Let me try to explain the scenario. I have two tables A (Columns - A1, A2, A3) & B (Columns - B1, B2, B3). I need to join table A with table B on A1.B2. For every join, table B has one or two records with different values for B3(X or Y). I wanna write one query where the JOIN query needs to pick the row with B3=X(if there's no other row with B3=Y); If two rows exists (B3=X & B3=Y), then the query needs to pick only the row with B3=Y (ignoring the row with B3=X).
Let me try to give some values to the tables & explain a little bit more.
Table A
********
A1 A2 A3
1 11 111
2 22 222
3 33 333
4 44 444
Table B
********
B1 B2 B3
6 1 X
7 1 Y
8 2 X
9 3 X
10 3 Y
11 4 X
Again.. JOIN is on A1.B2. The result should be as following,
JOIN Results
*************
A1 A2 A3 B1 B2 B3
1 11 111 7 1 Y
2 22 222 8 2 X
3 33 333 10 3 Y
4 44 444 11 4 X
Let me know if you guys have any clarification about my question.
Thanks in advance.
Yogi
You can pick the rows from table B with the ROW_NUMBER function if you partition by the join column and order by your "picking order" column:
SELECT b1, b2, b3,
ROW_NUMBER() OVER (PARTITION BY b2 ORDER BY b3 DESC) as rn
FROM b;
1 Y 1
1 X 2
2 X 1
3 Y 1
3 X 2
4 X 1
Then you can filter the first row, the one with rn=1:
SELECT b1, b2, b3
FROM (SELECT b1, b2, b3,
ROW_NUMBER() OVER (PARTITION BY b2 ORDER BY b3 DESC) as rn
FROM b)
WHERE rn=1;
7 1 Y
8 2 X
10 3 Y
11 4 X
The filtered rows can then be joined to table a:
SELECT *
FROM a
JOIN (
SELECT b1, b2, b3
FROM (SELECT b1, b2, b3,
ROW_NUMBER() OVER (PARTITION BY b2 ORDER BY b3 DESC) as rn
FROM b
)
WHERE rn=1
) bfilter ON a.a1 = bfilter.b2;
1 11 111 7 1 Y
2 22 222 8 2 X
3 33 333 10 3 Y
4 44 444 11 4 X
If 'X' and 'Y' are not actual values, you can extend the ORDER clause with a CASE statement to allow for general values:
ROW_NUMBER() OVER (PARTITION BY b2 ORDER BY
CASE b3 WHEN 'Y' THEN 1
WHEN 'X' THEN 2
...
END ASC)
Edit:
SELECT a1, a2, a3, b1, b2, b3
FROM (
SELECT a1, a2, a3, b1, b2, b3,
ROWNUMBER() OVER (PARTITION BY a1 ORDER BY
CASE WHEN a2=... AND b3=... THEN 1
WHEN a2=... AND b3=... THEN 2
...
END ASC)
FROM a JOIN b ON a.a1 = b.b2
)
WHERE rn = 1;
You can use left outer joins as follows
select A.A1, A.A2, A.A3,
nvl(BT1.B1, BT2.B1),
nvl(BT1.B2, BT2.B2),
nvl(BT1.B3, BT2.B3) from A
left outer join B BT1 on A.A1 = BT1.B2 and BT1.B3 = 'Y'
left outer join B BT2 on A.A1 = BT2.B2 and BT2.B3 = 'X'
A good explanation of the various joins is at http://www.codinghorror.com/blog/2007/10/a-visual-explanation-of-sql-joins.html
Here is, how I would do it:
Make the join
group by B2
take the max(B3)
That way you ensure that X is only picked, when there is no alphabetically higher value (Y) available
With UNION
select a.*,b.* from a,b
where a.a1=b.b2
and b.b3='Y'
union
select a.*,b.* from a,b
where a.a1=b.b2
and not exists (select bb.br from b bb where bb.b2=a.a1 and bb.b3='Y')
Without UNION
select a.*,b.* from a,b
where a.a1=b.b2
and (b.b3='Y'
or not exists (select bb.b3 from b bb where bb.b2=a.a1 and bb.b3='Y'))
The constraint here is that B has exactly 1 or 2 rows for each A's row
Related
Suppose if I have table1 as following
Category Brand Value
A A1 4
B B1 7
C C1 8
A A2 3
B B2 4
C C2 6
A A3 9
B B3 10
C C3 1
A A4 5
Now if I want to calculate rank for each brand but grouped by category how do I go about it?
