Attribute Closure Calculator
Computes X⁺, the set of every attribute determined by X under a set of functional dependencies. Almost every other question in normalisation is answered by running this one operation in different places.
How the closure is built
Start with X itself. Scan the dependencies: whenever the whole left side of a dependency is already inside the set, add its right side. Repeat until a full pass adds nothing. The order of scanning never changes the answer — the closure is unique.
The part people get wrong is compound determinants. For AB → C, having A is not enough: the closure only grows if both A and B are already in the set. A dependency you cannot use yet may become usable after a later one fires, which is why the scan repeats.
What the closure tells you
Superkey: if X⁺ contains every attribute of the relation, X is a superkey.
Candidate key: a superkey where no proper subset is also a superkey. Drop one attribute at a time and recompute; if some smaller set still determines everything, X was not minimal.
Whether a dependency is implied: X → Y holds under your set if and only if Y is inside X⁺. That is the practical test for redundant dependencies, and the basis for finding a minimal cover.
Naming your attributes
Single-letter shorthand is understood, so AB → C means two attributes on the left. Real column names work too — write student_id, name — and once anything lowercase or underscored appears, nothing is split apart. That rule is per grammar, not per word, so a column called "name" is never read as N, A, M, E.
Attributes that appear in no dependency
An attribute mentioned nowhere in the dependency list is still part of the relation, and it changes the answer: it belongs to every candidate key, because nothing else can determine it. Put those in the extra attributes box or the keys will come out wrong.