Обсуждение: err: select f() from i where (f()) in (select f() from x group by j);

Hi there.  Please forgive my igmorance.  

I have a table populated with lots of numbers, names, and timestamps.  I'd 
like to find the max of the sums of the numbers and the timestamp.

That is, find the timestamp and the sum of the numbers when the sum of 
the numbers is max.  Yow!

This does half -- the summing:

> select timestamp, sum(num) from timelines where (name='foo' or name='bar' or name='baz') group by timestamp;

If I create a view, with

> create view foo as select timestamp, sum(num) from timelines where (name='foo' or name='bar' or name='baz') group by
timestamp;

I get: 
Table    = foo
+----------------------------------+----------------------------------+-------+
|              Field               |              Type                | Length|
+----------------------------------+----------------------------------+-------+
| timestamp                        | datetime                         |     8 |
| sum                              | int2                             |     2 |
+----------------------------------+----------------------------------+-------+

(Note the fields' naming)  Still -- I'd like a more elegant way to do this,
than creating a view.

I came up with:

> select timestamp, max(sum) from timelines where (timestamp, sum) in (select timestamp, sum(num) from timelines where
(name='foo'or name='bar' or name='baz') group by timestamp);
 

...which returns 

< ERROR:  attribute 'sum' not found

The corresponding column in the view was titled ``sum.''  Hmmm.  So, perhaps 
fully writing the ``sum(num)'' is correct. 

> select timestamp from timelines where (timestamp, sum(num)) in (select timestamp, sum(num) from timelines where
(name='foo'or name='bar' or name='baz') group by timestamp);
 

...which returns

< ERROR:  parser: illegal use of aggregates or non-group column in target list

Or...

> select timestamp, max(sum(num)) from timelines where (timestamp, sum(num)) in (select timestamp, sum(num) from
timelineswhere (name='foo' or name='bar' or name='baz') group by timestamp);
 

...which returns

< ERROR:  type id lookup of 223198728 failed

This is making me itch.  Advice?  Wizardry?  Comments?  Flames?  
                    - chad

At 03:13 +0300 on 14/08/1999, Chad Miller wrote:


> If I create a view, with
>
> > create view foo as select timestamp, sum(num) from timelines where
>(name='foo' or name='bar' or name='baz') group by timestamp;
>
> I get:
> Table    = foo
> +-----------------------------+----------------------------------+-------+
> |              Field          |              Type                | Length|
> +-----------------------------+----------------------------------+-------+
> | timestamp                   | datetime                         |     8 |
> | sum                         | int2                             |     2 |
> +-----------------------------+----------------------------------+-------+
>
> (Note the fields' naming)  Still -- I'd like a more elegant way to do this,
> than creating a view.
>
> I came up with:
>
> > select timestamp, max(sum) from timelines where (timestamp, sum) in
>(select timestamp, sum(num) from timelines where (name='foo' or name='bar'
>or name='baz') group by timestamp);
>
> ...which returns
>
> < ERROR:  attribute 'sum' not found

The way to get rid of names that would make your life hard is to put field
aliases in the CREATE VIEW:

create view foo as select timestamp as ts_col, sum(num) as sum_col
from timelines, where ....;

Anyway, it's not the problem here. The problem is that you selected from
timelines instead of from foo.

As for a more elegant way of doing the same task, I'm not entirely sure,
because I don't have the latest PostgreSQL here, but here is a general idea:

SELECT timestamp, sum( num ) as the_sum
FROM timelines
WHERE (name='foo' or name='bar' or name='baz')
GROUP BY timestamp
ORDER BY the_sum DESC
LIMIT 1;

Herouth

--
Herouth Maoz, Internet developer.
Open University of Israel - Telem project
http://telem.openu.ac.il/~herutma