postgres-xc-developers Mailing List for Postgres-XC

I am ok if NODE_NAME branch is removed, since it is not required any more.

On Wed, Apr 11, 2012 at 1:40 PM, Michael Paquier
<mic...@gm...>wrote:

> Hi all,
>
> Could it be possible to clean up a bit the Github repository?
> More and more people are following our activities there and it might be
> clearer for everybody if we remove some branches like GTM_HA_*, NODE_NAME,
> PGXC_CLEAN...
> You should perhaps create a private repository the same way I do for
> personal developments:
> https://github.com/michaelpq/pgxc
>
> Regards,
> --
> Michael Paquier
> http://michael.otacoo.com
>
>
> ------------------------------------------------------------------------------
> Better than sec? Nothing is better than sec when it comes to
> monitoring Big Data applications. Try Boundary one-second
> resolution app monitoring today. Free.
> http://p.sf.net/sfu/Boundary-dev2dev
> _______________________________________________
> Postgres-xc-developers mailing list
> Pos...@li...
> https://lists.sourceforge.net/lists/listinfo/postgres-xc-developers
>
>

On 11 April 2012 12:48, Michael Paquier <mic...@gm...> wrote:

>
>
> On Wed, Apr 11, 2012 at 4:01 PM, Amit Khandekar <
> ami...@en...> wrote:
>
>> Hi Michael,
>>
>> Can you a bit explain the context? Do you just want a tuple descriptor,
>> given a table oid (which is available in Relation structure once the table
>> is opened) ? Or are you referring to a rowtype as in composite type ? There
>> is a function lookup_rowtype_tupdesc() which returns tupledesc, given a
>> composite type id.
>>
> A rowtype as a composite type I think.
>
> However, I am trying to solve a tricky case so you might understand easier
> with that:
> create table id (id serial primary key, name text);
> create view id_ordered as select * from id order by id;
> create rule update_id_ordered as on update to id_ordered do instead update
> id set name = new.name where id = old.id;
> postgres=# explain verbose update id_ordered set name = 'update 2' where
> id = 2;
>                                                        QUERY
> PLAN
>
> -------------------------------------------------------------------------------------------------------------------------
>  Update on public.id  (cost=0.00..20.00 rows=1000 width=74)
>    Node/s: dn1, dn2
>    Node expr: public.id.id
>    Remote query: UPDATE public.id SET name = $2 WHERE id = $3 AND ctid =
> $4 AND xc_node_id = $5 AND id = $1 AND $6 = $6
>    ->  Nested Loop  (cost=0.00..20.00 rows=1000 width=74)
>          Output: public.id.id, 'update 2'::text, public.id.id,
> public.id.ctid, public.id.xc_node_id, *id_ordered.**
>          ->  Data Node Scan on id  (cost=0.00..0.00 rows=1000 width=14)
>                Output: public.id.id, public.id.ctid, public.id.xc_node_id
>                Node/s: dn1
>                Remote query: SELECT id, ctid, xc_node_id FROM ONLY id
> WHERE (id = 2)
>          ->  Subquery Scan on id_ordered  (cost=0.00..10.00 rows=1000
> width=64)
>                Output: id_ordered.*, id_ordered.id
>                ->  Result  (cost=0.00..0.00 rows=1000 width=36)
>                      Output: public.id.id, public.id.name
>                      ->  Data Node Scan on id  (cost=0.00..0.00 rows=1000
> width=36)
>                            Output: public.id.id, public.id.name
>                            Node/s: dn1
>                            Remote query: SELECT id, name FROM ONLY id
> WHERE (id = 2)
>
> Here I got *id_ordered.** which is a row image of the view id_ordered.
> And I am looking for a way to find its type to push down its value and
> nullify it.
> I think get_rel_type_id can do it.
>
>
Are you referring to the other patch which you sent ? I am yet to go
through it and understand what do you mean by "nullifying it".

If you want to pass values for id_ordered.* parameter, I don't think we can
make a WHERE clause out of it unless we expand the row var into all of its
columns and then generate WHERE condition for each of the column.

But before that, I am feeling we should ultimately remove this logic of
adding all of the target list columns into the WHERE clause.  In your case,
the update query is:
   Remote query: UPDATE public.id SET name = $2 WHERE id = $3 AND ctid = $4
AND xc_node_id = $5 AND id = $1 AND $6 = $6
Here, the table being updated is public.id. How come you can add a WHERE
clause having id_ordered columns? The target list contains columns from
different table, so you should not include these columns right? I think in
your other patch, you were trying to exclude columns from other table?


I have also another problem, I am able to find the relation name easily by
> looking into the target list with the view index but the Oid of the view is
> not in its RangeTblEntry entry. Any ideas?
>

Yes, unfortunately RangeTblEntry does not have it, I know. You will want to
find some other way.

