[Tarantool-patches] [PATCH v4 2/6] sql: refactor create_table_def and parse

[Roman Khabibov]

Hi, Vlad and Nikita!

> On Sep 18, 2020, at 15:31, Nikita Pettik <korablev at tarantool.org> wrote:
> 
> On 17 Sep 16:43, Vladislav Shpilevoy wrote:
>> On 16.09.2020 15:27, Nikita Pettik wrote:
>>> On 12 Sep 00:51, Roman Khabibov wrote:
>>>> Move ck, fk constraint lists and autoincrement info from
>>>> struct create_table_def to struct Parse to make the code more
>>>> reusable when implementing <ALTER TABLE ADD COLUMN>.
>>>> 
>>>> Needed for #3075
>>>> ---
>>> 
>>> Does ANSI allow to include constraints in ADD COLUMN statement?
>> 
>> Yes.
>> 
>>> Why did you decide to add autoincrement to this list? Was there any
>>> discussion on this subj?
>> 
>> There is nothing to discuss. It is either a part of the standard or
>> not. Romain said it is (I didn't check though).
> 
> Unfortunatelly, this is false statement - autoincrement is not a part of
> standard (I've checked 2011 edition).

No no, Vlad. Maybe it was misunderstanding, I didn’t say that autoinc
is a part of the Standard. My point was the following:

The standard says that the column descriptions in <CREATE> and <ALTER>
are exactly the same. All consrtaints, <DEFAULT> and <COLLATE> are allowed.
But <AUTOINCREMENT> is not described in the standard. I suggested (Vlad
didn't mind) to continue this logic and support <AUTOINCREMENT> inside
<ALTER>, because there are no significant restrictions on this. Why can’t
we add a sequence to an already existing table?