Subquery query1 below works fine.
But when I put equi condition in -sort of- nested clause like query2, it shows error ORA-00904.
Is this wrong usage of correlated subquery or it is because of other reason?
--Query1: It shows expected result.
SELECT
O.ENAME
O,SAL
,(SELECT COUNT(*)
FROM SCOTT.EMP I
WHERE I.SAL>O.SAL --correlated to outer
) AS RESULT
from SCOTT.EMP O;
--Query2:ORA-00904: "O"."SAL": invalid identifier shows. How to modify to use correlated subquery?
SELECT
O.ENAME
O,SAL
,(
WITH TEMP AS
(
SELECT COUNT(*)
FROM SCOTT.EMP I
WHERE I.SAL>O.SAL --I have put equi condistion here
)
SELECT * FROM TEMP
) AS RESULT
from SCOTT.EMP O;
I believe the second option is a wrong use of correlated subqueries, not because of the comparison, but for the use of the with clause. I would like to remember that you should avoid correlated subqueries as much as possible.
The WITH clause, or subquery factoring clause, may be processed as an inline view or resolved as a temporary table. The advantage of the latter is that repeated references to the subquery may be more efficient as the data is easily retrieved from the temporary table, rather than being requeried by each reference.
In your third column of your second query you want to get the result from the inline view. The problem is that parsing of the inline view is done independently and therefore cannot have references to anything in the outer query.
SQL> create table emp ( ename varchar2(10) , sal number ) ;
Table created.
SQL> insert into emp values ( 'AAA' , 1000 ) ;
insert into emp values ( 'BBB' , 1000 ) ;
insert into emp values ( 'CCC' , 1000 ) ;
insert into emp values ( 'DDD' , 1000 ) ;
1 row created.
SQL> SQL>
1 row created.
SQL> SQL>
1 row created.
SQL> SQL>
1 row created.
SQL> select * from emp ;
ENAME SAL
---------- ----------
AAA 1000
BBB 1000
CCC 1000
DDD 1000
To write the query with inline view, the filter must be done in the outer query
SELECT
O.ENAME
O,SAL
,(
WITH TEMP AS
(
SELECT * FROM EMP
)
SELECT count(*) FROM TEMP t WHERE t.SAL>O.SAL
) AS RESULT
from EMP O;
O SAL RESULT
---------- ---------- ----------
AAA 1000 0
BBB 1000 0
CCC 1000 0
DDD 1000 0
It has been explained that the second query does not uses with correctly.
Let me suggest, however, that your query can be simpler and more efficiently phrased. For each employee, you want to count how many employees have a greater salary. Window functions are the way to go here:
select e.*, rank() over(order by salary desc) - 1 result
from scott.emp e
Related
I have the following select statments in Oracle 19:
select count(*) as facility_load_stg_cnt from facility_load_stg;
select count(*) as facility_cnt from facility;
select count(*) as loctn_typ_cnt from loctn_typ;
select count(*) as join_cnt from (
select f.facility_id, lt.loctn_typ_id
from facility_load_stg stg
inner join facility f on stg.facility_cd = f.facility_cd
inner join loctn_typ lt on stg.bldg = lt.loctn_typ_nm);
the object is simple, get the PK's for facility and loctn_typ (facility_id, loctn_typ_id) to insert into the table that will build the relationship.
The problem is when I run the above code, I get these results:
FACILITY_LOAD_STG_CNT
---------------------
987
FACILITY_CNT
------------
645
LOCTN_TYP_CNT
-------------
188
JOIN_CNT
----------
2905
why is the select with the join have 3x the rows of the facility_load_stg table? I am sure I am doing something silly, I just cannot see it.
Folks have asked to see the table definitions, here is the relavent parts:
create table FACILITY_MAINT_DATA.FACILITY_LOAD_STG
(
FACILITY_CD VARCHAR2(100) not null,
BLDG VARCHAR2(100)
)
create table FACILITY_MAINT_DATA.FACILITY
(
FACILITY_CD VARCHAR2(100),
FACILITY_ID NUMBER(14) not null
constraint PK_FACILITY
primary key
)
create table FACILITY_MAINT_DATA.LOCTN_TYP
(
LOCTN_TYP_ID NUMBER(14) default "FACILITY_MAINT_DATA"."LOCTN_TYP_ID_SEQ"."NEXTVAL" not null
constraint PK_LOCTN_TYP
primary key,
LOCTN_TYP_NM VARCHAR2(100),
)
The generall rule if you join and you encounter a higher count that you expect is that the join keys are not unique*
Here a simple example reproducting the same result as you have (limiting only to two tables).
create table tab1 as
select rownum id,
case when rownum <= 80 then 'CD_'||rownum else 'CD_X' end cd from dual connect by level <= 645;
create table tab2 as
select rownum id,
case when rownum <= 80 then 'CD_'||rownum
when rownum <= 85 then 'CD_X'
else 'CD_Y' end cd from dual connect by level <= 987;
select count(*) from tab1;
COUNT(*)
----------
645
select count(*) from tab2;
COUNT(*)
----------
987
select count(*)
from tab1
join tab2
on tab1.cd = tab2.cd;
COUNT(*)
----------
2905
Summary change the join to use the unique keys or limit the join only to such CDcolumns that are unique.
