SQLServerBang

SQL Server does not sniff for variable, it just simply uses the fixed value.   I'm going to use AdventureWorks 2019 in this post. --------------------Please run this script--------------- use AdventureWorks2019 go IF exists(select 1 from sys.tables where name='SalesOrderDetail' and schema_id=schema_id('dbo'))       drop table SalesOrderDetail go select * into SalesOrderDetail from [Sales].[SalesOrderDetail] go create statistics iProductID ON SalesOrderDetail(productid) with fullscan go dbcc traceon(3604,2363)---trace flag 2363 displays more detail about the selectivity go --------------------Please run this script---------------     Equality(=): all density DECLARE @pid INT = 0 SELECT * FROM SalesOrderDetail WHERE ProductID = @pid     456= 'All density'*card =0.003759399*121317(note, if there new rows inserted or deleted, you have to replace the 121317 with current card)   DBCC execution completed. If DBCC printed error messages, contact your system administ

The threshold of temp table in stored procedure is little bit tricky, it doesn’t exactly follow the regular threshold. First of all, you can’t use the modification_counter to determine when auto update statistics happens. Because the data in tempdb is deleted every time after execution, replying on the data that that already deleted is pointless. Secondly, the algorithm is different: Here is the original threshold of temp table that are not in stored procedure.   Here is the new algorithm. (note, compatibility of user database decides old threshold or sublinear threshold, the compatibility of tempdb does not matter) Table cardinality ( n ) Old threshold Sublinear threshold n  < 6 >=n+6 >=n+6 6 <=  n  <= 500 >=n+500 >=n+500 500<n  <=19682 >= n+(500 +0.2*n) or <= n-(500+0.2*n) >= n+(500 +0.2*n) or <= n-(500+0.2*n) n  >19682 500 + (0.20 *  n ) >= n+sqrt(1000*n) or <=n-sqrt(1000*n)     Here is an example in compatibility level 150 in SQL 2019