I have a little complex requirement on couple of tables which I am finding hard to crack.
There are 2 tables. TableA and TableB
TableA has a structure like:
-------------------------------------
ID COL1 COL2 CAT
-------------------------------------
1 RecAA RecAB 3
2 RecBA RecBB 3
3 RecCA RecCB 2
4 RecDA RecDB 2
5 RecEA RecEB 1
-------------------------------------
TableB has a structure like:
-----------------
COL3 TYPE
-----------------
RecAA 10
RecAA 11
RecAA 12
RecAB 10
RecAB 11
RecAB 12
RecAB 13
RecAB 14
RecBA 10
RecBA 11
RecBA 14
RecBA 15
RecBB 10
-----------------
Requirements:
Records in TableA should have CAT = 3.
Either COL1 or COL2 of TableA should be available in COL3 of TableB.
COL3 should definitely have TYPE in 10,11,12 and should have only that TYPE.
i.e As per the above requirements,
Of the records available in TableA, records with ID 1 and 2 have CAT = 3 in TableA
Both the records have atleast only value in COL3 of TableB. (Record with ID 1 in TableA has both COL1 and COL2 in TableB and record with ID 2 in TableA has COL1 in TableB)
RecAA record has Type 10,11,12 and only 10,11,12. So doesnt matter if RecAB has 10,11,12 or not. But RecBA and RecBB both does not have 10,11,12 types.
Therefore the result should be:
-------------------------------------
ID COL1 COL2 CAT
-------------------------------------
1 RecAA RecAB 3
-------------------------------------
What I tried:
WITH TEMP AS (SELECT COL3 FROM TableB GROUP BY COL3 HAVING SUM(CASE WHEN TYPE IN ('10','11','12') THEN 1 ELSE 0 END) = 0)
SELECT S.ID, S.COL1, S.COL2, S.CAT FROM TableA S
INNER JOIN TEMP T ON S.COL1 = T.COL3
WHERE S.CAT = 3;
Can someone please help on achieving this?
I think you're almost there, it's just your row selection in the CTE that seems problematic, and I think you need an OR:
WITH TEMP AS (
SELECT COL3
FROM TableB
GROUP BY COL3
HAVING SUM(POWER(2, TYPE - 10)) = 7 AND COUNT(*) = 3
)
SELECT
S.ID, S.COL1, S.COL2, S.CAT
FROM
TableA S
INNER JOIN TEMP T ON S.COL1 = T.COL3 OR S.COL2 = T.COL3
WHERE
S.CAT = 3;
I've subtracted 10 from each of your TYPEs to turn your 10,11,12 into 0,1,2 and then used POWER to turn them into 1, 2 and 4 which uniquely sum to 7 - (in other words your 10,11,12 became 2^(10-10), 2^(11-10) and 2^(12-10) which are 1, 2 and 4.. Which must then sum to 7).
I also mandate that there be a count of 3; the only way to get to 7 with three numbers that are powers of 2 is to have 1+2+4 which guarantees that 10,11,12 are present initially. If anything was missing, extra or repeated it wouldn't be 3 numbers that sum to 7
I think RecAB is excluded because even though it has 10,11,12 it also has 13,14 which cause it to be excluded..
You also seemed to be saying that COL3 should be present in either COL1 or COL2 of table A
You can use listagg analytic version to turn TYPE column into type_in_list column like below :
With temp_TableB (COL3, type_in_list) as (
SELECT distinct COL3, listagg(TYPE, ',') within group (order by TYPE)over(partition by COL3)
FROM TableB
)
select tA.*
--, tb.*
from tableA tA
INNER JOIN temp_TableB tB on (tA.COL1 = tB.COL3 or tA.COL2 = tB.COL3)
Where tA.CAT = 3
AND tB.TYPE_IN_LIST = '10,11,12'
;
Related
I have a query which will fetch two rows only and I want to bring second row data into columns with different column name.
Below is the original query result.
The expected result is like
Expected result.
Please help how shd I proceed, not able to figure out with PIVOT.
Here's one option; see comments within code.
SQL> with
2 your_query (column1, column2, column3) as
3 -- this is what your current query returns
4 (select 1, 'ABC', 123 from dual union all
5 select 2, 'XYZ', 456 from dual
6 ),
7 temp as
8 -- distinguish 1st from 2nd row
9 (select y.*,
10 row_number() over (order by column1) rn
11 from your_query y
12 )
13 -- finally, cross join two rows and conditionally display columns.
