Hi!
>>>>> "Sergei" == Sergei Golubchik <serg@stripped> writes:
<cut>
>> For example, MySQL may first do an ORDER by optimization and use a
>> bitmap that only fetches the row id + what is in the ORDER BY part and
>> then later it may fetch the whole rows.
Sergei> Monty, you've missed the point. I'm talking about bitmap changes in the
Sergei> middle of the scan. Like
Sergei> index_init
Sergei> index_read
Sergei> index_read
Sergei> ... bitmap change ...
Sergei> index_read
Sergei> index_read
Sergei> for the ORDER BY it's one bitmap for the first scan, another bitmap for
Sergei> the second. No problem here.
Agree, the above bitmap change shouldn't happen; Sorry for the
misunderstanding.
<cut>
Sergei> When a scan starts NDB sends the bitmap to the nodes and they start
Sergei> sending the data back. If bitmap changes, it has to abort the current
Sergei> scan (I presume) send new bitmap to the nodes and re-retrieve rows that
Sergei> were already sent under the old bitmap but not yet consumed by the
Sergei> server. You're right, it's not very hard. But it's expensive.
Thanks for the explanation. Agree that this is something that the
engine shouldn't have to do.
Regards,
Monty
