Re: [HACKERS] VACUUM in SP-GiST

2012-03-11 Thread Tom Lane
y...@mwd.biglobe.ne.jp (YAMAMOTO Takashi) writes: >> [ tgl's redesign of spgist vacuum algorithm ] > with a concurrent split moving leaf tuples between pages, can't it make > bulkdelete fail reporting some TIDs Yeah, you're right :-(. I've been chewing on this problem for awhile and I think I se

Re: [HACKERS] VACUUM in SP-GiST

2012-01-17 Thread YAMAMOTO Takashi
hi, > I started reading the spgist vacuum code last night, and didn't like it > at all. I found a number of smaller issues, but it seems to me that > the design is just fundamentally wrong. Unless I'm misunderstanding it, > the approach is to recursively traverse the tree in sort of the same way