[tarantool-patches] Re: [PATCH 1/3] sql: EXPLAIN through net.box leads to SEGFAULT

[n.pettik]

That’s why I don’t like any verbal agreements…

My point is if result-set features column names,
then those columns must have type. I fail to come up with
example where resulting set features column names,
but columns don’t have types.
On the other hand, mb it is worth to send type ANY instead of
UNKNOWN, it makes sense to a greater or lesser extent.

> On 22 Nov 2018, at 23:55, Мерген Имеев <imeevma at tarantool.org> wrote:
> 
> After discussion we decided that this patch will serve us
> part of patch-set that should close newly filed bug:
> https://github.com/tarantool/tarantool/issues/3832
> 
> The main Idea behind this ticket is that any statement
> should return column type for when it returns column name.
> And this ticket allow us to find an error without falling
> into SEGMENTATION FAULT or returning less metadata.
> 
> Четверг, 22 ноября 2018, 21:31 +03:00 от Vladislav Shpilevoy <v.shpilevoy at tarantool.org>:
> 
> It was just a question, "why did you decide to do not
> send type". It was not necessary to start sending UNKNOWN. I
> think, that unknown type should be sent never.