RE: How to emulate multiple DISTINCT aggregates?

[Suavi Ali Demir]

Let's look at the query differently to see what it is trying to say:

 

SELECT  count(some columns) FROM table_a a
WHERE ( some field1 criteria thats always true )
AND ( some field2 criteria thats always true )
AND (some field3 criteria thats always true )

The where clause will always be true. For a given row, if where clause is true, the tuples of that row will be returned. That would make 3 rows (all rows) returned in this case since where clause is always true. And which column we count does not matter i guess.

 

Regards,

Ali

[EMAIL PROTECTED] wrote:

That doesn’t make sense.

Each sub select statement should have returned the following

Field A {1,2}

Field B {1,2}

Field C {1,2,3}

 

Then in the original select, your counts should be {2,2,3};

 

Note: I used Informix.

 

There is something interesting going on. You’re using a dummy table as a wrapper for 3.

Now I wonder what the optimizer thinks of this and how, if at all, does it parallelize the query?

 

I’ll try my sql method on a similar table to yours, but I do think the response you saw wasn’t correct.

 

---

From: Suavi Ali Demir [mailto:[EMAIL PROTECTED]
Sent: Monday, August 28, 2006 3:52 PM
To: Derby Discussion; [EMAIL PROTECTED]
Subject: RE: How to emulate multiple DISTINCT aggregates?

 

That query needs to be modified little bit. It does not work when

my table contains:

 

1	1	1
2	2	2
1	1	3

 

result is 3,3,3, where as it should have been 2,2,3.

 

This one works:

 

SELECT
(SELECT COUNT(DISTINCT b.field1) FROM table_a b) as field1_count,
(SELECT COUNT(DISTINCT c.field2) FROM table_a c) as field2_count,
(SELECT COUNT(DISTINCT d.field3) FROM table_a d) as field3_count
FROM SYSIBM.SYSDUMMY1

 

Regards,

Ali

[EMAIL PROTECTED] wrote:

Hi,

Uhm you can't do what you want to do as written....

When you say SELECT COUNT(DISTINCT blah), it has to be unique to your query.
So no multiple COUNT(DISTINCTS xxx) allowed in a single query.
However here's a work around that may help...

SELECT COUNT(a.field1), COUNT(a.field2), COUNT(a.field3) ...
FROM table_a a
WHERE a.field1 in ( SELECT DISTINCT b.field1 FROM table_a b)
AND a.field2 in ( SELECT DISTINCT b.field2 FROM table_a b)
AND a.field3 in ( SELECT DISTINCT b.field3 FROM table_a b)
...

Well you get the idea.

It's then pretty straight forward.
Each subselect returns the distinct set of elements and then your outer
select returns the count on these unique subsets.

I just tried it in Informix (I happen to have a table with a couple of
string columns of data that I am working on.)
Should work in Derby. Its all standard SQL

But Hey! What do I know?

-G

> -----Original Message-----
> From: Robert Enyedi [mailto:[EMAIL PROTECTED]
> Sent: Monday, August 28, 2006 10:58 AM
> To: Derby Discussion
> Subject: How to emulate multiple DISTINCT aggregates?
>
> Hi,
>
> When trying to use multiple DISTINCT aggregates in Derby:
>
> /SELECT COUNT(distinct editable), COUNT( distinct visible )
> FROM item/
>
> the following error is reported:
>
> /ERROR 42Z02: Multiple DISTINCT aggregates are not supported at this
> time./
>
> For a simple query one could write:
>
> /SELECT COUNT(distinct editable) AS editable_count, COUNT( (SELECT
> DISTINCT visible FROM item) )
> FROM item/
>
> but this is actually pretty uncomfortable and unmaintainable in a larger
> query.
>
> Does anyone know an elegant overcome for this Derby limitation?
>
> Thanks,
> Robert