Dynamic SQL reads a local variable as a Table Variable? - tsql

This is NOT about using a table variable - this is about using a local variable to carry a db address within a dynamic SQL cursor which theoretically would work as follows:
-- Assume the global variables #sql, AnalysisLocation, and #sp_executeSql have been declared.
ALTER PROCEDURE [dbo].[sp_AggregateCompliance_Report]
#clientID int,
#InvScrDBLocation nvarchar(250),
#JoinFilter nvarchar(max) = '',
#Criteria nvarchar(max) = '',
#Year int = NULL
as
declare #sql nvarchar(4000)
set #sql = '
IF EXISTS (SELECT * FROM sys.tables WHERE name = ''tmp_Aggregate_Compliance_counts'')
TRUNCATE TABLE tmp_Aggregate_Compliance_counts
ELSE
CREATE TABLE tmp_Aggregate_Compliance_counts (
pfc_fk_prv_pkid int,
RxYear int,
RxMonth int,
Compliance decimal (6,5))
' print #sql EXEC sp_executesql #sql
SET #Criteria = isnull(case when #Criteria like 'WHERE %' then 'AND '+substring(#criteria,7,len(#criteria)-6) else #Criteria end ,'')
SET #Year = isnull(#year, year(getdate())-1)
set #sql = '
DECLARE #fk_cli_pkid INT
, #ServerAndDB_for_pfcAppended nvarchar(100)
DECLARE client_set CURSOR FOR
SELECT DISTINCT mtx.fk_cli_pkid, SettingValue+ ''.dbo.pfc_appended''
FROM mtx_ComplianceAndEarlyRefill_tracking AS mtx
JOIN prola7.Invoice_Screens.dbo.client_definition AS def
ON mtx.fk_cli_pkID = def.fk_cli_pkid
AND fk_lkSettings_pkID = 45
AND RecordStatus = 1
OPEN client_set
FETCH next FROM client_set
INTO #fk_cli_pkid, #ServerAndDB_for_pfcAppended
WHILE ##FETCH_STATUS = 0 BEGIN
INSERT INTO tmp_Aggregate_Compliance_counts (pfc_fk_prv_pkid, RxYear, RxMonth, Compliance)
SELECT pfc.pfc_fk_prv_pkid
, year(mtx.pfc_dateofservice) AS RxYear
, 0 AS RxMonth
, cast(mtx.Compliance as decimal (6,5))
FROM mtx_ComplianceAndEarlyRefill_tracking AS mtx
LEFT OUTER JOIN #ServerAndDB_for_pfcAppended AS pfc
ON mtx.pp_clientfile = pfc.pp_clientfile
AND mtx.pp_mirror_pkid = pfc.pp_mirror_pkid
AND mtx.fk_cli_pkid = #fk_cli_pkid
'+#JoinFilter+'
WHERE pfc.pfc_status = 0
AND year(mtx.pfc_dateofservice) = '+cast(#Year as nvarchar)+'
'+#Criteria+'
GROUP BY pfc.pfc_fk_prv_pkid, year(mtx.pfc_dateofservice)
FETCH next FROM client_set
INTO #fk_cli_pkid, #ServerAndDB_for_pfcAppended
END
CLOSE client_set
DEALLOCATE client_set
' print #sql EXEC sp_executesql #sql
This creates no syntax errors when compiling the dynamic code, however when calling this procedure: Msg 1087, Level 15, State 2, Line 27
Must declare the table variable "#ServerAndDB_for_pfcAppended".
When I use this type of structure passing the location variable in as a global variable from outside the procedure it accepts it correctly, however as a local variable it seems to default to presuming I intend it to be a table variable.
I do NOT want to create a table variable. Is this an impossible structure?

