Re: Review: GIN non-intrusive vacuum of posting tree

2017-03-16 21:27 GMT+05:00 David Steele <david(at)pgmasters(dot)net>:
> This patch applies cleanly and compiles at cccbdde.
>
> Jeff, any thoughts on Andrew's responses?

Hi, David!

I've got some updates on the matter of this patch, since the
understanding of the B-tree bothered me much.
Currently, I'm at PgConf.Russia, where I've contacted Theodor Sigaev,
and he answered my questions about the GIN.
0. I think that proposed patch is safe (deadlock free, does not
introduce new livelocks, all the resources guarded properly)
1. There _are_ high keys at the posting trees, they are just called
rightmost keys, but in fact they are high keys in terms of L&Y
algorithm.
2. Thus, L&S fully concurrent vacuum is possible, indeed, and
furthermore Theodor suggested that I should implement not only page
eviction, but also page merge and tree condence algorithm.
3. Eventually, I'll do that, certainly, but, currently, I can't
predict the time it'll take. I think I'll start somewhere in the
summer, may be right after GiST intrapage indexing.

As for now, I think that having this patch in PostgreSQL 10 is viable.

Best regards, Andrey Borodin.