Overview
- A Bloom filter uses a bit array plus multiple hash-derived positions to tell if an item is definitely absent or possibly present, trading exactness for far lower memory use.
- Standard sizing formulas let engineers pick the bit array size and number of hash functions from expected items and a target false-positive rate to control accuracy.
- The pieces show simple teaching code and a false-positive experiment while warning that those toy implementations are not production-grade and can give misleading results if used as-is.
- Calabro describes a one-hash variant that splits the bit array into coprime partitions so a single hash yields multiple bit positions, and he explains engine-level tweaks that make lookups cache-friendly.
- Production guidance stresses that standard Bloom filters cannot safely remove single items, that counting Bloom filters are required for deletion, and that teams should use battle-tested libraries (for example, Guava, Apache Commons, or Redis modules) and handle concurrency and monitoring in deployment.