mps-coderules/reactor/code/src/jetbrains/mps/unification/UnionFindTermGraphUnifier.java

336 lines
9.3 KiB
Java

/*
* Copyright 2015 JetBrains s.r.o.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package jetbrains.mps.unification;
import jetbrains.mps.unification.Substitution.FailureCause;
import jetbrains.mps.unification.Unification.SuccessfulSubstitution;
import java.util.*;
import static jetbrains.mps.unification.Term.Kind.*;
import static jetbrains.mps.unification.Substitution.FailureCause.*;
import static jetbrains.mps.unification.Unification.*;
/**
* This is an implementation of the "near linear" algorithm for solving syntactic unification
* as described in the paper linked below and also in the textbook of the same author.<sup>1</sup> <sup>2</sup>
*
* No recursive terms are allowed as a solution, meaning the "occurrs check" for variables
* is performed on the input. However, cyclic terms are allowed as input and can be unified, producing
* solutions binding variables to cyclic terms. Variables can also be passed by reference without altering the
* intuitive behaviour of the algorithm.
*
* If successful, the returned {@link Substitution} contains
* the variable bindings.
* <p/>
* The variables are sorted using {@link java.lang.Comparable} to ensure uniqueness of bindings
* whereas the substituted term is also a variable.
* <p/>
* <blockquote>
* 1. <i>Baader, Franz, and Wayne Snyder. "Unification Theory." Handbook of automated reasoning 1 (2001): 445-532.</i>
* 2. <i>Baader, Franz, and Tobias Nipkow. Term rewriting and all that. Cambridge University Press, 1999.</i>
* </blockquote>
*
* @author Fedor Isakov
*/
public class UnionFindTermGraphUnifier {
private Map<Object, Data> myData = new IdentityHashMap<Object, Data>();
private FailureCause myFailureCause = UKNOWN;
public Substitution unify(Term a, Term b) {
if (!unifClosure(a, b)) {
return failedSubstitution(myFailureCause);
}
return findSolution(a);
}
private boolean unifClosure(Term s, Term t) {
s = find(s);
t = find(t);
if (s == t) return true;
Term zs = getSchema(s);
Term zt = getSchema(t);
// a VAR always matches another node
if(zs.is(VAR) || zt.is(VAR)) {
union(s, t);
return true;
}
// dereference REF nodes
Term ds = zs.is(REF) ? zs.get() : zs;
Term dt = zt.is(REF) ? zt.get() : zt;
if (ds.is(VAR)) {
if (s != find(ds)) {
union(s, find(ds));
}
union(s, t);
return true;
}
else {
zs = ds;
}
if (dt.is(VAR)) {
if (t != find(dt)) {
union(t, find(dt));
}
union(t, s);
return true;
}
else {
zt = dt;
}
// use find 2nd time to account for dereferenced nodes
if (find(zs) == find(zt)) {
return true;
}
if (zs.is(FUN) && zt.is(FUN))
{
if (!eq(zs.symbol(), zt.symbol())) {
myFailureCause = SYMBOL_CLASH;
return false; // symbol clash
}
// union REF nodes only to each other
if (s.is(REF) == t.is(REF)) {
union(s, t);
}
Iterator<? extends Term> scit = zs.arguments().iterator();
Iterator<? extends Term> tcit = zt.arguments().iterator();
while (scit.hasNext() && tcit.hasNext()) {
if (!unifClosure(scit.next(), tcit.next())) {
return false; // arguments mismatch
}
}
// fail if different arguments count
return scit.hasNext() == tcit.hasNext();
}
else {
// something's wrong with the input
throw new IllegalStateException("invalid input");
}
}
private void union(Term s, Term t) {
int ssize = getSize(s);
int tsize = getSize(t);
// keep the order: the smaller class gets inserted under the bigger one
if (ssize < tsize) {
Term tmp = t; t = s; s = tmp;
}
else if (ssize == tsize && s.is(VAR) && t.is(VAR)) {
// ensure proper order of variables in the substitution