Thank you everyone for the comments, that got me looking at detail at the actual data and there are duplications. Now I just have to figure out why, just need some investigation:)
I have created a cursor that has two queries joined with inner join, but query is not compiling their is error at the end of first query but the same query is getting executed without cursor.
cursor data is
select * from
select rid,id, order from table1
inner join
select pid, name, order from table2
on table1.order = table2.order
original query is much bigger and complicated but end result would be this.
Their are compilation errors at the end of first query and those are generic nature, I guess may be syntax for creating a two joined queries is wrong (this is a wild guess though)
Error:
SQL statement ignored //at select word of first query
Missing right parenthesis //at the last word of first query
Example based on Scott's schema:
SELECT should contain column aliases if columns returned by those inline views share the same name; otherwise, you won't know which one you're using
inline views should have their own aliases; basically, that's always a good idea - prefix columns with table aliases, otherwise you'll soon forget which column belongs to which table
SQL> declare
2 cursor data is
3 select a.empno a_empno, b.ename b_ename
4 from (select empno, ename, deptno from emp) a
5 inner join
6 (select empno, ename, deptno from emp) b
7 on a.deptno = b.deptno
8 where rownum < 5;
9 begin
10 for data_r in data loop
11 dbms_output.put_line(data_r.b_ename);
12 end loop;
13 end;
14 /
SMITH
JONES
SCOTT
ADAMS
PL/SQL procedure successfully completed.
SQL>
You have to put your subqueries in parenthesis and add aliases for the subqueries:
cursor data is
select * from
(select rid,id, order from table1) table1
inner join
(select pid, name, order from table2) table2
on table1.order = table2.order
Here is another answer for you, with just small differences and with an example:
CREATE OR REPLACE PROCEDURE p_test(n_test in number)
AS
CURSOR data
IS
SELECT *
FROM
(SELECT rid
, id
, "order" or1
FROM table1) tab1
INNER JOIN
(SELECT pid
, name
, "order" or1
FROM table2 ) tab2
ON tab1.or1 = tab2.or1;
BEGIN
FOR data_i IN data LOOP
DBMS_OUTPUT.PUT_LINE(data_i.rid);
END LOOP;
END p_test;
Here is the DEMO
Code is written below, multiple nested query:
select *
from
(select employee_id,employee_id
from employees) a
where rownum <= 5
)
where rnum >= 10
If i give duplicate columns in the select its giving "column ambiguously defined" error.
employee_id,employee_id, rownum rnum
Firstly, your query is incorrect since you have ambiguously defined the columns. It will throw ORA-00918: column ambiguously defined.
You must use proper ALIAS to avoid the error. For example,
SELECT departments.department_id AS "dept_id",
employees.department_id AS "emp_Dept_id"
FROM...
Secondly, it is not at all a pagination query. Since you are alwyas going to pick random rows as there is no ORDER BY clause. You are not ordering the rows.
where rownum <= 5
where rnum >= 10
At last, how on earth could you try to fetch the rows beyond 10 when you have fetched ONLY 5 rows in the inner query? It will ALWAYS return zero rows.
The correct way of paging through data is:
SQL> SELECT empno
2 FROM (SELECT empno, rownum AS rnum
3 FROM (SELECT empno
4 FROM emp
5 ORDER BY sal)
6 WHERE rownum <= 8)
7 WHERE rnum >= 5;
EMPNO
----------
7654
7934
7844
7499
SQL>
when you give a.* it means you are trying to refer two columns with same name in a table which is not permitted.The column names in a table are unique
So select employee_id,employee_id from employees is not a problem
but
select a.* from (select employee_id,employee_id from employees)a is a problem
the sqlfiddle here
Also if you want records from 10 to 15 in your query then use like the below
select * from
(select a.*,Rownum rnum from
(select employee_id as emp_id1,employee_id as emp_id2
from employees order by 1)
a where rownum <= 15 ) where rnum >= 10
EDIT1:- If duplicate column is required use like the below
with emp1 as (select employee_id from employees),
emp2 as (select * from
(select a.*,rownum rnum from emp1 a order by 1)
where rownum <=15)
select b.*,c.*
from emp1 b,emp2 c
where b.employee_id=c.employee_id
and c.rnum >=10
I need to select columns from a table Table_A, However there is another table which has the same schema Table_B. The query should determine the from table dynamically. For ex. if Table_A has more rows then use Table_A else use Table_B.
Query something like this
select employee, salary, id from (condition to count rows and select the table )table;
Is this possible without using cursors and EXECUTE IMMEDIATE??.
