Unification (computer science)

Unification, in computer science and logic, is an algorithmic process by which one attempts to solve the satisfiability problem. The goal of unification is to find a substitution which demonstrates that two seemingly different terms are in fact either identical or just equal. Unification is widely used in automated reasoning, logic programming and programming language type system implementation.

Several kinds of unification are commonly studied: that for theories without any equations (the empty theory) is referred to as syntactic unification: one wishes to show that (pairs of) terms are identical. If one has a non-empty equational theory, then one is typically interested in showing the equality of (a pair of) terms; this is referred to as semantic unification. Since substitutions can be ordered into a partial order, unification can be understood as the procedure of finding a join on a lattice.

Unification on ground terms is just the ground word problem; because the ground word problem is undecidable, so is unification.

The first formal investigation of unification can be attributed to John Alan Robinson, who used first-order unification as a basic building block of his resolution procedure for first-order logic, a great step forward in automated reasoning technology, as it eliminated one source of combinatorial explosion: searching for instantiation of terms.

Syntactic unification of first-order terms

First-order terms

Main article: Term (logic)

Given a set of variable symbols X = {x,y,z,...}, a set of distinct constant symbols C = {a,b,c,...}, and a set of distinct function symbols F = {f,g,h,...}, a term is defined as any expression that can be generated by a finite number of applications of the following rules:

• Basis: any variable $x \in X$, and also any constant $a \in C$ is a term
• Induction: if $t_1,\ldots,t_k$ are terms then $f(t_1,\ldots,t_k)$ is term for finite k, 0 < k.

E.g. x,y,a and b are terms that can be generated directly from the Basis rule; f(a,x), g(b,y,x) are terms by a single application of the Induction rule to terms generated from the Basis; f(a,f(a,x)) is a term generated by a second application of the Induction rule to terms we have already generated; and so on... Often, for brevity, the constant symbols are regarded as function symbols taking zero arguments, the induction rule is relaxed to allow terms with zero arguments $f(t_0,\ldots,t_k)$, and a() is regarded as syntactically equal to a: see first-order terms. The distinction between constants (zero arity functions) and function symbols with arity greater than zero is often made to clearly distinguish Basis and Induction cases of terms for purposes of proofs. Often mathematicians fix the arity (number of arguments) of a function symbol (see Signature) while typically in syntactic unification problems a function symbol may have any (finite) number of arguments, and possibly may have different numbers of arguments in different contexts. e.g. f(f(a),f(x,y,z)) is a well formed term for unification problems.

Substitution

Main article: Substitution (logic)

A substitution is defined as a finite set of mappings from variables to terms $\{ x_0 \mapsto t_0,\dots, x_k \mapsto t_k \}$ where each mapping must be unique, because mapping the same variable to two different terms would be ambiguous. A substitution may be applied to a term u and is written $u \{ x_0 \mapsto t_0,\dots, x_k \mapsto t_k \}$, which means (simultaneously) replace every occurrence of each variable x_i in the term u with the term t_i for $0\leq i \leq k$. E.g. $f(x,a,g(z),y) \{ x \mapsto h(a,y), z \mapsto b \} = f(h(a,y),a,g(b),y)$

Syntactic unification problem on first-order terms

A syntactic unification problem on first-order terms is a conjunction of a finite number of potential equations on terms $t_0 \stackrel{?}{=} u_0 \And, \ldots, \And t_n \stackrel{?}{=} u_n$. To solve the problem, a substitution θ must be found so that the terms on LHS and RHS of each of the potential equalities become syntactically equivalent when the substitution is applied i.e. $t_0\theta = u_0\theta \And, \ldots, \And t_n\theta = u_n\theta$. Such a substitution θ is called a unifier. Alternatively a unification problem might have no solution. E.g. $x\stackrel{?}{=} z \And y \stackrel{?}{=} f(x)$ has a unifier $\{ x \mapsto z, y \mapsto f(z) \}$, because

$x \{ x \mapsto z, y \mapsto f(z) \} = z = z \{ x \mapsto z, y \mapsto f(z) \} = z$ and $y\{ x \mapsto z, y \mapsto f(z) \} = f(z) = f(x)\{ x \mapsto z, y \mapsto f(z) \} = f(z).$

Occurs check

Main article: Occurs check

If there is ever an attempt to unify a variable x with a term with a function symbol containing x as a strict subterm x=f(...,x,...) then x would have to be an infinite term, which contradicts the strict definition of terms that requires a finite number of applications of the induction rule. e.g x=f(x) does not represent a strictly valid term.