The error is caused by the fact that you are attempting to have a parametrised table name. This is not possible, and whenever a table name should be a parameter, a dynamic query is used, basically like this:
SET #sql = 'SELECT … FROM ' + #tablename + ' WHERE …'
I think, in your situation the cursor should be taken out of the dynamic query, except for the part that uses the parametrised table name. Something like this should probably do:
ALTER PROCEDURE [dbo].[sp_AggregateCompliance_Report]
#clientID int,
#InvScrDBLocation nvarchar(250),
#JoinFilter nvarchar(max) = '',
#Criteria nvarchar(max) = '',
#Year int = NULL
as
declare #sql nvarchar(4000)
set #sql = '
IF EXISTS (SELECT * FROM sys.tables WHERE name = ''tmp_Aggregate_Compliance_counts'')
TRUNCATE TABLE tmp_Aggregate_Compliance_counts
ELSE
CREATE TABLE tmp_Aggregate_Compliance_counts (
pfc_fk_prv_pkid int,
RxYear int,
RxMonth int,
Compliance decimal (6,5))
' print #sql EXEC sp_executesql #sql
SET #Criteria = isnull(case when #Criteria like 'WHERE %' then 'AND '+substring(#criteria,7,len(#criteria)-6) else #Criteria end ,'')
SET #Year = isnull(#year, year(getdate())-1)
DECLARE #fk_cli_pkid INT
, #ServerAndDB_for_pfcAppended nvarchar(100)
DECLARE client_set CURSOR FOR
SELECT DISTINCT mtx.fk_cli_pkid, SettingValue+ ''.dbo.pfc_appended''
FROM mtx_ComplianceAndEarlyRefill_tracking AS mtx
JOIN prola7.Invoice_Screens.dbo.client_definition AS def
ON mtx.fk_cli_pkID = def.fk_cli_pkid
AND fk_lkSettings_pkID = 45
AND RecordStatus = 1
OPEN client_set
FETCH next FROM client_set
INTO #fk_cli_pkid, #ServerAndDB_for_pfcAppended
WHILE ##FETCH_STATUS = 0 BEGIN
set #sql = '
INSERT INTO tmp_Aggregate_Compliance_counts (pfc_fk_prv_pkid, RxYear, RxMonth, Compliance)
SELECT pfc.pfc_fk_prv_pkid
, year(mtx.pfc_dateofservice) AS RxYear
, 0 AS RxMonth
, cast(mtx.Compliance as decimal (6,5))
FROM mtx_ComplianceAndEarlyRefill_tracking AS mtx
LEFT OUTER JOIN #ServerAndDB_for_pfcAppended AS pfc
ON mtx.pp_clientfile = pfc.pp_clientfile
AND mtx.pp_mirror_pkid = pfc.pp_mirror_pkid
AND mtx.fk_cli_pkid = #fk_cli_pkid
'+#JoinFilter+'
WHERE pfc.pfc_status = 0
AND year(mtx.pfc_dateofservice) = '+cast(#Year as nvarchar)+'
'+#Criteria+'
GROUP BY pfc.pfc_fk_prv_pkid, year(mtx.pfc_dateofservice)
' print #sql EXEC sp_executesql #sql
FETCH next FROM client_set
INTO #fk_cli_pkid, #ServerAndDB_for_pfcAppended
END
CLOSE client_set
DEALLOCATE client_set

Related

Output parameter from TSQL sp_executesql seems to be null

I'm sure there's a simple explanation. I'm trying to print out the Max ID for all tables with an Identity column.
The TSQL code below seems to run fine in the sense that the Results window contains cells with values, but the Messages window does not show: Table1 Row Count = 99
Seems as if the #Result variable is NULL when it gets to the PRINT command.
Am I missing something obvious or is the OUTPUT command working differently than I expect?
DECLARE #TABLE VARCHAR(100)
DECLARE #FIELD VARCHAR(100)
DECLARE #SQL NVARCHAR(300)
DECLARE #result BIGINT
DECLARE CUR CURSOR FOR
SELECT
[table] = t.name,
[colname] = c.name
FROM sys.schemas AS s
INNER JOIN sys.tables AS t
ON s.[schema_id] = t.[schema_id]
INNER JOIN sys.identity_columns c ON c.object_id = t.[object_id]
OPEN CUR
FETCH NEXT FROM CUR INTO #TABLE, #FIELD
WHILE ##FETCH_STATUS = 0
BEGIN
SET #SQL = 'SELECT MAX([' + #FIELD + ']) FROM [' + #TABLE + ']'
EXEC sp_executesql #SQL,
N'#result int OUTPUT',
#result OUTPUT
PRINT #TABLE + ' Row Count = ' + CONVERT(nvarchar, #result)
--NOTHING SEEMS TO BE PRINTED HERE
FETCH NEXT FROM CUR INTO #TABLE, #FIELD
END
CLOSE CUR
DEALLOCATE CUR

