Postgres Pro
Downloads
Home   >   mailing lists

Thread: getting the last N tuples of a query

getting the last N tuples of a query

From
"Edmundo Robles L."
Date:
Hi!

if a want the first   5,10,N tuples  of a query  (even without order)
i just  have to do a:
select  * from table limit 10;


but, What can i do to get the last  10 tuples ???
  i try to do:
select * from table limit -10;  :-D  but  that query  return 0 tuples .

So, what is the right way to do that  with no order????

Re: getting the last N tuples of a query

From
Ben Chobot
Date:
On Jul 8, 2010, at 4:17 PM, Edmundo Robles L. wrote:

> Hi!
>
> if a want the first   5,10,N tuples  of a query  (even without order)
> i just  have to do a:
> select  * from table limit 10;

That does not get the first 10 tuples, it merely gets 10 tuples. The database is free to return whichever 10 it can,
andin practice, the results will change given enough inserts or deletes. 

> So, what is the right way to do that  with no order????


Without an order by clause, there is no concept of "first" or "last". Once you have the order by clause, combine your
limitwith ascending or descending sorts to get the first or last, respectively.

Re: getting the last N tuples of a query

From
Kenichiro Tanaka
Date:
Hello.

I agree Ben.
But,I try your question as an SQL puzzle.
Doses this SQL meet what you want?

select * from wantlast offset (select count(*)-10 from wantlast);

--test case
create table wantlast(col1 int);
insert into wantlast select g from generate_series(1,1000) as g;

postgres=# select * from wantlast offset (select count(*)-10 from wantlast);
  col1
------
   991
   992
   993
   994
   995
   996
   997
   998
   999
  1000
(10 rows)

postgres=# analyze wantlast ;
ANALYZE
postgres=# explain  select * from wantlast offset (select count(*)-10
from wantlast);
                                 QUERY PLAN
--------------------------------------------------------------------------
  Limit  (cost=17.91..30.52 rows=900 width=4)
    InitPlan 1 (returns $0)
      ->  Aggregate  (cost=16.50..16.52 rows=1 width=0)
            ->  Seq Scan on wantlast  (cost=0.00..14.00 rows=1000 width=0)
    ->  Seq Scan on wantlast  (cost=0.00..14.00 rows=1000 width=4)
(5 rows)

*I try this test Postgresql8.4.4



> On Jul 8, 2010, at 4:17 PM, Edmundo Robles L. wrote:
>
>
>> Hi!
>>
>> if a want the first   5,10,N tuples  of a query  (even without order)
>> i just  have to do a:
>> select  * from table limit 10;
>>
> That does not get the first 10 tuples, it merely gets 10 tuples. The database is free to return whichever 10 it can,
andin practice, the results will change given enough inserts or deletes. 
>
>
>> So, what is the right way to do that  with no order????
>>
>
> Without an order by clause, there is no concept of "first" or "last". Once you have the order by clause, combine your
limitwith ascending or descending sorts to get the first or last, respectively. 
>

Re: getting the last N tuples of a query

From
Merlin Moncure
Date:
On Thu, Jul 8, 2010 at 9:09 PM, Kenichiro Tanaka
<ketanaka@ashisuto.co.jp> wrote:
> Hello.
>
> I agree Ben.
> But,I try your question as an SQL puzzle.
> Doses this SQL meet what you want?
>
> select * from wantlast offset (select count(*)-10 from wantlast);

that works, but for any non trivial query it's not optimal because it
runs the complete query twice.  if you are doing a lot of joins, etc.
(or your query involves volatile operations) you might want to avoid
this.

cursors can do it:
begin;
declare c scroll cursor for select generate_series(1,1000);
fetch last from c; -- discard result
fetch backward 10 from c; -- discard result
fetch 10 from c; -- your results
commit;

in 8.4 you can rig it with CTE:
with foo as (select generate_series(1,1000) v) select * from foo
offset (select count(*) - 10 from foo);

the advantage here is you are double scanning the query results, not
rerunning the query (this is not guaranteed to be a win, but it often
will be).

you can often rig it with arrays (dealing with non scalar type arrays
is only possible in 8.3+)
select unnest(a[array_upper(a, 1)-10:array_upper(a,1)]) from (select
array(select generate_series(1,1000) v) as a) q;

merlin