From: Paul Wilson
Date: February 21 2013 4:36am
Subject: Re: a doubt about page directory
List-Archive: http://lists.mysql.com/innodb/72
Message-Id: <CAKs83S-gdu5QZfrvkQpcy31h5TC=kE+k7G9o9=vskyAbKJBrBw@mail.gmail.com>
MIME-Version: 1.0
Content-Type: multipart/alternative; boundary=047d7b621f2a005c1804d634a102

--047d7b621f2a005c1804d634a102
Content-Type: text/plain; charset=windows-1252
Content-Transfer-Encoding: quoted-printable

UNSUBSCRIBE

On Wed, Feb 20, 2013 at 10:25 PM, Lixun Peng <penglixun@stripped> wrote:

> Hi Bohu,
>
> I think your understanding about Page Directory is wrong.
>
> In one page, the records usually are ordered in physical, Page Directory
> records some rows' offset.
> So if you have these rows in a page:
> [Infimum, offset:99]->[Supremum, offset:112]->[Key:0, offset 125]->...two
> rows...->[Key:3, offset: 221]->...three rows..->[Key:7, offset: 349]
> Then you maybe see these offsets in Page Directory:
> [99, 221, 349, 112]
>
> Why InnoDB need to use the page directory for speed up searching one row =
in
> a page?
> I think it can save parsing time, if no page directory, InnoDB need to
> parse entire page, and parse each rows one by one, and then you can find
> the row you need.
> But if has page directory, we can use these offsets to skip some rows, fo=
r
> example. we need to search Key=3D7, so we can go to offset 99, it's not K=
ey
> 7, and then we go to offset 221, it's not Key 7, too.
> Next, we go to 349, it's Key 7, it's the row I need.
>
> That's the page directory used for.
>
> Thanks,
> Lixun
>
>
> On Thu, Feb 21, 2013 at 11:42 AM, Bohu TANG <overred.shuttler@stripped
> >wrote:
>
> > Hello,
> >
> > I know the directory page is designed to faster read using binary searc=
h,
> > since the records of one page is unordered(in physical).
> > With the purpose of binary search, we also can do it in the same page:
> > 1) When you do first search in a page, we overwrite the page with the
> > ordered list(because one page is 16KB or others, but it=92s very small,=
 and
> > close to the disk transfer block in modern)
> > 2) The second time you search it you can do binary search internal  thi=
s
> > page, so do not need the page directory.
> >
> > Why not it?
> >
> > Thanks
> > ---
> > BohuTANG
> >
> > "The great artist is the simplifier."--- *Vincent Van Gogh*
> >
>
>
>
> --
> Senior MySQL Developer @ Taobao.com
>
> Mobile Phone: +86 18658156856 (Hangzhou)
> Gtalk: penglixun(at)gmail.com
> Twitter: http://www.twitter.com/plinux
> Blog: http://www.penglixun.com
>

--047d7b621f2a005c1804d634a102--