Dynamic SQL with nvarchar(max) variable (again)

I have read about a dozen articles here and I am still stumped with this issue.
I am building a dynamic select statement that will update a view on a monthly schedule.
set ansi_nulls on
go
set quoted_identifier on
go
alter procedure [dbo].[Proc_Name_SP]
as
begin
set nocount on
set quoted_identifier off
declare #dbname varchar(10), #schema_id int, #schema_name varchar(10),
#jacro varchar(10), #rec_cnt int, #tot_rec int
declare #SQL_Main nvarchar(max), #SQL_Final nvarchar(max),
#SQL_schema nvarchar(2000), #SQL_Union nvarchar(max)
declare iteration cursor global static for
-- Begin statement for cursor array
select distinct db, code
from linkedserver.db.schema.Directory
where current_stage = 'live'
order by db
-- End statement for cursor array
-- get total number of cursor iterations to know when to stop
-- "union" statements
select #tot_rec = count(*) from (select distinct db, code
from [linkedserver].db.schema.Directory
where current_stage = 'live') as cur
-- begin loop
open iteration
fetch first from iteration into #dbname, #jacro
while ##fetch_status=0
begin
-- the schema used is not consistent. Because of the linked server it was
-- necessary to get the Schema_ID from the sys.tables and then pull the
-- schema name from sys.schema
set #SQL_schema = 'select #sch_id = schema_id from [linkedserver].'+#dbname+'.sys.tables where name = ''Manuscript'''
execute sp_executesql #SQL_schema, N'#sch_id int OUTPUT', #sch_id = #schema_id output
--print #schema_id
set #SQL_schema ='select #sch_name = name from [linkedserver].'+#dbname+'.sys.schemas where schema_id = '+cast(#schema_id as varchar)+''
execute sp_executesql #SQL_schema, N'#sch_name nvarchar(10) OUTPUT', #sch_name = #schema_name output
--print #schema_name
--building Select statement
set #SQL_Main ='
select jcode.Code as BILLING_ACRO
,s.start_dt as BILLING_DATE
,cmpt_ms_nm as MANUSCRIPT
,isnull(jcode.billing_type, ''reviewed'') as Billing_type
from [linkedserver].'+#dbname+'.'+#schema_name+'.Manuscript as m
join [linkedserver].'+#dbname+'.'+#schema_name+'.Step as s on m.ms_id = s.ms_id and m.ms_rev_no = s.ms_rev_no
join (select j_id, Code, billing_type from [linkedserver].db.schema.Directory where db = '''+#dbname+''') as jcode on jcode.j_id = m.j_id
where jcode.Code = '''+#jacro+'''
and m.ms_rev_no = 0
and s.stage_id = 190
and isnull(cmpt_ms_nm, '''') <> ''''
and s.step_id = (select min(s2.step_id)
from [linkedserver].'+#dbname+'.'+#schema_name+'.Step as s2
where s2.stage_id = 190
and s2.ms_id = m.ms_id
and s2.ms_rev_no = m.ms_rev_no)
'
set #rec_cnt = isnull(#rec_cnt, 0) + 1
if #SQL_Union is null
begin
set #SQL_Union = #SQL_Main
end
else if #tot_rec <> #rec_cnt
begin
set #SQL_Union = #SQL_Union + ' union ' + #SQL_Main
end
else
begin
set #SQL_Union = #SQL_Union + #SQL_Main
end
--print #rec_cnt
fetch next from iteration into #dbname, #jacro --next database
end -- while ##FETCH_STATUS=0
close iteration
deallocate iteration
-- build new view
print len(#SQL_Union)
set #SQL_Final = '
ALTER VIEW [dbo].[View_Name_VW]
AS
'+#SQL_Union+'
'
execute sp_executesql #SQL_Final
--grab string variables to table for troubleshooting
insert into Output_SQL(SQL_Final, SQL_Final_Len, SQL_Union, SQL_Union_Len)
select #SQL_Final, LEN(#SQL_Final), #SQL_Union, LEN(#SQL_Union)
set nocount off
end
go
I have read that others have had problems with this type of truncation and I have tried multiple suggestions but in the end the I am getting capped at 68274 in this code with nvarchar(max). For troubleshooting, I am saving the results of the variables and the len of these variables to a table to eliminate the SSMS cap on the display of strings.
I have tried cast(#varible as nvarchar(max)) on the right side of the = sign. I have changed the data type lengths (as the select that is being built is not that large, it is just large after it has been union for each unique customer)
I am open to any suggestions as I have tried many variations of datatype declarations for these variables.

