SQL & normalisation

SQL is how you talk to a relational database, and normalisation is how you design tables that do not contradict themselves. The two meet in one rule: store each fact once. A table that repeats a department name beside every student will eventually disagree with itself — that is an update anomaly. Worked example — closure: with dependencies A -> B and B -> C, the closure of {A} is {A, B, C}, because A gives B and B gives C. Closure is the mechanical tool for every key question: compute it, do not guess. From closure you can name the violation precisely: a non-key attribute depending on part of a key breaks 2NF; depending on another non-key attribute breaks 3NF; a determinant that is not a superkey breaks BCNF.