A collection of standalone T-SQL scripts for exploring SQL Server query performance and behavior against the StackOverflow2010 sample database.
There's no build system, package manager, or test suite here — just .sql files meant to be opened in SQL Server Management Studio (or run via sqlcmd) against a SQL Server-compatible instance with the StackOverflow2010 database restored. None of the scripts contain a USE statement or connection details; point your session at the target database before running them.
Intentionally inefficient stored procedures that reproduce specific SQL Server performance problems — CPU pressure, join strategy choice, parameter sniffing — for teaching and diagnostics. Several files come in escalating variants (Mild → Medium → Hot → Nuclear) to show how much worse a pattern can get. A few (Hammer_CPU_Packaged.sql, Hammer_ParameterSniffing.sql, Hammer_StackOverflow.sql) pick a random branch on each execution, so running the same procedure repeatedly exercises different bad patterns. Hammer_Perf_Advanced.sql and Hammer_Perf_Diagnostics.sql are DMV-based diagnostic toolkits rather than stress generators.
A running catalog of T-SQL string functions, one section per function, each linking to its official Microsoft Learn documentation. Each demo query appends WHERE 1 = (SELECT 1) — a predicate that filters nothing but adds enough estimated cost to nudge the optimizer toward a parallel plan — so the resulting plans are useful for comparing how these functions behave under parallelism. Run one batch (GO-separated) at a time in SSMS with "Include Actual Execution Plan" enabled.
Captured actual execution plans (SET STATISTICS XML ON output) for the queries in Strings.sql, named string_<FUNCTION>_<p|s>.sqlplan — p if that query's plan went parallel, s if it stayed serial. These reflect one run against a specific dataset and server configuration (cost threshold for parallelism, MAXDOP), not a universal guarantee.