[tarantool-patches] Re: [PATCH 2/2] sql: add test on changes/total_changes

n.pettik korablev at tarantool.org
Thu Oct 4 13:49:50 MSK 2018

Accidentally, I sent last mail from wrong address, so it didn’t arrive to mailing list.
Original letter below:

> On 2 Oct 2018, at 16:40, Никита Петтик <kitnerh at gmail.com> wrote:
>>>> I have expected the same thing.
>>>> However:
>>>> 1. sqlite works in the same way.
>>> It doesn’t mean anything. SQLite features a wide range of really strange things.
>>>> 2. implementing this just need adding (nChanges) to savepoint
>>> Seems so, but it shouldn’t be complicated.
>> I suppose it is not complicated.
>>>> 3. I suppose that we should not implement this.
>>> Why? I guess it is matter of discussion.
>> I think that it is not important, because it is relatively useless counter.
>> 1. No one need to count exact number of changes.
> Without proofs such statements worth nothing.
>> 2. Changing the api require solid understanding of what total_changes is
>>    and why we need to change this.
>>    It strange for me that it do not do rollback on rollback, however, it meant to work this
>>    way and it also makes sense.
> It only means that SQLite developers decided to do so, nothing else.
>>> Anyway, I propose to decide what to do with TOTAL_CHANGES within this patch,
>>> since original issue sounds like:
>>> "Fix if easy, remove otherwise.”
>>> Adding initially wrong tests is likely to be huge mistake.
>> This thing works. Tests proof that.
> ‘Works' and ‘works and gives wrong result' are sort of different things.
> In this regard, rollback doesn’t work, ergo this counter works well
> only in half cases.
>> I believe that those tests are the artefact of this issue.
>> If you want change something, you need to file an issue and explain,
>> why this change is important.
> I guess there is no matter of discussion: it is very likely to be obvious
> that rollback should return old total_change counter (i.e. value before rollback).
> If you strictly disagree, you are welcome to discuss it with other members.

More information about the Tarantool-patches mailing list