On Jul 3, 2016, at 6:58 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> It looks to me like set_append_rel_pathlist builds a partial path for
> an appendrel only if there exist partial paths for all members of the
> appendrel. I don't understand this restriction. Why could you not
> build it from a regular path for each appendrel child? You'd have
> to farm out entire childrels to workers, which might be larger chunks
> of work than one could wish, but it seems better than failing to
> parallelize at all.
Absolutely. The problem is merely that we need some executor support that has not been written yet. See discussions of
ParallelAppend in the archives.
...Robert