What is sqlrace?

SQL Race is a teaching tool that demonstrates a common but tricky bug called a "race condition" in database applications. It shows that simply wrapping database operations in a transaction does not, by itself, prevent data corruption when multiple processes try to change the same data at the same time. A race condition happens when multiple workers all try to read and update the same piece of data at once, stepping on each other's toes. This tool runs a simple experiment: it starts with a counter set to 4,096 and asks four workers to subtract 1 from it repeatedly, over a thousand times each. If everything works correctly, the counter should reach zero. The tool lets you test different strategies for updating the database, ranging from a naive approach with no protection to more sophisticated methods. The demo's output tells the story clearly. When the workers use the naive approach, the final result is wrong, they lose some decrements because the workers overwrite each other's changes. Switching to a standard transaction does not fix the problem, the result is still incorrect. Only two methods get the right answer: locking the row so no one else can touch it until the update is complete, or using an atomic update where the database itself handles the subtraction safely. This project is aimed at developers and students learning about database concurrency. A practical example would be an e-commerce checkout where two customers are buying the last item in stock simultaneously, without proper safeguards, both might succeed, and the store oversells. The demo proves that you need specific techniques like row locking or atomic operations, not just transactions, to keep data accurate under concurrent load. The project is written in Go and requires a MySQL-compatible database to run. The README notes that the atomic method is also the fastest, since it avoids the overhead of locking, making it both correct and efficient.