Re: Reworks of CustomScan serialization/deserialization

On Wed, Mar 23, 2016 at 1:36 PM, Petr Jelinek <petr(at)2ndquadrant(dot)com> wrote:
> Ok, I am happy with it, marked it as ready for committer (it was marked as
> committed although it wasn't committed).

Thanks for fixing the status. I had forgotten about this thread.

I can't really endorse the naming conventions here. I mean, we've got
the main extensible nodes stuff in extensible.h, and then we've got
this stuff in custom_node.h (BTW, there is a leftover reference to
custom-node.h). There's no hint in the naming that this relates to
scans, and why is it extensible in one place and custom in another?

I'm not quite sure how to clean this up. At a minimum, I think we
should standardize on "custom_scan.h" instead of "custom_node.h". I
think that would be clearer. But I'm wondering if we should bite the
bullet and rename everything from "custom" to "extensible" and declare
it all in "extensible.h".

src/backend/nodes/custom_node.c:45: indent with spaces.
+ }

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