[tarantool-patches] Re: [PATCH 2/2] sql: transactional DDL

[n.pettik]

> On 24 Jul 2019, at 23:58, Vladislav Shpilevoy <v.shpilevoy at tarantool.org> wrote:
> On 24/07/2019 15:23, n.pettik wrote:
>> 
>> Thanks, the only dubious nit I can spot: should we
>> set initiateTTrans for DDL operations which consist
>> of one box operations?
> 
> Yeah, I thought about it, and even made it and reverted back
> several times having some doubts 'code consistency vs
> transaction necessity'.
> 
> After all I decided, that each DDL op should be transactional,
> even if consists of one statement. Because 1) it is consistent
> with other SQL DDL, 2) it protects us from the case when SQL
> DDL will complicate. I think, the latter matters the most, taking
> into account incoming information schema, when even a simple
> index creation will insert into _index, into something like
> _constraint_name to ensure name uniqueness, etc.

Ok, I was driven by the same thoughts.

Thanks, now LGTM.