On 5 Jul 2004 04:02:38 -0700, A.V.C. wrote:
Hello,
I found members of this group very helpful for my last queries.
Have one problem with CASE. I can use the column name alias in Order By Clause
but unable to use it in WHERE CLAUSE.
PLS TELL ME IF IT IS POSSIBLE TO USE IT IN WHERE CLAUSE AND SOME ALTERNATIVE.
QUERY:
SELECT
M.SECS =
CASE
WHEN NO_OF_SEC IS NULL THEN -1
WHEN NO_OF_SEC =0 THEN 1
ELSE NO_OF_SEC
END
FROM DOWNLOAD_MASTER M
WHERE M.SECS < 100
ORDER BY M.SECS
Hoping for a immediate reply.
thanks in advance
Hi A.V.C.,
Before answering your question, one remark about your query. I advise you
to remove "M." before "SECS". You are using an alias; not a column name.
The name "M.SECS" looks as if you're referring to a column named SECS in
the table named (or aliased) M. I expect the above query to throw an error
because column SECS can't be found in the table DOWNLOAD_MASTER. The order
by clause will probably not throw an error, but that is only because table
names (or aliases) are largely mostly disregarded by SQL Server when
evaluating an roder by clause.
To answer your question: no, this is not possible. To understand why, it
helps to know how an SQL query gets evaluated. Note that this is a
conceptual description; a good RDBMS will change the order of operation to
optimize; as long as the results remain the same that is not a problem.
Step 1: Evaluate FROM clause, build intermediate table from all rows in
the tables used, joined together on the conditions given. If old style
join syntax is used (with the ON conditions in the WHERE clause), this
step will yield the full carthesian product of all tables used.
Step 2: Evaluate WHERE clause, remove rows that don't match the criteria
from intermediate table.
Step 3: Evaluate GROUP BY clause, group rows together according to the
specified arguments.
Step 4: Evaluate HAVING clause, remove groups that don't match the
criteria from intermediate table.
Step 5: Evaluate SELECT clause, build result set to be returned from the
data in the intermediate table.
Step 6: Evaluate ORDER BY, perform sorting.
Officially, columns that are not included in the SELECT clause are not
available for sorting. Many products (like SQL Server) do allow this, but
it is a non-standard extension of the ISO/ANSI SQL-92 specification (and I
don't think that later SQL specifications included this).
Since the alias of a columns or expression is only effective from step 5
but the WHERE clause is evaluated as step 2, it is clear that an alias
can't be used in the WHERE clause.
You also ask for alternatives. In your case, you could try either "WHERE
COALESCE(NO_OF_SEC, -1) < 100" or "WHERE NO_OF_SEC < 100 OR NO_OF_SEC IS
NULL". In more complex cases, you might have to repeat the CASE expression
in the WHERE clause. If you dislike that redundancy, you can always use
the derived table approach. For your query, the equivalent with a derived
table would look like this:
SELECT SECS
FROM (SELECT SECS = CASE
WHEN NO_OF_SEC IS NULL THEN -1
WHEN NO_OF_SEC = 0 THEN 1
ELSE NO_OF_SEC
END
FROM DOWNLOAD_MASTER) AS D
WHERE SECS < 100
ORDER BY SECS
(untested)
Best, Hugo
--
(Remove _NO_ and _SPAM_ to get my e-mail address)
