not just spaces.

The server transform has substrings(). You can also use the form XYZ[StartPos, EndPos]. You can use convert(), change() or ereplace() to scrub this field. There lots of functions to do what you need.

Even PX has lots of ways to scrub this.