I am dealing with a huge list of columns (around 50) where i only need to group by one column. Is there anyway in SQL Server i can aggregate the columns by something such as
SELECT MAX(*)
FROM View1
GROUP BY Column1
instead of having to go through each one and specify an aggregate function. I have had a look online but cant find anything. Is there any advice or guidance someone can give me or is it just a case of going through each row?
Thanks
You can build query you need using system tables:
DECLARE #ViewName sysname = N'View1',
#query nvarchar(max),
#Column sysname = 'Column1'
SET #query = N'SELECT ' + #Column + ',' + CHAR(10)
SELECT #query = #query + N'MAX('+c.[name]+') as '+c.[name]+',' + CHAR(10)
FROM sys.views v
INNER JOIN sys.columns c
ON v.[object_id] = c.[object_id]
WHERE v.[name] = #ViewName AND c.[name] != #Column
SET #query = STUFF(#query,LEN(#query)-1,1,'') + 'FROM '+#ViewName + CHAR(10) + 'GROUP BY ' + #Column
PRINT #query
Output will be:
SELECT Column1,
MAX(Column2) as Column2,
MAX(Column3) as Column3,
...
MAX(ColumnN) as ColumnN
FROM View1
GROUP BY Column1
You can Ctrl+C Ctrl+V on new query window and execute, or execute it right here with:
EXEC (#query)
In case of tables - you need to use sys.tables
In case if view or table is not in default schema - you need to specify it manually.
SET #query = STUFF(#query,LEN(#query)-1,1,'') + 'FROM dbo.'+#ViewName + CHAR(10) + 'GROUP BY ' + #Column
No. Unfortunately, you're going to have to write out the columns.
If you need max value from different columns you can try as below
Select max(yourcolumn) from
(
Select col1 from yourtable
union all Select col2 from yourtable
union all Select col3 from yourtable
...
) a
No, you can not use max(*).
You will have to give a column name in max function like below
select max(column_name) from table_name;
You cannot write select max(*). This results in the error Incorrect syntax near '*'. Instead, you will need to specify the columns.
One way to get the maximum of multiple columns is to unpivot the table. An efficient way to do this is to use cross apply to generate separate row values for each column.
For example, the code below finds the maximum value across 3 different columns in all of the rows:
declare #test table
(
id int primary key clustered,
value1 int,
value2 int,
value3 int
)
insert into #test (id, value1, value2, value3)
values (1, 100, 0, 0), (2, 0, 50, 0), (3, 0, 0, 25)
select max(TestValue) -- returns 100
from #test
cross apply
(
values(value1),(value2), (value3)
) TestValues (TestValue)
Related
I generate comma seperated string and add single quite to each numbers
Here is how i do it
DECLARE #IDs NVARCHAR(max)
SELECT #IDs = COALESCE(#IDs +',', '') + ''''
+ Cast([mynos] AS NVARCHAR(255)) + ''''
FROM mytable
WHERE id = 22
If i print variable #IDs then i get below output
'78888','3333','1222'
When i use same variable in this query then query doesnt return any value
SELECT *
FROM table1
WHERE ids IN ( #IDs )
How to fix this?
It doesn't work as your query is effectively doing this:
SELECT *
FROM TABLE
WHERE Ids IN ('''78888'',''3333',''1222''');
Which would also be equivalent to:
SELECT *
FROM TABLE
WHERE Ids = '''78888'',''3333',''1222''';
If you want to do the query as you have done, you'll need to split your delomited data out again. As you're using SQL Server 2012, you can't make use of STRING_SPLIT, so you'll need to a different one; such as Jeff Moden's DelimitedSplit8K. Then you can do:
SELECT *
FROM TABLE
WHERE IDs IN (SELECT items
FROM dbo.DelimitedSplit8K (#IDs,','));
However, why are you not simply doing...