if(t.compareTo(s) < 0) {
Term tmp = t; t = s; s = tmp;
}
}
// union s and t classes by moving t under s
setSize(s, ssize + tsize);
appendVars(s, getVars(t));
Term zs = getSchema(s);
Term zt = getSchema(t);
if (zs.is(VAR) || (zs.is(REF) && zs.get().is(VAR) && !zt.is(VAR)))
{
setSchema(s, zt);
}
setRepresentative(t, s);
}
private Term find(Term s) {
Term term = getRepresentative(s);
if (term == s) {
return s;
}
// find representative and compress paths
List<Term> path = new ArrayList<Term>();
path.add(term);
for (Term t; (t = getRepresentative(term)) != term; ) {
path.add(t);
term = t;
}
for (Term p : path) {
setRepresentative(p, term);
}
return term;
}
private Substitution findSolution(Term s) {
return findSolution(s, EMPTY_SUBSTITUTION);
}
private Substitution findSolution(Term s, Substitution substitution) {
Term z = getSchema(find(s));
if (isAcyclic(z)) {
return substitution; // not part of a cycle
}
if (isVisited(z)) {
return failedSubstitution(CYCLE_DETECTED); // there exists a cycle
}
if (z.is(FUN)) {
setVisited(z, true);
for (Term c : z.arguments()) {
substitution = findSolution(c, substitution);
if (!substitution.isSuccessful()) {
break;
}
}
setVisited(z, false);
}
if (!substitution.isSuccessful()) {
return substitution;
}
setAcyclic(z, true);
SuccessfulSubstitution success = new SuccessfulSubstitution(substitution);
for (Term var : getVars(find(z))) {
if (var != z) {
Term val = z.is(REF) ? z.get() : z;
// Keep the order of variables within a binding
if (val.is(VAR) && val.compareTo(var) < 0) {
success.addBinding(val, var);
}
else {
success.addBinding(var, val);
}
}
}
return success;
}
private int getSize(Term n) {
if (!hasData(n)) return 1;
return getData(n).mySize;
}
private void setSize(Term n, int size) {
getData(n).mySize = size;
}
private Term getRepresentative(Term n) {
if (!hasData(n)) return n;
return getData(n).myClass;
}
private void setRepresentative(Term n, Term rep) {
getData(n).myClass = rep;
}
private Term getSchema(Term n) {
if (!hasData(n)) return n;
return getData(n).mySchema;
}
private void setSchema(Term n, Term schema) {
getData(n).mySchema = schema;
}
private List<Term> getVars(Term n) {
return hasData(n) ? getData(n).myVars : singletonVar(n);
}
private void appendVars(Term n, List<Term> vars) {
List<Term> newVars = new ArrayList<Term>(getVars(n));
newVars.addAll(vars);
getData(n).myVars = newVars;
}
private boolean isAcyclic(Term n) {
return hasData(n) && getData(n).myAcyclic;
}
private void setAcyclic(Term n, boolean acyclic) {
getData(n).myAcyclic = acyclic;
}
private boolean isVisited(Term n) {
return hasData(n) && getData(n).myVisited;
}
private void setVisited(Term n, boolean visited) {
getData(n).myVisited = visited;
}
private boolean hasData(Term n) {
Object key = n.is(VAR) ? String.valueOf(n.symbol()).intern() : n;
return myData.containsKey(key);
}
private Data getData(Term n) {
Object key = n.is(VAR) ? String.valueOf(n.symbol()).intern() : n;
if (myData.containsKey(key)) return myData.get(key);
Data data = new Data(n);
myData.put(key, data);
return data;
}
private static boolean eq(Object a, Object b) {
return a == null ? b == null : a.equals(b);
}
private static List<Term> singletonVar(Term n) {
return n.is(VAR) ?
Collections.singletonList(n) :
Collections.<Term>emptyList();
}
private static class Data {
int mySize = 1;
boolean myAcyclic = false;
boolean myVisited = false;
boolean myReferenced = false;
List<Term> myVars;
Term myClass;
Term mySchema;
Data(Term n) {
myClass = n;
mySchema = n;
myVars = singletonVar(n);
}
}
}