Top 2 of each group ordered SQL Server - sql-server

I have 8 records
Name Value Product
--------------------------
Abraham 4 A
Lincoln 6 B
Abraham 4 C
Lincoln 2 D
Lincoln 3 E
Lincoln 2 F
Abraham 1 G
Abraham 9 H
Abraham has 4 records and Lincoln too.
I need, from the SQL, the top 2 values from Abraham and the top 2 values from Lincoln
I've tried:
SELECT TOP 2 WITH TIES
NAME,
VALUE,
PRODUCT
FROM
blabla
JOIN
blabla...
ORDER BY
NAME
This takes 2 of each name, but not the most valuables, because I'm not setting a VALUE order by desc.
But I can't put the VALUE order by desc because of the top with ties..
What I need is something like making my TOP WITH TIES works only for ORDER BY NAME (if there is a way to do this, like restricting the top with ties to the first order by), but I need only the 2 max values of them.
The final result I need:
Abraham 9 H
Abraham 4 C
Lincoln 6 B
Lincoln 3 E
PS: This is only a simulation of what I want, the original query has like over 100 lines with union and stuff, so i think it was better to simplify.

You can do this with ROW_NUMBER and partitioning.
select Name
, Value
, Product
from
(
select Name
, Value
, Product
, RowNum = ROW_NUMBER() over(partition by Name order by value desc)
from SomeTable
) x
where x.RowNum <= 2
order by x.Name
, x.Value desc

A solution is a CROSS APPLY with TOP and ORDER BY. You start with the list of names and call a "function" that returns the TOP for each name.
SELECT
N.Name,
V.Value,
V.Product
FROM
(SELECT DISTINCT Y.Name FROM YourTable AS Y) AS N
CROSS APPLY (
SELECT TOP 2 WITH TIES
P.Value,
P.Product
FROM
YourTable AS P
WHERE
N.Name = P.Name
ORDER BY
P.Value DESC
) AS V
Keep in mind that using WITH TIES can make the TOP return more than the supplied number of rows (2 for your example), in case that there are ties of values.
If you need to show names which don't have products (in this example it won't work because both come from the same table), you can switch the CROSS APPLY for an OUTER APPLY, which behaves similar to a LEFT JOIN as it will return NULL values (but not 2 rows!).