SELECT *
FROM TABLE T
WHERE EXISTS (SELECT 1
FROM myTable mT
WHERE mT.Id = 22
AND mT.myNos = T.Ids);
You can use dynamic query #id is string variable not multi-value argument
DECLARE #IDs nVARCHAR(MAX)
SELECT #IDs = COALESCE(#IDs +',' ,'') + '''' + CAST([myNos] AS nVARCHAR(255)) + ''''
FROM myTable WHERE Id = 22
DECLARE #query nVARCHAR(MAX)
SET #query = "Select * from table1 where Ids in ("+#IDs+")"
EXECUTE sp_executesql #query
I tried below and it worked
select * from table1 where id in (select mynos from mytable where id = 22)
Thanks to #Larnu for giving me idea
Is there a way to access columns by their index within a stored procedure in SQL Server?
The purpose is to compute lots of columns. I was reading about cursors, but I do not know how to apply them.
Let me explain my problem:
I have a row like:
field_1 field_2 field_3 field_4 ...field_d Sfield_1 Sfield_2 Sfield_3...Sfield_n
1 2 3 4 d 10 20 30 n
I need to compute something like (field_1*field1) - (Sfield_1* Sfiled_1) / more...
So the result is stored in a table column d times.
So the result is a d column * d row table.
As the number of columns is variable, I was considering making dynamic SQL, getting the names of columns in a string and splitting the ones I need, but this approach makes the problem harder. I thought getting the column number by index could make life easier.
No, you can not use the ordinal (numeric) position in the SELECT clause.
Only in the ORDER BY clause can you use the ordinal position, because it's based on the column(s) specified in the SELECT clause.
First, as OMG Ponies stated, you cannot reference columns by their ordinal position. This is not an accident. The SQL specification is not built for dynamic schema either in DDL or DML.
Given that, I have to wonder why you have your data structured as you do. A sign of a mismatch between schema and the problem domain rears itself when you try to extract information. When queries are incredibly cumbersome to write, it is an indication that the schema does not properly model the domain for which it was designed.
However, be that as it may, given what you have told us, an alternate solution would be something like the following: (I'm assuming that field_1*field1 was meant to be field_1 * field_1 or field_1 squared or Power( field_1, 2 ) )
Select 1 As Sequence, field_1 As [Field], Sfield_1 As [SField], Sfiled_1 As [SFiled]
Union All Select 2, field_2, Sfield_2, Sfiled_2
...
Union All Select n, field_n, Sfield_n, Sfiled_n
Now your query looks like:
With Inputs As
(
Select 1 As Sequence, field_1 As [Field], Sfield_1 As [SField], Sfiled_1 As [SFiled]
Union All Select 2, field_2, Sfield_2, Sfiled_2
....
)
, Results As
(
Select Case
When Sequence = 1 Then Power( [Field], 2 ) - ( [SField] * [SFiled] )
Else 1 / Power( [Field], 2 ) - ( [SField] * [SFiled] )
End
As Result
From Inputs
)
Select Exp( Sum( Log( Result ) ) )
From Results
This might not be the most elegant or efficient but it works. I am using it to create a new table for faster mappings between data that I need to parse through all the columns / rows.
DECLARE #sqlCommand varchar(1000)
DECLARE #columnNames TABLE (colName varchar(64), colIndex int)
DECLARE #TableName varchar(64) = 'YOURTABLE' --Table Name
DECLARE #rowNumber int = 2 -- y axis
DECLARE #colNumber int = 24 -- x axis
DECLARE #myColumnToOrderBy varchar(64) = 'ID' --use primary key
--Store column names in a temp table
INSERT INTO #columnNames (colName, colIndex)
SELECT COL.name AS ColumnName, ROW_NUMBER() OVER (ORDER BY (SELECT 1))
FROM sys.tables AS TAB
INNER JOIN sys.columns AS COL ON COL.object_id = TAB.object_id
WHERE TAB.name = #TableName
ORDER BY COL.column_id;
DECLARE #colName varchar(64)
SELECT #colName = colName FROM #columnNames WHERE colIndex = #colNumber
--Create Dynamic Query to retrieve the x,y coordinates from table
SET #sqlCommand = 'SELECT ' + #colName + ' FROM (SELECT ' + #colName + ', ROW_NUMBER() OVER (ORDER BY ' + #myColumnToOrderBy+ ') AS RowNum FROM ' + #tableName + ') t2 WHERE RowNum = ' + CAST(#rowNumber AS varchar(5))
EXEC(#sqlCommand)
Please help me with this. I am totally stuck. I have coders block or something.
