Re: Multi-level hierarchy with parallel append can lead to an extra subplan.

On Tue, Jan 9, 2018 at 5:41 AM, Amit Khandekar <amitdkhan(dot)pg(at)gmail(dot)com> wrote:
> This subplan should not be there. It looks like that's because when
> we add up subpaths while preparing Parallel Append path containing mix of
> partial and non-partial child paths, in accumulate_append_subpath(),
> the subpath being added is an Append path because it's a multi-level
> partition. So in the condition "else if (special_subpaths != NULL)",
> both *subpaths and *special_paths get updated. But then it is not
> returning from there when it should. Instead, the control passes to
> the end of function where *subpaths is again populated :
> *subpaths = lappend(*subpaths, path);
> which leads to an extra child with partial subpath getting added.

Boy, what a dumb mistake on my part. Thanks for finding the bug,
Rajkumar. Thanks for analyzing it, Amit. I have committed the fix.

--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company