Re: PATCH: generate fractional cheapest paths in generate_orderedappend_path

Hi,

On 12/2/21 15:58, Arne Roland wrote:
> Afaiac we should add a simple testcase here, like I suggested in
> 477344d5f17c4a8e95d3a5bb6642718a
> <https://www.postgresql.org/message-id/477344d5f17c4a8e95d3a5bb6642718a%40index.de>.
> Apart from that I am not sure there is work to be done here.
>

Well, I mentioned three open questions in my first message, and I don't
think we've really addressed them yet. We've agreed we don't need to add
the incremental sort here, but that leaves

1) If get_cheapest_fractional_path_for_pathkeys returns NULL, should we
default to cheapest_startup or cheapest_total?

2) Should get_cheapest_fractional_path_for_pathkeys worry about
require_parallel_safe? I think yes, but we need to update the patch.

I'll take a closer look next week, once I get home from NYC, and I'll
submit an improved version for the January CF.

regards

--
Tomas Vondra
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company