Informal overview

Given two input terms s and t, (syntactic) unification is the process which attempts to find a substitution that structurally identifies s and t. If such a substitution exists, we call this substitution a unifier of s and t.

In theory, some pairs of input terms may have infinitely many unifiers. However, for most applications of unification, it is sufficient to consider a most general unifier (mgu). An mgu is useful because all other unifiers are instances of the mgu.

In particular, first-order unification is the syntactic unification of first-order terms (terms built from variable and function symbols). Higher-order unification, on the other hand, is the unification of higher-order terms (terms containing some higher-order variables).

The set of all syntactically equivalent terms is variously called the free theory (because it is a free object), the empty theory (because the set of equational sentences is empty), or the theory of uninterpreted functions (because unification is done on uninterpreted terms.)

The theoretical properties of a particular unification algorithm depend upon the variety of term used as input. First-order unification, for instance, is decidable, efficiently so, and all unifiable terms have (unique, up to variable renamings) most general unifiers. Higher-order unification, on the other hand, is generally undecidable and lacks most general unifiers, though Huet's higher-order unification algorithm seems to work sufficiently well in practice.

Aside from syntactic unification, another form of unification, called semantic unification (known also as equational-unification, E-unification) is also widely used. The key difference between the two notions of unification is how two unified terms should be considered 'equal'. Syntactic unification attempts to find a substitution making the two input terms structurally identical. However, semantic unification attempts to make the two input terms equal modulo some equational theory. Some particularly important equational theories have been widely studied in the literature. For instance, AC-unification is the unification of terms modulo associativity and commutativity.

Unification is a significant tool in computer science. First-order unification is especially widely used in logic programming, programming language type system design, especially in type inferencing algorithms based on the Hindley-Milner type system, and automated reasoning. Higher-order unification is also widely used in proof assistants, for example Isabelle and Twelf, and restricted forms of higher-order unification (higher-order pattern unification) are used in some programming language implementations, such as lambdaProlog, as higher-order patterns are expressive, yet their associated unification procedure retains theoretical properties closer to first-order unification. Semantic unification is widely used in SMT solvers and term rewriting algorithms.

Definition of unification for first-order logic

Let p and q be sentences in first-order logic.

UNIFY(p,q) = U where subst(U,p) = subst(U,q)

Where subst(U,p) means the result of applying substitution U on the sentence p. Then U is called a unifier for p and q. The unification of p and q is the result of applying U to both of them.^[1]

Let L be a set of sentences, for example, L = {p,q}. A unifier U is called a most general unifier for L if, for all unifiers U' of L, there exists a substitution s such that applying s to the result of applying U to L gives the same result as applying U' to L:

subst(U',L) = subst(s,subst(U,L)).

Unification in logic programming

The concept of unification is one of the main ideas behind logic programming, best known through the language Prolog. It represents the mechanism of binding the contents of variables and can be viewed as a kind of one-time assignment. In Prolog, this operation is denoted by the equality symbol =, but is also done when instantiating variables (see below). It is also used in other languages by the use of the equality symbol =, but also in conjunction with many operations including +, -, *, /. Type inference algorithms are typically based on unification.

In Prolog:

1. A variable which is uninstantiated—i.e. no previous unifications were performed on it—can be unified with an atom, a term, or another uninstantiated variable, thus effectively becoming its alias. In many modern Prolog dialects and in first-order logic, a variable cannot be unified with a term that contains it; this is the so called occurs check.
2. Two atoms can only be unified if they are identical.
3. Similarly, a term can be unified with another term if the top function symbols and arities of the terms are identical and if the parameters can be unified simultaneously. Note that this is a recursive behavior.

Type inference

Unification is used during type inference, for instance in the functional programming language Haskell (programming language). On one hand, the programmer does not need to provide type information for every function, on the other hand it is used to detect typing errors. The Haskell expression 1:['a','b','c'] is not correctly typed, because the list construction function ":" is of type a->[a]->[a] and for the first argument "1" the polymorphic type variable "a" has to denote the type Int whereas "['a','b','c']" is of type [Char], but "a" cannot be both Char and Int at the same time.