I have the following table
ID Name Cost Included
---- ---------- ------- ----------
1 Package1 10.00 Yes
2 Package2 20.00 No
3 Package3 20.00 Yes
I would like to crosstab this information, to display like the following example,there will be more columns in the table.
Type Package1 Package2 Package3
----- ------------ ----------- ----------
Name Package1 Package2 Package3
Cost 10.00 20.00 30.00
Included Yes No Yes
It seems to me that you are trying to build a product comparison list. If this is true, you might unpivot the table first and then join individual records together.
The 'transponded' part unpivots the columns. All columns must be of compatible types or converted to one. I choose varchar(100). transponded returns table with three columns, ID from ProductInfo, Type as column name and Value as value of corresponding column.
Select part joins together info on as many product as demanded by adding another left join transponded tn on t1.Type = tnType and tn.ID = #parametern. This part seems as a hassle, but when I tried to do this part with pivot I failed to get column in proper order - pivot sorted names in Type. It would however demand dynamic sql generation. This solution is fixed providing that you add enough joins for maximum products you wish to compare at once. I belive it would not be over 5.
=1, =2 and =3 should be replaced by parameters. The query should be hosted in stored procedure.
; with transponded as
(
select ID, Type, Value
from
(
select ID,
Name,
cast (Cost as varchar(100)) Cost,
cast (case when Included = 1 then 'Yes' else 'No' end as varchar(100)) Included
from ProductInfo
) p
unpivot (Value for Type in (Name, Cost, Included) ) a
)
select t1.Type,
t1.Value Product1,
t2.Value Product2,
t3.Value Product3
from transponded t1
left join transponded t2
on t1.Type = t2.Type
and t2.id = 2
left join transponded t3
on t1.Type = t3.Type
and t3.id = 3
where t1.id = 1
In short, transpond one record at time and join to another transponded record by Type column.
Oh, and here is a Sql Fiddle playground.
There is no easy way to do this, as the pivot will need to be aggregated by column. Given that adding columns to the input table would cause a maintenance issue where these values will not be presented to the output until the code is changed wherever it is used, I'd say you're probably best doing it once with a stored procedure, which will dynamically generate the output you're looking for based on the schema of the input table.