14 -- MAX is here to avoid empty "cells"
15 select max(case when a.rn = 1 then a.column1 end) as col1,
16 max(case when a.rn = 1 then a.column2 end) as col2,
17 max(case when a.rn = 1 then a.column3 end) as col3,
18 --
19 max(case when b.rn = 2 then b.column1 end) as col4,
20 max(case when b.rn = 2 then b.column2 end) as col5,
21 max(case when b.rn = 2 then b.column3 end) as col6
22 from temp a cross join temp b;
COL1 COL COL3 COL4 COL COL6
---------- --- ---------- ---------- --- ----------
1 ABC 123 2 XYZ 456
SQL>
I have 2 tables - tab1 , tab2 with following data
tab1 data:
OID Label
1 MX1
1 MX2
1 MX3
2 MX4
2 MX5
tab2 data:
OID ID Label
1 5678
1 2347
1 9687
2 4567
2 3455
The join condition between these two tables is oid column.I need to create a process which will update Label column from tab1 to Label column of tab2.It doesn't matter which label gets assigned to which record of tab2 for a particular oid. The only check that needs to happen is that both the tables should have same number of records for an oid.The final outcome should be the following
tab2 data:
OID ID Label
1 5678 MX1
1 2347 MX2
1 9687 MX3
2 4567 MX4
2 3455 MX5
Again, it doesn't matter which label gets assigned to tab2 for a particular oid,but the same label cannot be repeated for a particular oid.What would be the best way to write a code for this requirement?
Here is a sql solution:
merge into tab2
using
(
select t2."id" as ide,t1."label" labela from
(select rownum n,"label","oid" from tab1 order by "oid")t1,
(select rownum n, a2.* from tab2 a2 order by "oid")t2
where t1.n=t2.n and
t1."oid"=t2."oid"
) tb4
on (tab2."id" = tb4.ide)
when matched then
update set tab2."label" = tb4.labela;
Result:
oid| id | label
-----------------
1 5678 mx1
1 2347 mx2
1 9687 mx3
2 4567 mx4
2 3455 mx5
Sample tables:
SQL> select * from tab1 order by oid, label;
OID LAB
---------- ---
1 mx1
1 mx2
1 mx3
2 mx4
2 mx5
SQL> select * from tab2 order by oid, id;
OID ID LAB
---------- ---------- ---
1 2347
1 5678
1 9687
2 3455
2 4567
SQL>
This is query that returns desired result:
SQL> with
2 t1 as (select oid, label, rowid rwid,
3 row_number() over (partition by oid order by label) rn
4 from tab1
5 ),
6 t2 as (select oid, id, rowid rwid,
7 row_number() over (partition by oid order by id) rn
8 from tab2
9 )
10 select b.oid, b.id, a.label
11 from t1 a join t2 b on a.oid = b.oid and a.rn = b.rn;
OID ID LAB
---------- ---------- ---
1 2347 mx1
1 5678 mx2
1 9687 mx3
2 3455 mx4
2 4567 mx5
SQL>
A few options I tried: correlated update won't work because of
ORA-01779: cannot modify a column which maps to a non key-preserved table
SQL> update (
2 with
3 t1 as (select oid, label, rowid rwid,
4 row_number() over (partition by oid order by label) rn
5 from tab1
6 ),
7 t2 as (select oid, id, rowid rwid, label,
8 row_number() over (partition by oid order by id) rn
9 from tab2
10 )
11 select b.oid, b.id, b.label b_label, a.label a_label
12 from t1 a join t2 b on a.oid = b.oid and a.rn = b.rn
13 )
14 set b_label = a_label;
set b_label = a_label
*
ERROR at line 14:
ORA-01779: cannot modify a column which maps to a non key-preserved table
SQL>
MERGE won't work because of
ORA-01732: data manipulation operation not legal on this view
SQL> merge into
2 (select oid, id, label, row_Number() over (partition by oid order by id ) rn from tab2) b
3 using (select oid, label, row_number() over (partition by oid order by label) rn from tab1) a
4 on (a.oid = b.oid and
5 a.rn = b.rn)
6 when matched then update set
7 b.label = a.label;
(select oid, id, label, row_Number() over (partition by oid order by id ) rn from tab2) b
*
ERROR at line 2:
ORA-01732: data manipulation operation not legal on this view
SQL>
Merge would accept a view (created with create view ...), but a view has to be updateable; this one can't be because it contains analytic function.
What's left is a PL/SQL procedure:
SQL> begin
2 for cur_r in (with
3 t1 as (select oid, label, rowid rwid,
4 row_number() over (partition by oid order by label) rn
5 from tab1
6 ),
7 t2 as (select oid, id, rowid rwid,
8 row_number() over (partition by oid order by id) rn
9 from tab2
10 )
11 select b.rwid, a.label
12 from t1 a join t2 b on a.oid = b.oid and a.rn = b.rn
13 )
14 loop
15 update tab2 b set
16 b.label = cur_r.label
17 where b.rowid = cur_r.rwid;
18 end loop;
19 end;
20 /
PL/SQL procedure successfully completed.
