12

I am studying the s-m-n theorem and the concept reminded me of currying.

From wikipedia article about s-m-n theorem:

the theorem says that for a given programming language and positive integers m and n, there exists a particular algorithm that accepts as input the source code of a program with m + n free variables, together with m values. This algorithm generates source code that effectively substitutes the values for the first m free variables, leaving the rest of the variables free.

From an article about currying:

Intuitively, currying says "if you fix some arguments, you get a function of the remaining arguments"

Seems like the same idea to me. The only reason why I am unsure is that the materials I came across on s-m-n don't mention currying (and vice versa), so I wanted to consult it to make sure I actually get it.

emanek
  • 223
  • 1
  • 6
  • Indeed. Some computability proofs have a lambdish flavor. The s-m-n theorem, very roughly, allows to pretend that indexes of recursive functions are lambda terms, so that given $\phi_i(-,-)$ we can craft the informal $g(x) = #\lambda y. \phi_i(x,y)$ and claim that $g$ is primitive recursive. Even the second recursion theorem proof (which exploits s-m-n) is Church's fixed point combinator in disguise, hidden behind $s()$ uses. The key point here is that even if the enumeration $\phi_i$ is defined enumerating, say, TMs (or Java, or ...) we can still pretend we have lambdas! – chi Sep 04 '17 at 14:00
  • Well, s-m-n makes an existential statement while the existence of a curried function provides the "compiler". But the idea is the same. – Raphael Sep 04 '17 at 14:05

1 Answers1

16

Yes, it is the same thing.

Currying is a concept from $\lambda$-calculus. It is a transformation between $A \times B \to C$ and $A \to (B \to C)$. Think of this as "if we have a function of two arguments of types $A$ and $B$, then we may fix the first argument (of type $A$), and we will get a function of the remaining argument (of type $B$)". In fact, this transformation is an isomorphism. This is made mathematically precise by mathematical models of (typed) $\lambda$-calculus, which are cartesian closed categories.

There is a category of numbered sets. A numbered set is a pair $(A, \nu_A)$ where $A$ is a set and $\nu_A : \mathbb{N} \to A$ is a partial surjection, i.e., a map from numbers onto $A$, which may also be undefined. If $\nu_A(n) = x$ then we say that $n$ is a code of $x$. In computability theory there are many examples. Whenever we encode some information with a number we get a numbered set. For instance, there is a standard numbering $\varphi$ of partial computable functions, so that $\varphi_n(k)$ is the number computed by the partial computable function encoded by $n$ when applied to $k$. (The result may be undefined.)

A morphism of numbered sets is a realized map $f : (A, \nu_A) \to (B, \nu_B)$, which means that there exists $n \in \mathbb{N}$ such that $f(\nu_A(k)) = \nu_B(\varphi_n(k))$ for all $k$ in the domain of $\nu_A$. This looks complicated, but all it says is that $\varphi_n$ does to codes what $f$ does to elements. It is the mathematical way of saying that "program $\phi_n$ implements function $f$".

Here is the punchline: the category of numbered sets is cartesian closed. We may therefore interpret the typed $\lambda$-calulus in it, and ask what program implements the currying operation. The answer is: the program given by the s-m-n theorem.

Andrej Bauer
  • 30,396
  • 1
  • 70
  • 117
  • Interesting. Is that category closely related to $PER(A)$? $\nu_A$ seems to induce a PER. – chi Sep 04 '17 at 14:06
  • 1
    Yes, the two categories are equivalent, and the third equivalent version is that of modest sets (lookup "modest sets and assemblies"). – Andrej Bauer Sep 04 '17 at 14:55