Like for prolog an algorithm for type inference can be given:

1. Any type variable unifies with any type expression, and is instantiated to that expression. A specific theory might restrict this rule with an occurs check.
2. Two type constants unify only if they are the same type.
3. Two type constructions unify only if they are applications of the same type constructor and all of their component types recursively unify.

Due to its declarative nature, the order in a sequence of unifications is (usually) unimportant.

Note that in the terminology of first-order logic, an atom is a basic proposition and is unified similarly to a Prolog term.

Higher-order unification

Many applications require one to consider the unification of typed lambda-terms instead of first-order terms. Such unification is often called higher-order unification. A well studied branch of higher-order unification is the problem of unifying simply typed lambda terms modulo the equality determined by αβη conversions. Such unification problems do not have most general unifiers. While higher-order unification is undecidable,^[2][3] Gérard Huet gave a semi-decidable (pre-)unification algorithm^[4] that allows a systematic search of the space of unifiers (generalizing the unification algorithm of Martelli-Montanari^[5] with rules for terms containing higher-order variables. Huet^[6] and Gilles Dowek^[7] have written articles surveying this topic.

Dale Miller has described what is now called higher-order pattern unification.^[8] This subset of higher-order unification is decidable and solvable unification problems have most-general unifiers. Many computer systems that contain higher-order unification, such as the higher-order logic programming languages λProlog and Twelf, often implement only the pattern fragment and not full higher-order unification.

In computational linguistics, one of the most influential theories of ellipsis is that ellipses are represented by free variables whose values are then determined using Higher-Order Unification (HOU). For instance, the semantic representation of "Jon likes Mary and Peter does too" is like(j; m)R(p) and the value of R (the semantic representation of the ellipsis) is determined by the equation like(j; m) = R(j). The process of solving such equations is called Higher-Order Unification.^[9]

Examples of syntactic unification of first-order terms

In the Prolog syntactical convention a symbol starting with an upper case letter is a variable name; a symbol that starts with a lowercase letter is a function symbol; the comma is used as the logical and operator. In maths the convention is to use lower case letters at the end of the alphabet as variable names (e.g. x,y,z); letters f,g,h as function symbols, and letters a,b,c as constants, and constants are often regarded as functions that take no arguments; logical "and" is represented by & or $\and$

Prolog Notation	Maths Notation	Unifying Substitution	Explanation
a = a	a = a	{}	Succeeds. (tautology)
a = b	a = b	fail	a and b do not match
X = X	x = x	{}	Succeeds. (tautology)
a = X	a = x	$\{ x \mapsto a \}$	x is unified with the constant a
X = Y	x = y	$\{ x \mapsto y \}$	x and y are aliased
f(a,X) = f(a,b)	f(a,x) = f(a,b)	$\{ x \mapsto b \}$	function and constant symbols match, x is unified with the constant b
f(a) = g(a)	f(a) = g(a)	fail	f and g do not match
f(X) = f(Y)	f(x) = f(y)	$\{ x \mapsto y \}$	x and y are aliased
f(X) = g(Y)	f(x) = g(y)	fail	f and g do not match
f(X) = f(Y,Z)	f(x) = f(y,z)	fail	Fails. The terms have different arity
f(g(X)) = f(Y)	f(g(x)) = f(y)	$\{ y \mapsto g(x) \}$	Unifies y with the term g(x)
f(g(X),X) = f(Y,a)	f(g(x),x) = f(y,a)	$\{ x \mapsto a, y \mapsto g(a) \}$	Unifies x with constant a, and y with the term g(a)
X = f(X)	x = f(x)	should fail	Fails in strict first-order logic and many modern Prolog dialects (enforced by the occurs check). Succeeds in traditional Prolog, unifying x with infinite expression x=f(f(f(f(...)))) which is not strictly a term.
X = Y, Y = a	x = y & y = a	$\{ x \mapsto a, y \mapsto a \}$	Both x and y are unified with the constant a
a = Y, X = Y	a = y & x = y	$\{ x \mapsto a, y \mapsto a \}$	As above (unification is symmetric, and transitive)
X = a, b = X	x = a & b = x	fail	Fails. a and b do not match, so x can't be unified with both