SQL> select * from tab2 order by oid, id;
OID ID LAB
---------- ---------- ---
1 2347 mx1
1 5678 mx2
1 9687 mx3
2 3455 mx4
2 4567 mx5
SQL>
Maybe someone has another idea; I'd like to see it & learn something new.
Currently, I have a table MY_TABLE like below:
ID ACCT_TYPE CREDIT_AMT DEBIT_AMT
-- --------- ---------- ---------
1 CDT_01 4 (null)
1 DBT_01 (null) 6
One ID can have multiple ACCT_TYPE like above, and each type has its own amount.
I want to just select the row which has ACCT_TYPE like 'CDT_%' but also the total_amount column which is the total of credit_amt and debit_amt column for the same ID.
My expected output like below:
ID ACCT_TYPE TOTAL_AMT
-- --------- ---------
1 CDT_01 10
I tried with this select statement below but it's no use, I think it's because of different ACCT_TYPE:
Select ID, ACCT_TYPE, SUM(NVL(CREDIT_AMT, 0) + NVL(DEBIT_AMT, 0)) TOTAL_AMT
FROM MY_TABLE WHERE ACCT_TYPE LIKE 'CDT_%' GROUP BY ID, ACCT_TYPE;
Here is the output of the select statement above:
ID ACCT_TYPE TOTAL_AMT
-- --------- ---------
1 CDT_01 4
I just begin to learn some query so I don't know is it really possible to get my expected output.
One way to do it is like below:
with inputs (ID, ACCT_TYPE, CREDIT_AMT, DEBIT_AMT) as
(
select 1, 'CDT_01', 4, null from dual union all
select 1, 'DBT_01', null, 6 from dual
),
prep as
(
select t.*, sum(nvl(credit_amt,0)) over (partition by id) + sum(nvl(debit_amt,0)) over (partition by id) as sum_per_id
from inputs t
)
select id, acct_type, sum_per_id
from prep
where acct_type like 'CDT_%';
Output:
A correlated subquery might be one option; sample data (thank you, #Ranagal) in lines #1 - 5; query that does the job begins at line #6.
SQL> with inputs (ID, ACCT_TYPE, CREDIT_AMT, DEBIT_AMT) as
2 (
3 select 1, 'CDT_01', 4, null from dual union all
4 select 1, 'DBT_01', null, 6 from dual
5 )
6 select a.id,
7 a.acct_type,
8 (select sum(nvl(b.credit_amt, 0)) +
9 sum(nvl(b.debit_amt , 0))
10 from inputs b
11 where b.id = a.id
12 ) total_amt
13 from inputs a
14 where acct_type like 'CDT%';
ID ACCT_T TOTAL_AMT
---------- ------ ----------
1 CDT_01 10
SQL>
With this, I mean Inserting millions of records in tables. I know how to insert data using loops but for inserting millions of data it won't be a good approach.
I have two tables
CREATE TABLE test1
(
col1 NUMBER,
valu VARCHAR2(30),
created_Date DATE,
CONSTRAINT pk_test1 PRIMARY KEY (col1)
)
/
CREATE TABLE test2
(
col2 NUMBER,
fk_col1 NUMBER,
valu VARCHAR2(30),
modified_Date DATE,
CONSTRAINT pk_test2 PRIMARY KEY (col2),
FOREIGN KEY (fk_col1) REFERENCES test1(col1)
)
/
Please suggest a way to insert some dummy records upto 1 million without loops.
As a fairly simplistic approach, which may be enough for you based on your comments, you can generate dummy data using a hierarchical query. Here I'm using bind variables to control how many are created, and to make some of the logic slightly clearer, but you could use literals instead.
