EXPERT RESPONSE
To answer your question, yes, you can do it.
You didn't say whether you wanted exactly
one S and and P, or at least one S and one P,
so I'll show you how to do it either way.
Let's start with this query:
select id
, loc_numb
, loc_prec
from item_loc
where id in
( select id
from item_loc
where loc_prec in ('S','P')
group by id
having count(*) = 2 )
Let's look first at the subselect. The subselect returns all S and P
rows in the table. Then it groups them by ID. If a particular ID
has exactly 2 rows, then it is selected, and both of its
rows are included in the result set of the outer query.
But is this correct? We are assuming that there is one S and
one P, but there's a "gotcha" in that logic. The subquery
also chooses every ID that
has two S's and no P's, or two P's and no S's.
And if there's a particular ID that happens
to have one S and two P's? It would not be chosen!
We could fix this by using
having count(*) >= 2
But this still lets an ID with only three P's or only three S's through.
This is familiar OR behaviour, the result of
"mingling" the selected rows in the
IN?('S','P') condition.
Let's see what happens when we split the S and P conditions
into separate subselects, so that the selected rows
are not mingled.
select id
, loc_numb
, loc_prec
from item_loc
where id in
( select id
from item_loc
where loc_prec = 'S'
group by id
having count(*) = 1 )
and id in
( select id
from item_loc
where loc_prec = 'P'
group by id
having count(*) = 1 )
This query lists ID's which have exactly one S and one P.
Use an equality condition in the HAVING clause to
specify an exact number of rows.
To find one or more rows, use an inequality.
For example, to find every ID with one S and at least one P,
the query is:
select id
, loc_numb
, loc_prec
from item_loc
where id in
( select id
from item_loc
where loc_prec = 'S'
group by id
having count(*) = 1 )
and id in
( select id
from item_loc
where loc_prec = 'P'
group by id
having count(*) >= 1 )
If you think about it for a moment,
a HAVING clause that says COUNT(*)?>=?1
is redundant. It's easy to see why, if
you consider how a query with a GROUP BY clause is processed.
First, rows are retrieved based on WHERE conditions.
Then the selected rows are assembled into groups, typically by sorting.
Then the groups are processed. There will not be
a group with no rows, since during selection,
no rows were selected for that group. Therefore each group
that comes out of row selection
and goes into group processing has to have one or more rows.
This answer is continued
|
