|
According to
Slow Query Using Pipeline Function TABLE with Plan 'COLLECTION ITERATOR PICKLER FETCH' (Doc ID 2224317.1)
"Statement is using pipeline function TABLE(). Statistics are not considered when using pipeline function."
And the note suggests using the cardinality hint to tell CBO the number of rows in the table, or raising dynamic sampling level, or using extensible optimizer.
Also, Jonathan Lewis's article at
https://jonathanlewis.wordpress.com/2010/11/30/collection-costs/
is useful. Read from "Here’s a suggestion for use with the table() operator" to the end.
|
|