function application

**natural deduction** metalanguage, practical foundations

**type theory** (dependent, intensional, observational type theory, homotopy type theory)

**computational trinitarianism** = **propositions as types** +**programs as proofs** +**relation type theory/category theory**

A function $f$ is defined by its association to each input value $x$ (belonging to some allowable domain of values) of an output value, usually denoted $f(x)$ or $f x$. The process of passing from $f$ and $x$ to $f(x)$ is called **function application**, and one speaks of **applying** $f$ to $x$ to produce $f(x)$.

The determination of the allowable domain for $x$, given $f$, depends a bit on foundational choices. In type theory and structural set theory, all functions have a type (a function type, naturally) which specifies their domain and codomain. In material set theory, a function is sometimes defined to be simply a particular sort of set of ordered pairs, with its domain specified implicitly as the set of elements occurring as first components of some such pair. (However, even in material set theory it is sometimes important for a function to come with a specified domain and/or codomain, in which case it can be defined to be an ordered triple.)

In formalized logic and type theory, $f$, $x$, and $f(x)$ are terms (or more precisely, metavariables? standing for terms), and the process of function application is a rule of term formation. This is something which belongs to the realm of syntax. On propositions ((-1)-truncated types) this is the modus ponens deduction rule.

Under a denotational semantics, each of these terms denotes a particular object, and we also refer to the object denoted by $f(x)$ as the result of applying the object denoted by $f$ to the object denoted by $x$. For instance, in a material set-theoretic semantics, $f$ would denote a set of ordered pairs such that for any $a$, there is at most one $b$ such that $(a,b)\in f$, and $x$ would denote some $a$ such that there *does* exist such a $b$, and $f(x)$ would denote that uniquely specified $b$. The distinction between the terms $f$, $x$, and $f(x)$ and what they denote is usually (and harmlessly) blurred in ordinary mathematical practice, but when studying logic and type theory formally it becomes important.

Under an operational semantics, by contrast, the “meaning” of the term $f(x)$ lies in how it is “evaluated”. Usually this proceeds by beta-reduction and related rules. For instance, if $f$ is the term $\lambda x. x*x$ and $x$ is the term $s(s(0))$ (the numeral two?), then $f(x)$ is $(\lambda x.x*x)(s(s(0)))$ which beta-reduces to $s(s(0))*s(s(0))$. The definition of $*$ can then be invoked to cause futher beta-reductions resulting in $s(s(s(s(0))))$ (the numeral four).

Last revised on January 26, 2014 at 07:12:56. See the history of this page for a list of all contributions to it.