>
>
>
>> -Amit
>>
>>
>> On 11 April 2012 11:44, Michael Paquier <mic...@gm...>wrote:
>>
>>> Hi all,
>>>
>>> Is there an easy way to get a rowtype with a table Oid?
>>> A rowtype is a type dedicated to a table to represent a complete row of
>>> the table.
>>> Regards,
>>> --
>>> Michael Paquier
>>> http://michael.otacoo.com
>>>
>>>
>>> ------------------------------------------------------------------------------
>>> Better than sec? Nothing is better than sec when it comes to
>>> monitoring Big Data applications. Try Boundary one-second
>>> resolution app monitoring today. Free.
>>> http://p.sf.net/sfu/Boundary-dev2dev
>>> _______________________________________________
>>> Postgres-xc-developers mailing list
>>> Pos...@li...
>>> https://lists.sourceforge.net/lists/listinfo/postgres-xc-developers
>>>
>>>
>>
>
>
> --
> Michael Paquier
> http://michael.otacoo.com
>

Hi all,

Could it be possible to clean up a bit the Github repository?
More and more people are following our activities there and it might be
clearer for everybody if we remove some branches like GTM_HA_*, NODE_NAME,
PGXC_CLEAN...
You should perhaps create a private repository the same way I do for
personal developments:
https://github.com/michaelpq/pgxc

Regards,
-- 
Michael Paquier
http://michael.otacoo.com

On Wed, Apr 11, 2012 at 4:01 PM, Amit Khandekar <
ami...@en...> wrote:

> Hi Michael,
>
> Can you a bit explain the context? Do you just want a tuple descriptor,
> given a table oid (which is available in Relation structure once the table
> is opened) ? Or are you referring to a rowtype as in composite type ? There
> is a function lookup_rowtype_tupdesc() which returns tupledesc, given a
> composite type id.
>
A rowtype as a composite type I think.

However, I am trying to solve a tricky case so you might understand easier
with that:
create table id (id serial primary key, name text);
create view id_ordered as select * from id order by id;
create rule update_id_ordered as on update to id_ordered do instead update
id set name = new.name where id = old.id;
postgres=# explain verbose update id_ordered set name = 'update 2' where id
= 2;
                                                       QUERY
PLAN
-------------------------------------------------------------------------------------------------------------------------
 Update on public.id  (cost=0.00..20.00 rows=1000 width=74)
   Node/s: dn1, dn2
   Node expr: public.id.id
   Remote query: UPDATE public.id SET name = $2 WHERE id = $3 AND ctid = $4
AND xc_node_id = $5 AND id = $1 AND $6 = $6
   ->  Nested Loop  (cost=0.00..20.00 rows=1000 width=74)
         Output: public.id.id, 'update 2'::text, public.id.id,
public.id.ctid, public.id.xc_node_id, *id_ordered.**
         ->  Data Node Scan on id  (cost=0.00..0.00 rows=1000 width=14)
               Output: public.id.id, public.id.ctid, public.id.xc_node_id
               Node/s: dn1
               Remote query: SELECT id, ctid, xc_node_id FROM ONLY id WHERE
(id = 2)
         ->  Subquery Scan on id_ordered  (cost=0.00..10.00 rows=1000
width=64)
               Output: id_ordered.*, id_ordered.id
               ->  Result  (cost=0.00..0.00 rows=1000 width=36)
                     Output: public.id.id, public.id.name
                     ->  Data Node Scan on id  (cost=0.00..0.00 rows=1000
width=36)
                           Output: public.id.id, public.id.name
                           Node/s: dn1
                           Remote query: SELECT id, name FROM ONLY id WHERE
(id = 2)

Here I got *id_ordered.** which is a row image of the view id_ordered. And
I am looking for a way to find its type to push down its value and nullify
it.
I think get_rel_type_id can do it.

I have also another problem, I am able to find the relation name easily by
looking into the target list with the view index but the Oid of the view is
not in its RangeTblEntry entry. Any ideas?



> -Amit
>
>
> On 11 April 2012 11:44, Michael Paquier <mic...@gm...> wrote:
>
>> Hi all,
>>
>> Is there an easy way to get a rowtype with a table Oid?
>> A rowtype is a type dedicated to a table to represent a complete row of
>> the table.
>> Regards,
>> --
>> Michael Paquier
>> http://michael.otacoo.com
>>
>>
>> ------------------------------------------------------------------------------
>> Better than sec? Nothing is better than sec when it comes to
>> monitoring Big Data applications. Try Boundary one-second
>> resolution app monitoring today. Free.
>> http://p.sf.net/sfu/Boundary-dev2dev
>> _______________________________________________
>> Postgres-xc-developers mailing list
>> Pos...@li...
>> https://lists.sourceforge.net/lists/listinfo/postgres-xc-developers
>>
>>
>


-- 
Michael Paquier
http://michael.otacoo.com

Hi Michael,

Can you a bit explain the context? Do you just want a tuple descriptor,
given a table oid (which is available in Relation structure once the table
is opened) ? Or are you referring to a rowtype as in composite type ? There
is a function lookup_rowtype_tupdesc() which returns tupledesc, given a
composite type id.

-Amit


On 11 April 2012 11:44, Michael Paquier <mic...@gm...> wrote:

> Hi all,
>
> Is there an easy way to get a rowtype with a table Oid?
> A rowtype is a type dedicated to a table to represent a complete row of
> the table.
> Regards,
> --
> Michael Paquier
> http://michael.otacoo.com
>
>
> ------------------------------------------------------------------------------
> Better than sec? Nothing is better than sec when it comes to
> monitoring Big Data applications. Try Boundary one-second
> resolution app monitoring today. Free.
> http://p.sf.net/sfu/Boundary-dev2dev
> _______________________________________________
> Postgres-xc-developers mailing list
> Pos...@li...
> https://lists.sourceforge.net/lists/listinfo/postgres-xc-developers
>
>

Hi all,

Is there an easy way to get a rowtype with a table Oid?
A rowtype is a type dedicated to a table to represent a complete row of the
table.
Regards,
-- 
Michael Paquier
http://michael.otacoo.com