I have demonstrated how this can be done, using the data you have supplied. This data is stored in a temp table (not #temp, because the stored proc won't work with temporary tables), populated thus:
CREATE TABLE temp (
_key int,
package_name varchar(50),
cost float,
included bit
)
INSERT INTO temp VALUES(1,'Package1', 10.00, 1)
INSERT INTO temp VALUES(2,'Package2', 20.00, 0)
INSERT INTO temp VALUES(3,'Package3', 20.00, 1)
The stored procedure retrieves a list of values based on the #pivot_field parameter, and uses these values as a column list to be inserted after the "Type" field. It then unions the pivot field and all other fields together to generate the rows, pivoting one column at a time. The procedure is as follows:
CREATE PROCEDURE usp_get_pivot (#table_name nvarchar(255), #pivot_field nvarchar(255)) AS
BEGIN
CREATE TABLE #temp (val nvarchar(max))
DECLARE #sql NVARCHAR(MAX), #cols NVARCHAR(MAX), #col NVARCHAR(255)
SET #sql = 'SELECT DISTINCT ' + #pivot_field + ' FROM ' + #table_name
INSERT INTO #temp EXEC sp_executesql #sql;
SET #cols = (SELECT '[' + val + '],' FROM #temp FOR XML PATH(''))
SET #cols = SUBSTRING(#cols, 1, LEN(#cols)-1)
SET #SQL = N'SELECT ''' + #pivot_field + ''' as [type], *
FROM (SELECT ' + #pivot_field + ', ' + #pivot_field + ' as ' + #pivot_field + '1 FROM ' + #table_name + ') AS source_table
PIVOT (max(' + #pivot_field + '1) FOR ' + #pivot_field + ' IN (' + #cols + ')) AS pivot_table'
DECLARE csr CURSOR FOR
SELECT c.name FROM sys.columns c, sys.objects o
WHERE c.object_id = o.object_id AND o.name = #table_name
AND c.name <> #pivot_field
ORDER BY column_id
OPEN csr
FETCH NEXT FROM csr INTO #col
WHILE ##FETCH_STATUS = 0
BEGIN
SET #sql = #sql + ' UNION ALL
SELECT ''' + #col + ''' as [type], *
FROM (SELECT ' + #pivot_field + ', CAST(' + #col + ' AS VARCHAR) AS ' + #col + ' FROM ' + #table_name + ') AS source_table
PIVOT (max(' + #col + ') FOR ' + #pivot_field + ' IN (' + #cols + ')) AS pivot_table'
FETCH NEXT FROM csr INTO #col
END
CLOSE csr
DEALLOCATE csr
DROP TABLE #temp
EXEC sp_executesql #sql
END
You should be able to simply copy and paste the procedure into management studio, create the data is shown above and execute the procedure with:
EXEC usp_get_pivot 'temp', 'package_name'
If number of packages is not static there is no option for you I think. PIVOT clause can produce only static/defined number of columns.
You may do some table-to-table rewriting using multiple statements - but still you have to face with static number of columns.
But you may set it to for example to 10 and then display up to 10 packages, having NULL-s in rest of columns if there are less packages.
You may also use dynamic SQL to have dynamic number of columns - but it will be a headache.
If you're going to export this data to Excel - do not pivot it at SQL - do a transposition in Excel (it's under "paste special").
Basically what i have at this stage is the following.
SELECT [Type],
MAX(Beginner) AS [Beginner],
MAX(Intermediate) AS [Intermediate],
MAX(Advanced) AS [Advanced]
FROM
(
SELECT
'Name' AS TYPE,
CASE WHEN Name='Beginner' THEN Name END AS [Beginner],
CASE WHEN Name='Intermediate' THEN Name END AS [Intermediate],
CASE WHEN Name='Advanced' THEN Name END AS [Advanced]
FROM Administration.Package
UNION ALL
SELECT
'Price' AS TYPE,
CASE WHEN Name='Beginner' THEN CAST(Price AS VARCHAR) END AS [Beginner],
CASE WHEN Name='Intermediate' THEN CAST(Price AS VARCHAR) END AS [Intermediate],
CASE WHEN Name='Advanced' THEN CAST(Price AS VARCHAR) END AS [Advanced]
FROM Administration.Package
)A
GROUP BY [Type]
But it does not feel right to have the union for each and every column.
Is there a way to access columns by their index within a stored procedure in SQL Server?
The purpose is to compute lots of columns. I was reading about cursors, but I do not know how to apply them.
Let me explain my problem:
I have a row like:
field_1 field_2 field_3 field_4 ...field_d Sfield_1 Sfield_2 Sfield_3...Sfield_n
1 2 3 4 d 10 20 30 n
I need to compute something like (field_1*field1) - (Sfield_1* Sfiled_1) / more...
So the result is stored in a table column d times.
So the result is a d column * d row table.
As the number of columns is variable, I was considering making dynamic SQL, getting the names of columns in a string and splitting the ones I need, but this approach makes the problem harder. I thought getting the column number by index could make life easier.
No, you can not use the ordinal (numeric) position in the SELECT clause.
