Hi Tom,

in our ETL tool we have a process by which a flat file is read and transformed. A table is created with columns to hold the flat file content and columns for the functions that are applied sequentially per row. the file content is loaded via bulk copy. After the bulk copy a series of update statements are applied into the function columns, again applied sequentially because updates are dependent on previous updates in many cases. All these tables have PCTFREE = 10.

The total count of columns is over 255 in all these examples and i was expecting all rows to be chained because of this. However in one case i have ~5% chaining, in others i have 99% chaining.

Notes
These are edge cases..normally files are not so wide. We are looking at separating the tables.
We cannot alter the order of the columns
We cannot drop any columns

Examples

Flat File Col Count Function Col Count % Chain
261 116 7.5%
252 73 99%

So some questions:
1. Can the diff in chaining be explained by which columns are being referred to in the update statements? The whole splitting of the table (in memory terms) from the right-hand side perhaps
2. Can PCTFREE settings have any positive impact on chaining with wide tables?
3. Are there any other tactics you can suggest?