Loic Dachary wrote:
>
> Sergei Golubchik writes:
> >
> > Ok, I cannot say that much. I didn't ever look at INSERT DELAYED before,
> > but when I did, after your mail, looks that it's ideally suited for grouping
> > fulltext index updates (it's better then hooking on selects).
>
> Glad that you see it as a good idea.
>
> > But delayed_insert::handle_inserts calls mi_write to modify table files.
> > So there should be some way to tell mi_write not to create keys for
> > row data (probably, via mi_extra() call ?).
>
> mi_extra seems exactly fit for that purpose thanks.
>
> > And how will you tell find_all_keys which are "previously inserted records" ?
>
> The idea is to write another function that behaves mostly like find_all_keys
> except it gets rows from the delayed insert cache instead. This function would
> be called *after* the rows are inserted.
>
> > > It occurs to me that we might as well do the same for all the
> indexes
> > > types, no need to be selfish :-)
> >
> > Yes, but it's most important for fulltext keys, as quite a few rows can
> > produce a lot of fulltext index entries.
>
> Yes, anyway I'll only focus on fulltext keys for now, no need to accumulate
> problems.
>
> Do you think it's a good idea to cc: this discussion to
> internals@stripped or will this be considered noise ?
Yes, I think it is a good idea - this is exactly what internals list is for!
--
MySQL Development Team
__ ___ ___ ____ __
/ |/ /_ __/ __/ __ \/ / Sasha Pachev <sasha@stripped>
/ /|_/ / // /\ \/ /_/ / /__ MySQL AB, http://www.mysql.com/
/_/ /_/\_, /___/\___\_\___/ Provo, Utah, USA
<___/
