On 8/1/19, David Raymond wrote:
> [autovacuum] doesn't let you skip ahead in the overflow page chain.
Actually it does. Sort of. Sometimes. There is an optimization that
lets the btree layer "guess" the next page in an overflow chain and
verify its guess using the pointer-map information. If
https://www.sqlite.org/fileformat2.html
is the page with the nitty gritty for the file format.
Overflow pages are stored in a singly linked list of pages, so you have to
traverse through all of the pages to get to the end. So while you may know
right away that you want the 20th overflow page for
On Thu, Aug 1, 2019 at 5:44 PM Jens Alfke wrote:
> > On Jul 31, 2019, at 5:02 AM, Hick Gunter wrote:
> > SQLite stores rows in a compressed format that requires decoding. To
> access the nth field, all the fields that come before it need to be decoded.
>
> My understanding is that it’s just a ma
> On Jul 31, 2019, at 5:02 AM, Hick Gunter wrote:
>
> SQLite stores rows in a compressed format that requires decoding. To access
> the nth field, all the fields that come before it need to be decoded.
My understanding is that it’s just a matter of a byte-count before each field.
So getting t
SQLite stores rows in a compressed format that requires decoding. To access the
nth field, all the fields that come before it need to be decoded. If there is a
large blob stoed in a blob field, any field after that will suffer a
performance penalty (unless, of course, both fields are required).
5 matches
Mail list logo