First, parent rows:
var parent_rows number;
var avg_children_per_parent number;
exec :parent_rows := 5;
exec :avg_children_per_parent := 3;
-- create dummy parent rows
insert into test1 (col1, valu, created_date)
select level,
dbms_random.string('a', dbms_random.value(1, 30)),
trunc(sysdate) - dbms_random.value(1, 365)
from dual
connect by level <= :parent_rows;
which might generate rows like:
COL1 VALU CREATED_DA
---------- ------------------------------ ----------
1 rYzJBVI 2016-11-14
2 KmSWXfZJ 2017-01-20
3 dFSTvVsYrCqVm 2016-07-19
4 iaHNv 2016-11-08
5 AvAxDiWepPeONGNQYA 2017-01-20
Then child rows, which a random fk_col1 in the range generated for the parent:
-- create dummy child rows
insert into test2 (col2, fk_col1, valu, modified_date)
select level,
round(dbms_random.value(1, :parent_rows)),
dbms_random.string('a', dbms_random.value(1, 30)),
trunc(sysdate) - dbms_random.value(1, 365)
from dual
connect by level <= :parent_rows * :avg_children_per_parent;
which might generate:
select * from test2;
COL2 FK_COL1 VALU MODIFIED_D
---------- ---------- ------------------------------ ----------
1 2 AqRUtekaopFQdCWBSA 2016-06-30
2 4 QEczvejfTrwFw 2016-09-23
3 4 heWMjFshkPZNyNWVQG 2017-02-19
4 4 EYybXtlaFHkAYeknhCBTBMusGAkx 2016-03-18
5 4 ZNdJBQxKKARlnExluZWkHMgoKY 2016-06-21
6 3 meASktCpcuyi 2016-10-01
7 4 FKgmf 2016-09-13
8 3 JZhk 2016-06-01
9 2 VCcKdlLnchrjctJrMXNb 2016-05-01
10 5 ddL 2016-11-27
11 4 wbX 2016-04-20
12 1 bTfa 2016-06-11
13 4 QP 2016-08-25
14 3 RgmIahPL 2016-03-04
15 2 vhinLUmwLwZjczYdrPbQvJxU 2016-12-05
where the number of children varies for each parent:
select fk_col1, count(*) from test2 group by fk_col1 order by fk_col1;
FK_COL1 COUNT(*)
---------- ----------
1 1
2 3
3 3
4 7
5 1
To insert a million rows instead, just change the bind variables.
If you needed more of a relationship between the children and parents, e.g. so the modified date is always after the created date, you can modify the query; for example:
insert into test2 (col2, fk_col1, valu, modified_date)
select *
from (
select level,
round(dbms_random.value(1, :parent_rows)) as fk_col1,
dbms_random.string('a', dbms_random.value(1, 30)),
trunc(sysdate) - dbms_random.value(1, 365) as modified_date
from dual
connect by level <= :parent_rows * :avg_children_per_parent
) t2
where not exists (
select null from test1 t1
where t1.col1 = t2.fk_col1 and t1.created_date > t2.modified_date
);
You may also want non-midnight times (I set everything to midnight via the trunc() call, based on the column name being 'date' not 'datetime'), or some column values null; so this might just be a starting point for you.
I have an Oracle table I've compiled using an Informatica workflow. It's failing an integrity check because the following queries return a different number of rows:
SELECT DISTINCT * FROM table // 4,000 rows
SELECT * FROM table // 4,006 rows
The table consists of 17 fields, none of which are unique keys (obviously). How can I find the 6 duplicate rows?
For returning duplicate rows.
select * from
(SELECT cd.*,
ROW_NUMBER ()
OVER (PARTITION BY column1,column2...column2
ORDER BY column_names)
seq_no
FROM table cd)
where seq_no>1;
For example i have create one sample_table below for your better understanding.
create table sample_table
(
id1 number,
id2 number
)
i have inserted below data into table
ID1 ID2
1 2
1 2
1 2
2 3
2 3
2 3
In above data set we have 6 rows but only two rows are distinct.
By using below queries we can get distinct rows and non-distinct rows.
SELECT cd.*,
ROW_NUMBER ()
OVER (PARTITION BY id1
ORDER BY id1)
seq_no
FROM sample_table cd
after partition the table with the help of id1 we will get the below results
ID1 ID2 SEQ_NO
1 2 1
1 2 2
1 2 3
2 3 1
2 3 2
2 3 3
Then if you want to see the distinct rows use below query
select * from
(SELECT cd.*,
ROW_NUMBER ()
OVER (PARTITION BY id1
ORDER BY id1)
seq_no
FROM sample_table cd)
where seq_no=1;
if you want to see duplicate set use below query
select * from
(SELECT cd.*,
ROW_NUMBER ()
OVER (PARTITION BY id1
ORDER BY id1)
seq_no
FROM sample_table cd)
where seq_no>1;
A posibiliy is to use a analytical function to count the rows in the same group and I don't see how you can write the query without writing all the columns in some clause:
select *
from (
Select a.*, count(*) over (partition by column1, column2, ..., column17) as cnt
from your_table a
)
where cnt>1
This should get 12 rows, because 6 are duplicated.
A basic sql query would be:
select col1, col2, ..., col17
from table
group by col1, col2, ..., col17
having count(*) > 1;