How do i can resolve this promblem?

I have an table include 1 column ID and 12 columns month (Month01 to Month12). If i have an parameter #month=9, how can i compare to get column Month01 to Month09 without Month10, Month11, Month12. Please help me!
Simo's answer is more elegant, but below is an alternative "basic" way to do this:
DECLARE #month INT
SET #month = 9
IF #Month = 1
BEGIN
SELECT Id, Month01
FROM MonthlyData
END
IF #Month = 2
BEGIN
SELECT Id, Month01, Month02
FROM MonthlyData
END
-- Repeat for months 3-9
IF #Month = 9
BEGIN
SELECT Id, Month01, Month02, Month03, Month04,
Month05, Month06, Month07, Month08, Month09
FROM MonthlyData
END
-- Repeat for months 10-12
See my SQL Fiddle.
declare #str varchar(1000)
SELECT #str= coalesce(#str + ', ', '') + a.name
FROM (
Select name from sys.columns
where object_id = OBJECT_ID('table_months')
and isnumeric(right(name,2))= 1 and name like 'month%'
and cast(right(name,2) as int) < '03') a
-- Instead of '03' you use a variable and assign required moths number
declare #sql nvarchar(100)
set #sql = 'select ID, '+ #str+' from table_months'
exec sp_executesql #sql
you may try this out also :)
DECLARE #month int = 9
DECLARE #Sql nvarchar(100) =''
DECLARE #pos int = 1
WHILE(#pos <= #month)
BEGIN
SET #Sql = #Sql + (',month' + right('00'+ rtrim(#pos), 2))
SET #pos = #pos + 1
END
SET #Sql= 'SELECT ID ' +#Sql + ' FROM table'
EXEC sp_executesql #Sql
DECLARE #month int
DECLARE #tableName nvarchar(128)
SET #month = 9;
SET #tableName = 'months' -- YOUR TABLE's NAME
DECLARE #query nvarchar(2048)
SET #query =
'SELECT id, ' + STUFF((SELECT DISTINCT ', ' + name
FROM sys.columns WHERE name LIKE'Month%'
AND object_id = OBJECT_ID(#tableName)
AND CONVERT(INT, RIGHT(name, 2)) <= #month
FOR XML PATH(''), TYPE).value('.', 'NVARCHAR(MAX)'),1,1,'')
+ ' FROM ' + #tableName;
EXEC(#query)