Only in the ORDER BY clause can you use the ordinal position, because it's based on the column(s) specified in the SELECT clause.
First, as OMG Ponies stated, you cannot reference columns by their ordinal position. This is not an accident. The SQL specification is not built for dynamic schema either in DDL or DML.
Given that, I have to wonder why you have your data structured as you do. A sign of a mismatch between schema and the problem domain rears itself when you try to extract information. When queries are incredibly cumbersome to write, it is an indication that the schema does not properly model the domain for which it was designed.
However, be that as it may, given what you have told us, an alternate solution would be something like the following: (I'm assuming that field_1*field1 was meant to be field_1 * field_1 or field_1 squared or Power( field_1, 2 ) )
Select 1 As Sequence, field_1 As [Field], Sfield_1 As [SField], Sfiled_1 As [SFiled]
Union All Select 2, field_2, Sfield_2, Sfiled_2
...
Union All Select n, field_n, Sfield_n, Sfiled_n
Now your query looks like:
With Inputs As
(
Select 1 As Sequence, field_1 As [Field], Sfield_1 As [SField], Sfiled_1 As [SFiled]
Union All Select 2, field_2, Sfield_2, Sfiled_2
....
)
, Results As
(
Select Case
When Sequence = 1 Then Power( [Field], 2 ) - ( [SField] * [SFiled] )
Else 1 / Power( [Field], 2 ) - ( [SField] * [SFiled] )
End
As Result
From Inputs
)
Select Exp( Sum( Log( Result ) ) )
From Results
This might not be the most elegant or efficient but it works. I am using it to create a new table for faster mappings between data that I need to parse through all the columns / rows.
DECLARE #sqlCommand varchar(1000)
DECLARE #columnNames TABLE (colName varchar(64), colIndex int)
DECLARE #TableName varchar(64) = 'YOURTABLE' --Table Name
DECLARE #rowNumber int = 2 -- y axis
DECLARE #colNumber int = 24 -- x axis
DECLARE #myColumnToOrderBy varchar(64) = 'ID' --use primary key
--Store column names in a temp table
INSERT INTO #columnNames (colName, colIndex)
SELECT COL.name AS ColumnName, ROW_NUMBER() OVER (ORDER BY (SELECT 1))
FROM sys.tables AS TAB
INNER JOIN sys.columns AS COL ON COL.object_id = TAB.object_id
WHERE TAB.name = #TableName
ORDER BY COL.column_id;
DECLARE #colName varchar(64)
SELECT #colName = colName FROM #columnNames WHERE colIndex = #colNumber
--Create Dynamic Query to retrieve the x,y coordinates from table
SET #sqlCommand = 'SELECT ' + #colName + ' FROM (SELECT ' + #colName + ', ROW_NUMBER() OVER (ORDER BY ' + #myColumnToOrderBy+ ') AS RowNum FROM ' + #tableName + ') t2 WHERE RowNum = ' + CAST(#rowNumber AS varchar(5))
EXEC(#sqlCommand)
I need a query that will return a table where each column is the count of distinct values in the columns of another table.
I know how to count the distinct values in one column:
select count(distinct columnA) from table1;
I suppose that I could just make this a really long select clause:
select count(distinct columnA), count(distinct columnB), ... from table1;
but that isn't very elegant and it's hardcoded. I'd prefer something more flexible.
This code should give you all the columns in 'table1' with the respective distinct value count for each one as data.
