On 24.05.2011 13:28, Roy Lyseng wrote:
> Hi Martin,
>
> On 24.05.11 13.02, Martin Hansson wrote:
>> Roy Lyseng skrev 2011-05-24 11.09:
>>> On 24.05.11 10.40, Jorgen Loland wrote:
>>>> Martin Hansson wrote:
>>>>> #At file:///data0/martin/bzrroot/wl5855/mysql-trunk-prototype/
>>>>> based on
>>>>> revid:azundris@stripped
>>>>>
>>>>> 3368 Martin Hansson 2011-05-23
>>>>> Replaced the st_join_table typedef with a C++ class declaration.
>>>>
>>>> +1!
>> Hehe, this was actually not intended for anyone to see. I just needed
>> to do this
>> in order to comfortably start prototyping wl#5855. But apparently, it
>> is in
>> demand. May I push?
>>>>
>>>> Feel free to encapsulate as well :)
>> I have to leave some fun for the rest of you lot, don't I?
>>
>>>>
>>>>
>>>
>>> JOIN_TAB is not in compliance with our class naming rules.
>> True. But regardless of what other name I might pick it would cause
>> controversy,
>> not to mention all documentation that would have to be rewritten, or
>> the merge
>> conflicts. At least it is consistent with class JOIN, and a step in
>> the right
>> direction IMHO. g
> Given that Evgeny is planning a major refactoring around JOIN and
> JOIN_TAB, I'd say it is better to wait until those plans are finalized.
Is there a WL or description for this refactoring?
The main integration point for Clusters 'pushed join' feature is the
JOIN_TAB structure, and we would rather not
see a refactoring in this area which changes a lot in how we integrate
with the optimizer.
Ole John
>
> Thanks,
> Roy
>