T-SQL How To Count Of Records For Each Column

I have a table with over 120 columns and need to determine which column is used the least. I tried using sql queries to do this, but found T-SQL a bit simpler.
I tried the following but my count comes out as 0 for every column.
Declare data1 Cursor for select column_name
from information_schema.columns
where table_name = 'repository'
Declare #mField nvarchar(255)
Declare #count int
Open data1;
fetch next from data1 into #mField;
set #count = -1;
while ##fetch_status = 0
begin
select #count = count(#mField)
from repository where tablereference =
'central' and ( #mField!= null )
print #mField+' ' ;
print #count;
Fetch next from data1 into #mField;
end
close data1;
deallocate data1;
You can't count values like this because you are only testing if #mField is NULL. The column name isn't substituted.
COUNT ignores NULLs anyway so if you want to count non-null values, do this:
DECLARE #sql varchar(4000)
SET #sql = 'SELECT COUNT(*) AS Total '
SELECT #sql = #sql + ', COUNT(' + QUOTENAME(column_name) + ') AS ' + QUOTENAME(column_name)
from information_schema.columns
where table_name = 'repository'
SET #sql = #sql + ' FROM repository'
EXEC (#sql)
This queries the table once for all columns
You need to use some dynamic sql in the middle to acheive your aim here.
Declare data1 Cursor for select column_name
from information_schema.columns
where table_name = 'repository'
Declare #mField nvarchar(255)
Open data1;
fetch next from data1 into #mField;
while ##fetch_status = 0
begin
exec ('
declare #count int
select #count = count([' + #mField + '])
from repository where tablereference =
''central'' and ( [' + #mField + '] is not null)
if #count < 10
begin
print ''' + #mField + ' '' ;
print #count;
end
')
Fetch next from data1 into #mField;
end
close data1;
deallocate data1;
Your count(#mField) is a count of the litteral value that happens to be in #mField, its not resolving the field name into COUNT(fldBlah), you would need to use dynamic SQL for that.
Declare data1 Cursor FAST_FORWARD for select column_name
from information_schema.columns where table_name = 'repository'
Declare #mField nvarchar(255)
Declare #SQL varchar(1024)
Declare #results table (col_name varchar(128), non_nulls int)
Open data1;
fetch next from data1 into #mField;
while (##FETCH_STATUS = 0) begin
set #SQL = 'SELECT ''' + #mField + ''', count(' + #mField + ') from repository where tablereference = ''central'''
insert #results
exec(#SQL)
Fetch next from data1 into #mField;
end
close data1;
deallocate data1;
select * from #results
For an output like;
col_name non_nulls
[tablereference] 5
[another_col] 1
Where there are 5 non-null values in column tablereference etc
You need to use IS NOT NULL instead of != NULL

How to get sp_executesql result into a variable?

