Thread: What is an 'unused item pointer'
I can't seem to find info about this in the docs, or on the web. -- Jim C. Nasby, Sr. Engineering Consultant jnasby@pervasive.com Pervasive Software http://pervasive.com work: 512-231-6117 vcard: http://jim.nasby.net/pervasive.vcf cell: 512-569-9461
"Jim C. Nasby" <jnasby@pervasive.com> writes: > I can't seem to find info about this in the docs, or on the web. http://developer.postgresql.org/docs/postgres/storage-page-layout.html Currently, when a tuple is reclaimed by VACUUM, we just mark its item pointer as unused (and hence recyclable). I think it might be safe to decrease pd_lower if there are unused pointers at the end of the page's pointer array, but we don't currently do that. In any case there could still be unused pointers embedded within the array. regards, tom lane
On Sat, Sep 24, 2005 at 07:19:10PM -0400, Tom Lane wrote: > "Jim C. Nasby" <jnasby@pervasive.com> writes: > > I can't seem to find info about this in the docs, or on the web. > > http://developer.postgresql.org/docs/postgres/storage-page-layout.html > > Currently, when a tuple is reclaimed by VACUUM, we just mark its item > pointer as unused (and hence recyclable). I think it might be safe to > decrease pd_lower if there are unused pointers at the end of the page's > pointer array, but we don't currently do that. In any case there could > still be unused pointers embedded within the array. I suspect that if the free space in pages isn't collapsed during vacuum then it probably doesn't matter, but I don't know if that's the case. If it is collapsed then my intuition is that tables that are fairly narrow would benefit from adjusting pd_lower because it wouldn't take too many item pointers to equate to a tuple. This would probably be even more pronounced if the tuples were generally small but could vary to a larger size. Sounds like a good newbie TODO? -- Jim C. Nasby, Sr. Engineering Consultant jnasby@pervasive.com Pervasive Software http://pervasive.com work: 512-231-6117 vcard: http://jim.nasby.net/pervasive.vcf cell: 512-569-9461
"Jim C. Nasby" <jnasby@pervasive.com> writes:
> On Sat, Sep 24, 2005 at 07:19:10PM -0400, Tom Lane wrote:
>> Currently, when a tuple is reclaimed by VACUUM, we just mark its item
>> pointer as unused (and hence recyclable). I think it might be safe to
>> decrease pd_lower if there are unused pointers at the end of the page's
>> pointer array, but we don't currently do that.
> Sounds like a good newbie TODO?
Uh, no, because the $64 question is whether it actually *is* safe, or
perhaps would be safe with more locking than we do now. I'm not sure of
the answer myself, and would have zero confidence in a newbie's answer.
Decreasing pd_lower would definitely be a win if we can do it free or
cheaply. If it requires significant additional locking overhead, then
maybe not.
regards, tom lane
On Sun, Sep 25, 2005 at 12:09:24AM -0400, Tom Lane wrote: > "Jim C. Nasby" <jnasby@pervasive.com> writes: > > On Sat, Sep 24, 2005 at 07:19:10PM -0400, Tom Lane wrote: > >> Currently, when a tuple is reclaimed by VACUUM, we just mark its item > >> pointer as unused (and hence recyclable). I think it might be safe to > >> decrease pd_lower if there are unused pointers at the end of the page's > >> pointer array, but we don't currently do that. > > > Sounds like a good newbie TODO? > > Uh, no, because the $64 question is whether it actually *is* safe, or > perhaps would be safe with more locking than we do now. I'm not sure of > the answer myself, and would have zero confidence in a newbie's answer. > > Decreasing pd_lower would definitely be a win if we can do it free or > cheaply. If it requires significant additional locking overhead, then > maybe not. Ok, sounds like a non-newbie TODO then. :) -- Jim C. Nasby, Sr. Engineering Consultant jnasby@pervasive.com Pervasive Software http://pervasive.com work: 512-231-6117 vcard: http://jim.nasby.net/pervasive.vcf cell: 512-569-9461