Related

SQL Server 2008 Is it Possible to Have Select Top Return Nulls

(Select top 1 pvd.Code from PatientVisitDiags pvd
where pvd.PatientVisitId = pv.PatientVisitId
Order By pvd.Listorder) as "DX1",
(Select top 1 a.code from (Select top 2 pvd.Code,pvd.ListOrder from PatientVisitDiags pvd
where pvd.PatientVisitId = pv.PatientVisitId
Order By pvd.Listorder)a order by a.ListOrder DESC ) as "DX2",
(Select top 1 a.code from (Select top 3 pvd.Code,pvd.ListOrder from PatientVisitDiags pvd
where pvd.PatientVisitId = pv.PatientVisitId
Order By pvd.Listorder)a order by a.ListOrder DESC ) as "DX3",
(Select top 1 a.code from (Select top 4 pvd.Code,pvd.ListOrder from PatientVisitDiags pvd
where pvd.PatientVisitId = pv.PatientVisitId
Order By pvd.Listorder)a order by a.ListOrder DESC ) as "DX4",
(Select top 1 a.code from (Select top 5 pvd.Code,pvd.ListOrder from PatientVisitDiags pvd
where pvd.PatientVisitId = pv.PatientVisitId
Order By pvd.Listorder)a order by a.ListOrder DESC ) as "DX5"
The above code is what I am using currently (It is not optimal but is only being used once for a one time Data Export).
In the database that we are currently exporting from, there is a table PatientVisitDiags that has columns "ListOrder" and "Code". There can be between 1 and 5 codes. The ListOrder holds the number of that code. For example:
ListOrder|Code |
1 |M51.27 |
2 |M54.17 |
3 |G83.4 |
I am trying to export the Code to its corresponding Column in the new table(DX1,DX2..etc). If I sort by ListOrder I can get them in the order I need (Row 1 to DX1 | Row 2 to DX2 etc.) However when I run the above SQL code, If the source table only has 3 Codes DX4 and DX5 will repeat DX3. For Example:
DX1 |DX2 |DX3 |DX4 |DX5
M51.27 |M54.17 |G83.4 |G83.4 |G83.4
Is there a way to have TOP return NULL values if you Select TOP more than what is given? SQL Sever 2008 does not allow for OFFSET/FETCH, this is what I normally would have done given the option to select individual rows.
TL:DR
ID | Name
1 | Joe
2 | Eric
3 | Steve
4 | John
If I have a table like above and run
SELECT TOP 5 Name FROM Table
Is there anyway to return?
Joe
Eric
Steve
John
NULL
What you're really doing is pivoting. So pivot! Try this little query:
WITH Top5 AS (
SELECT TOP 5
Dx = 'DX' + Convert(varchar(11), Row_Number() OVER (ORDER BY pvd.Listorder)),
pvd.Code
FROM dbo.PatientVisitDiags pvd
WHERE pvd.PatientVisitId = #patientVisitId
)
SELECT *
FROM
Top5 t
PIVOT (Max(Code) FOR Dx IN (DX1, DX2, DX3, DX4, DX5)) p
;
To answer your second question about getting an unpivoted rowset, basically do the same thing but provide the 5 rows somehow and left join to the desired data.
WITH Data AS (
SELECT TOP 5
Seq = Row_Number() OVER(ORDER BY ID),
Name
FROM dbo.Table
ORDER BY ID
)
SELECT
n.Seq,
t.Name
FROM
(VALUES
(1), (2), (3), (4), (5) -- or a numbers-generating CTE perhaps
) n (Seq)
LEFT JOIN Top 5 t
ON n.Seq = t.Seq
;
Side note
The fact that you're doing this:
where pvd.PatientVisitId = pv.PatientVisitId
tells me you're not using ANSI joins. Stop. Don't do that any more. Put this join condition in the ON clause of a JOIN. It's the year 2016... why are you using join syntax from the last century?
Oh, and prefix the schema on the table names. Look it up--you'll find actual performance reasons why you should do that. It's not just about the time taken to find the correct schema, but also about the execution plan cache...
one at a time - answering the last question
create a table with a bunch of null
select top (5) col
from
(
select col from table1
union
select nulCol from nullTable
) tt
order by tt.col

how to do a FOR EACH loop in T-Sql using join?

I have a table PRODUCT that is basically set up so there is an PRODUCTID, PRODUCTNAME... it looks sort of like this
PRODUCTID PRODUCTNAME
100 PNAME1
101 PNAME2
102 PNAME3
Now I have to insert a record into new table PRODUCTMAPPING for each row in the PRODUCT.
so my new table PRODUCTMAPPING should look like this
PRODUCTMAPPINGID PRODUCTID
1 100
1 101
1 102
2 100
2 101
2 102
3 100
and so on ....
I tried doing while but I need it using Join.
Can we acheive this using joins ?
Thanks in advance.
One way;
select
row_number() over(partition by a.PRODUCTID order by a.PRODUCTID) PRODUCTMAPPINGID,
a.PRODUCTID
from PRODUCT a, PRODUCT b
Using LOOP
The following example specifies that the JOIN operation in the query is performed by a LOOP join.
Select sp.Name,spqh.quota
FROM Sales.SalesPersonQuotaHistory AS spqh
INNER LOOP JOIN Sales.SalesPerson AS sp
ON spqh.SalesPersonID = sp.SalesPersonID
WHERE sp.SalesYTD > 2500000.00;
GO
Refer this MSDN link
INSERT
INTO dbo.PRODUCTMAPPING
(
PRODUCTMAPPINGID
,PRODUCTID
)
SELECT pmv.PRODUCTMAPPINGID
,p.PRODUCTID
FROM dbo.Product p
CROSS JOIN
(
SELECT pm.ProductMappingID
FROM dbo.ProductMappingValues pmv -- WHERE DO THESE COME FROM?
) pmv

