maria-developers team mailing list archive

[Alexi1952 <Alexi1952@xxxxxxxxx>]

24.09.09, 12:04, "Kristian Nielsen" <knielsen@xxxxxxxxxxxxxxx>:

> Alexi1952  writes:
> > Agree. BTW  tables_ok() is just the only member I had already #ifdef'ed out from Rpl_filter for client context.
> Ah, I see.
> > As for your suggestion to have a separate class, is it OK to do something like this?
> >
> > class Binlog_filter
> > {
> >   < ... all members from Rpl_filter except for tables_ok()
> >      ... (will also check carefully for other members) ...>
> > };
> >
> > class Rpl_filter: public Binlog_filter
> > {
> >   <... tables_ok() ...>
> > };
> Yes, that sounds good.
> >
> > BTW in this case declaring
> >
> >     Binlog_filter* binlog_filter;
> >
> > will look like more natural than
> >
> >     Rpl_filter* binlog_filter;
> >
> > (why indeed *replication filter* in mysqlbinlog which actully *doesn't replicate* :)
> Indeed :-)

Sorry. It's a bit hasty decision. For WL#40 we have to have a modification of
tables_ok(TABLE_LIST*) to support table-rules. So it should be something
like this:

class A_filter  /* TODO: choose more appropriate name*/
{
  < ... all members from Rpl_filter except for tables_ok() ...>
};

class Binlog_filter: public A_filter
{
  <... tables_ok() for client
        with appropriate argument instead of TABLE_LIST ..>
};

class Rpl_filter: public A_filter
{
  <... tables_ok() for replication...>
};

Note. This is also preliminary because curently I have no final/clear decision how to do WL40 for SBR.
(I'm itching to detach the parser but this is too huge task within these binlog WL's)


> _______________________________________________
> Mailing list: https://launchpad.net/~maria-developers
> Post to     : maria-developers@xxxxxxxxxxxxxxxxxxx
> Unsubscribe : https://launchpad.net/~maria-developers
> More help   : https://help.launchpad.net/ListHelp

-- 
Новая Яндекс.Почта http://mail.yandex.ru/promo/new/sign