🔗 Your COUNT(DISTINCT) is too slow: Approximations and sampling in PostgreSQL

Can you afford to wait for precise distinct counts in Postgres? Elizabeth Garrett Christensen reveals that exact counts can be prohibitively slow, especially with large datasets. She outlines various strategies for approximating unique counts using features like the built-in TABLESAMPLE for random sampling, the HyperLogLog (HLL) for efficient distinct counting, and introduces Apache DataSketches for more complex scenarios. The article offers practical examples, including how to set up a sample page_views table with 10 million rows, and examines the benchmarks of different methods to highlight performance differences. You'll learn how accurate data analysis can still be quick and accessible without sacrificing too much precision.