Normally, you would use dynamic SQL for this sort of thing. That would involve either using the DBMS_SQL package, EXECUTE IMMEDIATE or doing an OPEN <<cursor>> FOR <<SQL statement string>>.
If you really want to use static SQL, you could query both tables and only return one set of results. I cannot envision a situation where this would really make sense but you can certainly do it
Create a FOO and a FOO2 table. FOO2 has two rows to the one row from FOO
SQL> create table foo( col1 number );
Table created.
SQL> create table foo2( col1 number );
Table created.
SQL> insert into foo values( 1 );
1 row created.
SQL> insert into foo2 values( 1 );
1 row created.
SQL> insert into foo2 values( 2 );
1 row created.
Run the query. This will return all the data from FOO2
SQL> ed
Wrote file afiedt.buf
1 select col1
2 from (select the_union.*,
3 max(cnt) over () max_cnt
4 from (select col1, count(*) over () cnt from foo
5 union all
6 select col1, count(*) over () from foo2) the_union)
7* where cnt = max_cnt
SQL> /
COL1
----------
1
2
Insert more rows into FOO. Now the same query will return all the data from FOO
SQL> insert into foo values( 3 );
1 row created.
SQL> insert into foo values( 5 );
1 row created.
SQL> commit;
Commit complete.
SQL> select col1
2 from (select the_union.*,
3 max(cnt) over () max_cnt
4 from (select col1, count(*) over () cnt from foo
5 union all
6 select col1, count(*) over () from foo2) the_union)
7 where cnt = max_cnt;
COL1
----------
1
3
5
As I said, though, I cannot fathom a situation where it would actually make sense to do this.
I am completely guessing at what you are really trying to do, but I am thinking you want to use a synonym. I am guessing that there is some sort of event that triggers when you should be using TableA vs. TableB. Create a synonym "my_table" pointing to TableA and have your view select from "my_table". Then whenever you want your view to point to TableB instead, you just switch your synonym to point to TableB, and you don't have to do anything to the view.
For more info, read about synonyms in the Concepts Guide.
I use following query to insert a new row:
insert into table1 (c1, c2, c3) (select c1, c2, c3 from table2 where some_condition)
This works finely if there is a row in table2 that satisfies some_condition. But if there are no rows, nothing is inserted.
Are there any way to specify default values to insert if select returns empty result set? I want to do that in one sql query.
This isn't very pretty, but it does what you want, You'd need to test with your environment to see if it performs well enough
SQL> drop table so_tgt;
Table dropped.
SQL>
SQL> create table so_src (
2 c1 varchar2(6)
3 ,c2 varchar2(6)
4 ,c3 varchar2(6)
5 );
Table created.
SQL>
SQL> insert into so_src values ( 'foo','bar','moo' );
1 row created.
SQL>
SQL> create table so_tgt as select * from so_src where 1 = 0;
Table created.
SQL>
SQL> /* Test for existing row insert */
SQL> insert into so_tgt
2 with x as ( select s.*, 1 as r
3 from so_src s
4 where c1='foo'
5 union
6 select 'x','y','z',0 as r /* DEFAULT VALUES */
7 from dual )
8 select c1,c2,c3
9 from x
10 where r = ( select max(r) from x ) ;
1 row created.
SQL>
SQL> select * from so_tgt;
C1 C2 C3
------ ------ ------
foo bar moo
SQL> truncate table so_tgt;
Table truncated.
SQL>
SQL> /* Test for default row insert */
SQL> insert into so_tgt
2 with x as ( select s.*, 1 as r
3 from so_src s
4 where c1='far'
5 union
6 select 'x','y','z',0 as r /* DEFAULT VALUES */
7 from dual )
8 select c1,c2,c3
9 from x
10 where r = ( select max(r) from x ) ;
1 row created.
SQL>
SQL> select * from so_tgt;
C1 C2 C3
------ ------ ------
x y z
SQL> truncate table so_tgt ;
Table truncated.
The quick and dirty way, if you don't mind repeating some_condition and where some_condition doesn't depend on the values in table2 is:
insert into table1 (c1,c2,c3)
select c1, c2, c3 from table2 where some_condition
union select defaultvalue1, defaultvalue2, defaultvalue3 from dual where not (some_condition)
If some_condition does depend on values in table2, then you can do (untested):
insert into table1 (c1,c2,c3)
select nvl(t2.c1, defaultvalue1), nvl(t2.c2, defaultvalue2), nvl(t2.c2, defaultvalue3)
from dual left join (select c1,c2,c3 from table2 where some_condition) t2
on 1 = 1
If I'm right, this query will always return at least one row, but if no rows showed up on the right side, then the t2 values will all be returned as null and so the nvl can be used to provide your default values.
Edit: Small caveat. This assumes that the values returned from table2 will not be null or that if they are, you want the default values.