I have a piece of dynamic SQL I need to execute, I then need to store the result into a variable.
I know I can use sp_executesql but can't find clear examples around about how to do this.
If you have OUTPUT parameters you can do
DECLARE #retval int
DECLARE #sSQL nvarchar(500);
DECLARE #ParmDefinition nvarchar(500);
DECLARE #tablename nvarchar(50)
SELECT #tablename = N'products'
SELECT #sSQL = N'SELECT #retvalOUT = MAX(ID) FROM ' + #tablename;
SET #ParmDefinition = N'#retvalOUT int OUTPUT';
EXEC sp_executesql #sSQL, #ParmDefinition, #retvalOUT=#retval OUTPUT;
SELECT #retval;
But if you don't, and can not modify the SP:
-- Assuming that your SP return 1 value
create table #temptable (ID int null)
insert into #temptable exec mysp 'Value1', 'Value2'
select * from #temptable
Not pretty, but works.
DECLARE #vi INT
DECLARE #vQuery NVARCHAR(1000)
SET #vQuery = N'SELECT #vi= COUNT(*) FROM <TableName>'
EXEC SP_EXECUTESQL
#Query = #vQuery
, #Params = N'#vi INT OUTPUT'
, #vi = #vi OUTPUT
SELECT #vi
DECLARE #tab AS TABLE (col1 VARCHAR(10), col2 varchar(10))
INSERT into #tab EXECUTE sp_executesql N'
SELECT 1 AS col1, 2 AS col2
UNION ALL
SELECT 1 AS col1, 2 AS col2
UNION ALL
SELECT 1 AS col1, 2 AS col2'
SELECT * FROM #tab
Return values are generally not used to "return" a result but to return success (0) or an error number (1-65K). The above all seem to indicate that sp_executesql does not return a value, which is not correct. sp_executesql will return 0 for success and any other number for failure.
In the below, #i will return 2727
DECLARE #s NVARCHAR(500)
DECLARE #i INT;
SET #s = 'USE [Blah]; UPDATE STATISTICS [dbo].[TableName] [NonExistantStatisticsName];';
EXEC #i = sys.sp_executesql #s
SELECT #i AS 'Blah'
SSMS will show this
Msg 2727, Level 11, State 1, Line 1
Cannot find index 'NonExistantStaticsName'.
If you want to return more than 1 value use this:
DECLARE #sqlstatement2 NVARCHAR(MAX);
DECLARE #retText NVARCHAR(MAX);
DECLARE #ParmDefinition NVARCHAR(MAX);
DECLARE #retIndex INT = 0;
SELECT #sqlstatement = 'SELECT #retIndexOUT=column1 #retTextOUT=column2 FROM XXX WHERE bla bla';
SET #ParmDefinition = N'#retIndexOUT INT OUTPUT, #retTextOUT NVARCHAR(MAX) OUTPUT';
exec sp_executesql #sqlstatement, #ParmDefinition, #retIndexOUT=#retIndex OUTPUT, #retTextOUT=#retText OUTPUT;
returned values are in #retIndex and #retText
Declare #variable int
Exec #variable = proc_name
DECLARE #ValueTable TABLE
(
Value VARCHAR (100)
)
SELECT #sql = N'SELECT SRS_SizeSetDetails.'+#COLUMN_NAME+' FROM SRS_SizeSetDetails WHERE FSizeID = '''+#FSizeID+''' AND SRS_SizeSetID = '''+#SRS_SizeSetID+'''';
INSERT INTO #ValueTable
EXEC sp_executesql #sql;
SET #Value='';
SET #Value = (SELECT TOP 1 Value FROM #ValueTable)
DELETE FROM #ValueTable
This worked for me:
DECLARE #SQL NVARCHAR(4000)
DECLARE #tbl Table (
Id int,
Account varchar(50),
Amount int
)
-- Lots of code to Create my dynamic sql statement
insert into #tbl EXEC sp_executesql #SQL
select * from #tbl
Here's something you can try
DECLARE #SqlStatement NVARCHAR(MAX) = ''
,#result XML
,#DatabaseName VARCHAR(100)
,#SchemaName VARCHAR(10)
,#ObjectName VARCHAR(200);
SELECT #DatabaseName = 'some database'
,#SchemaName = 'some schema'
,#ObjectName = 'some object (Table/View)'
SET #SqlStatement = '
SELECT #result = CONVERT(XML,
STUFF( ( SELECT *
FROM
(
SELECT TOP(100)
*
FROM ' + QUOTENAME(#DatabaseName) +'.'+ QUOTENAME(#SchemaName) +'.' + QUOTENAME(#ObjectName) + '
) AS A1
FOR XML PATH(''row''), ELEMENTS, ROOT(''recordset'')
), 1, 0, '''')
)
';
EXEC sp_executesql #SqlStatement,N'#result XML OUTPUT', #result = #result OUTPUT;
SELECT DISTINCT
QUOTENAME(r.value('fn:local-name(.)', 'VARCHAR(200)')) AS ColumnName
FROM #result.nodes('//recordset/*/*') AS records(r)
ORDER BY ColumnName
This was a long time ago, so not sure if this is still needed, but you could use ##ROWCOUNT variable to see how many rows were affected with the previous sql statement.
This is helpful when for example you construct a dynamic Update statement and run it with exec. ##ROWCOUNT would show how many rows were updated.
Here is the definition