On 12/16/2010 10:47 AM, Daogang Qu wrote:
> 2010-12-16 18:12, Alfranio Correia wrote:
>> Hi Daogang,
>>
>>
>> On 12/16/2010 08:55 AM, Daogang Qu wrote:
>>> Hi Alfranio,
>>> Thanks for your comment. See reply in-line.
>>> Please review:
>>> http://lists.mysql.com/commits/127027
>>>
>>> Best Regards,
>>>
>>> Daogang
>>>
>>> 2010-12-11 05:05, Alfranio Correia wrote:
>>>> Hi Daogang,
>>>>
>>>>
>>>> Great work,
>>>>
>>>>
>>>> STATUS
>>>> ------
>>>>
>>>> Conditionally approved.
>>>>
>>>> REQUEST
>>>> ------
>>>>
>>>> 1 - We are not writing directly to the index_file so I think
>>>> we can replace the init_io_cache by a simple seek.
>>> We are using init_io_cache for efficiently reading from index file
>>> instead of writing.
>>
>> You don't need to call init_io_cache everytime you need to read the
>> index file. You just need to call it once, after opening the file.
>>
>> I think you should replace other calls to init_io_cache by a seek,
>> this would me more efficient as these calls to init_io_cache were
>> used because after the open there was a init_io_cache with write
>> properties.
> But for every write and purge to the index file, the index file will be
> closed and deleted, and then renamed from crash_safe_index_file.
> So we need init_io_cache in the process. No chance for seek.
>
I still don't agree with that.
Notice that you open the index with init_io_cache(...READ...) and
calling init again is only need if you had openned it with WRITE,
otherwise a seek is enough.
I will check this next week.
Cheers.