DECLARE #TableName VarChar (Max) = 'table1'
DECLARE #SqlString VarChar (Max)
set #SqlString = (
SELECT DISTINCT
'SELECT ' +
RIGHT (ColumnList, LEN (ColumnList)-1) +
' FROM ' + Table_Name
FROM INFORMATION_SCHEMA.COLUMNS COL1
CROSS AppLy (
SELECT ', COUNT (DISTINCT [' + COLUMN_NAME + ']) AS ' + '''' + COLUMN_NAME + ''''
FROM INFORMATION_SCHEMA.COLUMNS COL2
WHERE COL1.TABLE_NAME = COL2.TABLE_NAME
FOR XML PATH ('')
) TableColumns (ColumnList)
WHERE
1=1 AND
COL1.TABLE_NAME = #TableName
)
EXECUTE (#SqlString)
try this (sql server 2005 syntax):
DECLARE #YourTable table (col1 varchar(5)
,col2 int
,col3 datetime
,col4 char(3)
)
insert into #YourTable values ('abcdf',123,'1/1/2009','aaa')
insert into #YourTable values ('aaaaa',456,'1/2/2009','bbb')
insert into #YourTable values ('bbbbb',789,'1/3/2009','aaa')
insert into #YourTable values ('ccccc',789,'1/4/2009','bbb')
insert into #YourTable values ('aaaaa',789,'1/5/2009','aaa')
insert into #YourTable values ('abcdf',789,'1/6/2009','aaa')
;with RankedYourTable AS
(
SELECT
ROW_NUMBER() OVER(PARTITION by col1 order by col1) AS col1Rank
,ROW_NUMBER() OVER(PARTITION by col2 order by col2) AS col2Rank
,ROW_NUMBER() OVER(PARTITION by col3 order by col3) AS col3Rank
,ROW_NUMBER() OVER(PARTITION by col4 order by col4) AS col4Rank
FROM #YourTable
)
SELECT
SUM(CASE WHEN col1Rank=1 THEN 1 ELSE 0 END) AS col1DistinctCount
,SUM(CASE WHEN col2Rank=1 THEN 1 ELSE 0 END) AS col2DistinctCount
,SUM(CASE WHEN col3Rank=1 THEN 1 ELSE 0 END) AS col3DistinctCount
,SUM(CASE WHEN col4Rank=1 THEN 1 ELSE 0 END) AS col4DistinctCount
FROM RankedYourTable
OUTPUT:
col1DistinctCount col2DistinctCount col3DistinctCount col4DistinctCount
----------------- ----------------- ----------------- -----------------
4 3 6 2
(1 row(s) affected)
and it's hardcoded.
It is not hardcoding to provide a field list for a sql statement. It's common and acceptable practice.
This won't necessarily be possible for every field in a table. For example, you can't do a DISTINCT against a SQL Server ntext or image field unless you cast them to other data types and lose some precision.
I appreciate all of the responses. I think the solution that will work best for me in this situation (counting the number of distinct values in each column of a table from an external program that has no knowledge of the table except its name) is as follows:
Run "describe table1" and pull out the column names from the result.
Loop through the column names and create the query to count the distinct values in each column. The query will look something like "select count(distinct columnA), count(distinct columnB), ... from table1".
Raj More's answer works well if you don't need to consider null as a value as count(distinct...) does not count null.
Here is a modification to count values including null by converting values to a string and replacing null with "NULL AS SOME IMPOSSIBLE STRING":
DECLARE #TableName VarChar (1024) = 'tableName'
DECLARE #SqlString VarChar (Max)
set #SqlString = (
SELECT DISTINCT
'SELECT ' +
RIGHT (ColumnList, LEN (ColumnList)-1) +
' FROM ' + Table_Name
FROM INFORMATION_SCHEMA.COLUMNS COL1
CROSS AppLy (
SELECT ', COUNT (DISTINCT coalesce(cast([' + COLUMN_NAME + '] as varchar),
''NULL AS SOME IMPOSSIBLE STRING'')) AS ' + '''' + COLUMN_NAME + ''''
FROM INFORMATION_SCHEMA.COLUMNS COL2
WHERE COL1.TABLE_NAME = COL2.TABLE_NAME
FOR XML PATH ('')
) TableColumns (ColumnList)
WHERE
COL1.TABLE_NAME = #TableName
)
EXECUTE (#SqlString)
DISTINCT is evil. Do COUNT/GROUP BY