Re: Parallel INSERT (INTO ... SELECT ...)

Thanks for the feedback.
Posting an updated set of patches. Changes are based on feedback, as
detailed below:

[Hou]
- Deallocate list returned from RelationGetIndexList() using
list_free() after finished using list
- Regard foreign and temporary tables as parallel-restricted (instead
of parallel unsafe) for Insert
- pfree() conbin returned from TextDatumGetCString() after finished using it
- Make parallel-safety checks of index expressions more efficient,
remove some redundant function calls
- Add a few more test cases to cover certain untested parallel-safety
check cases
- Remove repeated table_close() on return, by moving table_open() &
table_close() to a higher level
- Reduce Insert parallel-safety checks required for some SQL, by
noting that the subquery must operate on a relation (check for
RTE_RELATION in subquery range-table)

[Zhihong Yu]
- Minor change to patch comment
- Wrap long line
- Remove intermediate local variable

[Tsunakawa-san]
- Update RELATION_IS_LOCAL macro to reinstate previously-removed check
on the relation being newly created in the current transaction (and so
assumed accessible only to the current backend), but for
non-parallel-mode only (since now it may be accessible to parallel
workers)
- Remove braces for one-line else
- Fix code comment
- Rename PrepareParallelMode() for plan execution, so that it's not
misinterpreted as a general function for preparation of parallel-mode

[Misc]
- Fix bug in query re-writer - hasModifyingCTE is not set in
re-written non-SELECT queries having a CTE

Regards,
Greg Nancarrow
Fujitsu Australia