Something like
Select rank() (over value)
from table
group by category
Expected output is this:
Category Brand Value Rank
A A3 9 1
A A4 5 2
A A1 4 3
A A2 3 4
B B3 10 1
B B1 7 2
B B2 4 3
C C1 8 1
C C2 6 2
C C3 1 3
Maybe you are looking for something like this.
See this official documentation on DENSE_RANK for more details
select brand, category, dense_rank() over(partition by category order by value desc) as dr
from table
You may add a PARTITION BY clause to your RANK() call, specifying the category as the partition.
SELECT RANK() OVER (PARTITION BY category ORDER BY value) rnk
FROM yourTable
ORDER BY category, rnk;
Suppose I have a table a with columns A1,A2,A3,A4 and table b with columns B1,B2,B3,B4.
I want to find a records which are having different values in column A1,A2 and B1,B2
Ex.
A1 A2 A3 A4 B1 B2 B3 B4
12 10 10 12 12 10 10 12
14 14 10 12 15 10 10 12
15 10 10 10 15 10 10 10
IT SHOULD RETURN
14 14 10 10
I tried:
SELECT A1,A2
FROM A
EXCEPT
SELECT B1,B2
FROM B;
However, it returned on A1,A2 columns instead of all columns
use left join
select a.* from tableA a
left join tbaleB b
on a.A1=b.B1 and a.A2=b.B2 and a.A3=b.B3 and a.A4=b.B4
where b.B1 is null
You can try below using left join
select * from tableA
left join tableB on A1=B1 and A2=B2
where B1 is null and B2 is null
I would use not exists:
select a.*
from a
where not exists (select 1
from b
where a.a1 = b.b1 and a.a2 = b.b2
);
I have tables like that: (C1-C2 varchar(10), C3-Number int)
WaitingData
C1 C2 C3 Number
A B 1 10
A B 2 0
A B 3 4
X B 4 2
CompletedData
C1 C2 C3 Number
A B 1 5
A B 2 2
A B 3 0
X B 4 12
I am using the query below to represent the data:
Select wd.C1,wd.C2,wd.C3,wd.Number as NW,cdd.Number as NC
into #AllData
from (Select C1,C2,C3,sum(Number) from WaitingData group by C1,C2,C3) wd
outer apply (Select C1,C2,C3,sum(Number)
from CompletedData cd
where wd.C1=cd.C1 and wd.C2=cd.C2 and wd.C3=cd.C3
) cdd
Select * from #AllData
union
Select C1='Total',C2='Total',C3=-1, sum(NW),sum(NW)
from #AllData
This is giving me an output like:
C1 C2 C3 NW NC
A B 1 10 5
A B 2 0 2
A B 3 4 0
X B 4 2 12
Total Total -1 16 19
However, I want to hide the rows that has no NW but calculate its regarding values while calculating the Total row (see NC below). The output I want is like:
C1 C2 C3 NW NC
A B 1 10 5
A B 3 4 0
X B 4 2 12
Total Total -1 16 19
I could not find a way to provide an output like this. Any help would be so appreciated!
------------------------------EDIT---------------------------------------
------------------------------EDIT---------------------------------------
When I have data in the tables like below, the outer apply is not working like I want, it does not include the data A B 2.