SQL Select Count Statement

I have two tables, one for Employees and the second for records. I want to get total entries for each employee and order the results by max total entry like:
Daniel 3
David 1
tblEmployee:
EID Name
1 Daniel
2 David
tblEntry:
ID Column1 EID
1 XX 1
2 XX 1
3 XX 2
4 XX 1
try this:
select emp.EID,emp.Name,COUNT(etr.EID)
as total_entries from Employee emp join Entry etr
on emp.EID=etr.EID
group by emp.EID,emp.Name
You must use group by
select count(*) from tblEmployee ee, tblEntry e where ee.eid = e.eid group by ee.name
There are several variations on this, and you don't say what version of SQL Server you're using, but I like doing it this way:
;
using A
as (
select EID
, RecordCount = COUNT(*)
from tblEntry
group by
EID
)
select a.EID
, e.Name
, a.RecordCount
from A
join tblEmployee e
on A.EID = e.EID
order by
RecordCount desc
I like doing it this way rather than joining and then summarizing because you only have to group on the minimum number of fields. EID in tblEntry is likely to already have an index on it, while Name in tblEmployee may not.

how to select a row based on a single distinct value

If i have 4 rows and want to select rows based on a single column's distinct values and dont mind which info it uses for the rest of the row, how do i do this? There doesn't seem to be a 'distinct' function for single cols whilst maintaining rest of row data.
eg
Name, value
john 1
john 2
michael 3
michael 5
result
john 1
michael 5
note it could have been john 2 or michael 3, i dont care which row for John or Michael it uses for the rest of the values.
GROUP BY Name and use any aggregate function with the value MIN or MAX since you don't care about the value of it:
SELECT Name, MIN(value)
FROM table
GROUP BY Name
Try this
select a.* from TAbleName a
inner join
(
select distinct name,min(Id) as id from TAbleName
group by name
) as b
on a.name= b.name
and a.id=b.id

Get union of two table and taking data with a condition

I have two tables
table-a
id name
100 asd
101 ass
102 gdd
103 hgf
104 cvd
105 erf
table-b
id filter
100 red
101 blue
100 green
100 yellow
102 black
102 red
103 dark
Table-a is the master table and that have all the id's.but Table two is the one which has 'filter' data.
from these two table I want to find out all those 'id's which does not have minimum 2 filters.
note that table-b does not have all the itemnumbers in table-a, and i want all that itemnumber irrespective of if that is in table-a or table-b.I have tried inner joining these two tables and getting data out but nothing worked.
Select A.ID, A.Name, count(*)
from tableA A
LEFT JOIN tableB B on A.ID = B.ID
Group By A.ID, A.name
having count(*) <= 1
LEFT JOIN gives all records from A and only those in B which match.
The group by ID and name let us count the number of filters found in
each
The having says give me any items with a count less than or
equal to 1. (or less than the minimum 2)
Thus results would be.
101 ass 1
103 hgf 1
104 cvd 0
105 erf 0
select
*
from
table-a a
left join (
select id, count(id) as c from table-b group by id
) v on a.id = v.id
where isnull(v.id, 0) < 2
I think this would work in SQL Server (tested in SQLite and usually the two are fairly compatible when it comes to inline view syntax). But syntax issues aside, inline views can make working with sets easier to visualize.
select TA.id, name
from TA
inner join
(
select id from TA
where not exists (select id from TB where TA.id = TB.id)
UNION
select id from TB
group by id having count(filter) < 2
) as FOO
on TA.id = FOO.id
The default behavior of UNION is to remove duplicates.
The first UNIONed set consists of the ids from table A that have no filter (no counterpart in the filters table B).
The second UNIONed set consists of the ids from the filters table, table B, that have only 1 filter.
We inner join those unioned sets back to Table A to get the entity Name.

Resources