A Unification Algorithm

Given a unification problem G consisting of a finite multi-set of potential equations on terms $s_0,t_0, \ldots, s_n,t_n, \text{ } G = \{ s_0 \stackrel{\text{?}}{=} t_0, \ldots , s_n \stackrel{\text{?}}{=} t_n \}$, the algorithm applies term rewriting rules to transform the multi-set of potential equations to an equivalent multi-set of equations of the form $\{ x_0 \stackrel{\text{?}}{=} u_0, \ldots, x_m \stackrel{\text{?}}{=} u_m \}$ where $x_0,\ldots,x_m$ are unique variables (appearing exactly once on the LHS of one equation, and nowhere else). A multi-set of this form can be read as a substitution. If there is no solution the algorithm terminates with $\bot$. The set of variables in a term t is written as Vars(t), and the set of variables in all terms on LHS or RHS of potential equations in a problem G is similarly written as Vars(G). The operation of substituting all occurrences of variable x in problem G with term t is denoted $G\{ x \mapsto t\}$. For brevity constant symbols are regarded as function symbols having zero arguments.

$G \cup \{ t \stackrel{\text{?}}{=} t \} \Rightarrow G$

$G \cup \{ f(s_0,\dots,s_k) \stackrel{\text{?}}{=} f(t_0,\ldots,t_k) \} \Rightarrow G \cup \{ s_0 \stackrel{\text{?}}{=} t_0, \ldots, s_k \stackrel{\text{?}}{=} t_k \}$

$G \cup \{ f(s_0,\dots,s_k) \stackrel{\text{?}}{=} g(t_0,\ldots,t_m) \} \Rightarrow \bot \text{ if } f \neq g \or k \neq m$

$G \cup \{ f(s_0,\dots,s_k) \stackrel{\text{?}}{=} x \} \Rightarrow G \cup \{ x \stackrel{\text{?}}{=} f(s_0,\dots,s_k) \}$

$G \cup \{ x \stackrel{\text{?}}{=} t \} \Rightarrow G\{ x \mapsto t \} \cup \{ x \stackrel{\text{?}}{=} t \} \text{ if } x \in Vars(G) \and x \notin Vars(t)$

$G \cup \{ x \stackrel{\text{?}}{=} f(s_0,\dots,s_k) \} \Rightarrow \bot \text{ if } x \in Vars(f(s_0,\dots,s_k))$

Proof of Termination

For the proof of termination consider a 3-tuple <NUVN,NLHS,EQN> where NUVN is the number of non-unique variables, NLHS is the number of function symbols and constants on the LHS of potential equations, and EQN is the number of equations. The application of these rewrite rules in any order terminates because NUVN remains the same or is reduced by each rewrite. Where NUVN remains the same, NLHS remains the same or is reduced by each rewrite. Where NUVN remains the same and NLHS remains the same, EQN is reduced by each rewrite.

See also

• Admissible rule
• Semantic Unification
• Explicit substitution in lambda calculus

Notes

1. ^ Russell, Norvig: Artificial Intelligence, A Modern Approach, p. 277
2. ^ Warren Goldfarb: The undecidability of the second-order unification problem [1]
3. ^ Gérard Huet: The undecidability of unification in third order logic
4. ^ Gérard Huet: A Unification Algorithm for typed Lambda-Calculus []
5. ^ Martelli, Montanari: An Efficient Unification Algorithm
6. ^ Gérard Huet: Higher Order Unification 30 Years Later
7. ^ Gilles Dowek: Higher-Order Unification and Matching. Handbook of Automated Reasoning 2001: 1009-1062
8. ^ Dale Miller: A Logic Programming Language with Lambda-Abstraction, Function Variables, and Simple Unification, Journal of Logic and Computation, 1991, pp. 497--536 [2]
9. ^ "Claire Gardent, Michael Kohlhase, Karsten Konrad, A multi-level, Higher-Order Unification approach to ellipsis (1997)". http://en.scientificcommons.org/48410. 

References

• F. Baader and T. Nipkow, Term Rewriting and All That. Cambridge University Press, 1998.
• F. Baader and W. Snyder, Unification Theory. In J.A. Robinson and A. Voronkov, editors, Handbook of Automated Reasoning, volume I, pages 447–533. Elsevier Science Publishers, 2001.
• Joseph Goguen, What is Unification?.
• Alex Sakharov, "Unification" from MathWorld.