WaitingData
C1 C2 C3 Number
A B 1 10
A B 3 4
X B 4 2
CompletedData
C1 C2 C3 Number
A B 1 5
A B 2 2
X B 4 12
And the output would be like:
C1 C2 C3 NW NC
A B 1 10 5
A B 3 4 NULL
X B 4 2 12
Total Total -1 16 17
In this situation, what can I do to count "2" NC value having by A B 2 on the final result and see NC as 19 instead 17, except inserting all the records that included by CompletedData but WaitingData? (need an efficient way)
Wrap the final result with one more select and exclude rows where NW = 0.
select * from
(
Select * from #AllData
union
Select C1='Total',C2='Total',C3=-1, sum(NW),sum(NC)
from #AllData
) t
where NW <> 0
Edit: Using a full join to get all values from both tables.
with t as
(select coalesce(w.c1,c.c1) as c1,coalesce(w.c2,c.c2) as c2,coalesce(w.c3,c.c3) as c3
, coalesce(w.number,0) as nw , coalesce(c.number,0) as nc
from waitingdata w
full join completeddata c on w.c1 = c.c1 and w.c2=c.c2 and w.c3=c.c3)
select * from
(select * from t
union all
Select C1='Total',C2='Total',C3=-1, sum(NW),sum(NC)
from t) x where nw <> 0
You can do all of this in one query, without temporary tables, intermediate results, subqueries, or UNION by using the ROLLUP operator:
SELECT
WD.C1,
WD.C2,
WD.C3,
SUM(WD.Number) AS NW,
SUM(CD.Number) AS NC
FROM
dbo.WaitingData WD
LEFT OUTER JOIN CompletedData CD ON
CD.C1 = WD.C1 AND
CD.C2 = WD.C2 AND
CD.C3 = WD.C3
GROUP BY
WD.C1,
WD.C2,
WD.C3
WITH ROLLUP
HAVING
GROUPING_ID(WD.C1, WD.C2, WD.C3) IN (0, 7) AND
SUM(WD.Number) <> 0
I have one table1
c1 c2
a 1
a 2
a 3
b 1
b 2
b 3
I have one table2
c3 c4
A I
B II
C III
If Cross Join Table1 & Table2 into Table3
c1 c2 c3 c4
a 1 A I
a 2 A I
a 3 A I
b 1 A I
b 2 A I
b 3 A I
a 1 B II
a 2 B II
a 3 B II
b 1 B II
b 2 B II
b 3 B II
a 1 C III
a 2 C III
a 3 C III
b 1 C III
b 2 C III
b 3 C III
If you add one column in Table1 and cross join with Table2 and insert into Table3
the order of Table3 is missing.
So my question is, how to maintain the order based on the Table1?
Don't worry about the order of inserts. You should include whichever field you want to order by in the insert and use order by when selecting the data.
Try this -
select *
from table3
order by c1, c2
That's as good as your going to get with your current schema.
Below is the sample data:
c1 c2 c3 c4 c5
1 a1 a 1 1
2 a2 a 2 1
3 a3 a 3 1
4 a4 a 4 1
5 b1 b 1 1
6 b2 b 2 1
7 b3 b 3 1
8 b4 b 4 1
9 a1 c 3 1
I want to get the the below details:
c1 c2 c3 c4 c5
1 a1 a 1 1
5 b1 b 1 1
9 a1 c 3 1
C1 is primary key, the criteria is for any given unique(c2) where c4 is the lowest, I want to return the contents(all the 5 columns) of the row.
Try this:
SELECT t1.*
FROM Table1 t1
INNER JOIN
(
SELECT c3, MIN(c4) c4
FROM Table1
GROUP BY c3
) t2 ON t1.c3 = t2.c3 ANd t1.c4 = t2.c4
SQL Fiddle Demo
Update:1 In SQL the returned results is a set set(unless you specify an ORDER BY clause, it is a cursor in this case), wherein the order is not guaranteed. This is a standard. You should use an ORDER BY clause if you want to guarantee a specific order. In your case , the results is not guaranteed to be ordered like 1 5 9. Add ORDER BY c1 instead.
The ORDER BY clause might be crucial in some cases, for example, if want to get the top three rows, or the maximum one, in this case you have to specify an ORDER BY clause.
So if you wants to persist a specific order the you have specify an ORDER BY.
1 As noted by #Fahim Parker, see the comments below.
select c1,c2,c3,c4,c5
from table
where c4= (select min(c4) from table as f where f.c4 = table.c4);
i hope that helps