Thread: Fwd: Test of Algorithm || Indexing Scheme

On 11.11.2013 13:19, Rohit Goyal wrote:
> Hi All,
>
> I want to implement and test my indexing approach.
> I would like to know which are the main files to look for b tree indexing
> scheme modification.
>
> It would be great, if can share some helpful links which are needed to
> understand how to modify and test an indexing scheme in postgresql.

You didn't give any details on what kind of changes you want to make, so 
hard to tell. The code for the btree access method is in 
src/backend/access/nbtree. See the README in that directory for the gory 
details.

Also consider implementing your algorithm as a GiST opclass, instead of 
changing the internals.

- Heikki

On 11.11.2013 14:12, Rohit Goyal wrote:
> Hi,
> Thanks for reply.
> I actually want to make some changes in all operations on index like
> insert, update, delete.
> for example, i want store a new customized value for every key inserted in
> b tree instead of storing Tuple Id as value.

That sounds pretty exotic, I can't imagine what good the index will do 
if it doesn't contain tuple IDs.

> Can you also pls explain me more about appoach to follow to use my
> algorithm as GiST opclass. I dont have much knowledge abt this also?

See user manual, and the examples in the server code and contrib.

- Heikki

<p dir="ltr">Hi,<br /> Can you pls tel me how can achieve which sound pretty exotic to you. <p dir="ltr">Can u please
sharesome relevant documents.  I havw no clue from where to start. <p dir="ltr">Regards<br /> Rohit<div
class="gmail_quot<blockquoteclass=" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">On
11.11.201314:12, Rohit Goyal wrote:<br /><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc
solid;padding-left:1ex">Hi,<br /> Thanks for reply.<br /> I actually want to make some changes in all operations on
indexlike<br /> insert, update, delete.<br /> for example, i want store a new customized value for every key inserted
in<br/> b tree instead of storing Tuple Id as value.<br /></blockquote><br /> That sounds pretty exotic, I can't
imaginewhat good the index will do if it doesn't contain tuple IDs.<br /><br /><blockquote class="gmail_quote"
style="margin:00 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"> Can you also pls explain me more about appoach to
followto use my<br /> algorithm as GiST opclass. I dont have much knowledge abt this also?<br /></blockquote><br /> See
usermanual, and the examples in the server code and contrib.<br /><br /> - Heikki<br /></div> 

On 11/12/2013 03:26 AM, Rohit Goyal wrote:
> Hi, 
> 
> Actually, 
> 
> I want to test an algorithm in which I will use store tuple_id in some
> other data structure and value of the key of index will contain some
> random created by me. Could you please tel me which file need to be
> change for it.
> 
> you said something about usage of GiST opclass. Can you please elaborate
> ur opinion in detail. It would be of a great help to me.

I think you really need to take a few steps back and explain *why* you
feel you want to modify the b-tree values.

Are you attempting to implement a covering index or index-organized table?

"some random created by me" tells us around about nothing about what
you're trying to do.


-- Craig Ringer                   http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training & Services

On Tue, Nov 12, 2013 at 11:53 AM, Craig Ringer <craig@2ndquadrant.com> wrote:
> On 11/12/2013 03:26 AM, Rohit Goyal wrote:
>> Hi,
>>
>> Actually,
>>
>> I want to test an algorithm in which I will use store tuple_id in some
>> other data structure and value of the key of index will contain some
>> random created by me. Could you please tel me which file need to be
>> change for it.
>>
>> you said something about usage of GiST opclass. Can you please elaborate
>> ur opinion in detail. It would be of a great help to me.
>
> I think you really need to take a few steps back and explain *why* you
> feel you want to modify the b-tree values.

+1

>
> Are you attempting to implement a covering index or index-organized table?
>
> "some random created by me" tells us around about nothing about what
> you're trying to do.




-- 
Regards,

Atri
l'apprenant