rename Node->StateNode
[sbp.git] / src / edu / berkeley / sbp / Forest.java
index 7fde86e..636ad4f 100644 (file)
@@ -1,78 +1,91 @@
+// Copyright 2006-2007 all rights reserved; see LICENSE file for BSD-style license
+
 package edu.berkeley.sbp;
-import edu.berkeley.sbp.*;
-import edu.berkeley.sbp.Sequence.Position;
 import edu.berkeley.sbp.util.*;
 import java.io.*;
 import java.util.*;
-import java.lang.reflect.*;
-
-/** an efficient representation of a collection of trees (Tomita's shared packed parse forest) */
-public abstract class Forest<T> /*extends PrintableTree<Forest.Body<T>>*/
-    implements IntegerMappable,
-               GraphViz.ToGraphViz {
 
-    private static int master_idx = 0;
-    private final int idx = master_idx++;
-    public int toInt() { return idx; }
+/**
+ *   <font color=blue>
+ *   An efficient representation of a collection of trees (Tomita's
+ *   shared packed parse forest).
+ *   </font>
+ */
+public abstract class Forest<NodeType> implements GraphViz.ToGraphViz {
 
     /** assume that this forest contains exactly one tree and return it; otherwise throw an exception */
-    public abstract Tree<T> expand1() throws Ambiguous;
+    public abstract Tree<NodeType> expand1() throws Ambiguous;
 
     /** expand this forest into a set of trees */
-    public void expand(HashSet<Tree<T>> ht) { expand(ht, new HashSet<Forest<T>>(), null); }
-    public abstract void expand(HashSet<Tree<T>> ht, HashSet<Forest<T>> ignore, Tree<T> bogus);
-    public abstract void gather(HashSet<Forest<T>> ignore);
-    public HashSet<Tree<T>> expand(boolean toss) {
-        HashSet<Tree<T>> ret = new HashSet<Tree<T>>();
-        expand(ret);
-        return ret;
+    public Iterable<Tree<NodeType>> expand() {
+        HashSet<Tree<NodeType>> ht = new HashSet<Tree<NodeType>>();
+        expand(ht, new HashSet<Forest<NodeType>>(), null);
+        return ht;
     }
 
-    public static <T> Forest<T> create(Input.Region loc, T tag, Forest<T>[] tokens, boolean unwrap) {
-        return new Body<T>(loc, tag, tokens, unwrap);
+    /** returns the input Region which this Forest was parsed from */
+    public abstract Input.Region getRegion();
+
+    // Package-Private //////////////////////////////////////////////////////////////////////////////
+
+    public static <NodeType> Forest<NodeType> create(Input.Region region, NodeType head, Forest<NodeType>[] children) {
+        return create(region, head, children, new boolean[children==null ? 0 : children.length]); }
+    public static <NodeType> Forest<NodeType> create(Input.Region region, NodeType head, Forest<NodeType>[] children, boolean[] lifts) {
+        if (region == null) throw new RuntimeException("invoked Forest.create(region=null) -- this should never happen");
+        return new One<NodeType>(region, head, children, lifts);
     }
-    public abstract void edges(GraphViz.Node n);
-    public boolean ambiguous() { return false; }
 
-    // Body //////////////////////////////////////////////////////////////////////////////
+    abstract void expand(HashSet<Tree<NodeType>> ht, HashSet<Forest<NodeType>> ignore, Tree<NodeType> bogus);
+    abstract void gather(HashSet<Forest<NodeType>> ignore);
+    abstract void edges(GraphViz.StateNode n);
+    boolean ambiguous() { return false; }
+    
+    // One //////////////////////////////////////////////////////////////////////////////
 
-    public /*protected*/ static class Body<T> extends Forest<T> /* extends PrintableTree<Forest<T>> implements */ {
+    /** A "single" forest with a head and child subforests */    
+    private static class One<NodeType> extends Forest<NodeType> {
 
         private final Input.Region      location;
-        private final T                 tag;
-        private final Forest<T>[]       tokens;
-        private final boolean           unwrap;
+        private final NodeType                head;
+        private final Forest<NodeType>[]       children;
+
+        /** if true, the last child's children are considered children of this node */
+        private final boolean[]         lifts;
+
+        public Input.Region getRegion() { return location; }
 
-        private Body(Input.Region loc, T tag, Forest<T>[] tokens, boolean unwrap) {
+        private One(Input.Region loc, NodeType head, Forest<NodeType>[] children, boolean[] lifts) {
             this.location = loc;
-            this.tag = tag;
-            this.tokens = tokens==null ? emptyForestArray : new Forest[tokens.length];
-            if (tokens != null) System.arraycopy(tokens, 0, this.tokens, 0, tokens.length);
-            if (tokens != null) for(int i=0; i<tokens.length; i++) if (tokens[i]==null) throw new Error(i+"");
-            this.unwrap = unwrap;
+            this.head = head;
+            if (head==null) throw new RuntimeException("invoked Forest.create(,null,,,) -- this should never happen");
+            this.children = children==null ? emptyForestArray : new Forest[children.length];
+            if (children != null) System.arraycopy(children, 0, this.children, 0, children.length);
+            if (children != null) for(int i=0; i<children.length; i++) if (children[i]==null) throw new Error(i+"");
+            this.lifts = lifts;
         }
 
-        public Tree<T> expand1() throws Ambiguous {
-            Tree<T>[] ret = new Tree[tokens.length];
-            for(int i=0; i<tokens.length; i++)
-                ret[i] = tokens[i].expand1();
-            return new Tree<T>(location, tag, ret, unwrap);
+        public Tree<NodeType> expand1() throws Ambiguous {
+            Tree<NodeType>[] ret = new Tree[children.length];
+            for(int i=0; i<children.length; i++) ret[i] = children[i].expand1();
+            return new Tree<NodeType>(location, head, ret, lifts);
         }
-        public void gather(HashSet<Forest<T>> hf) {
+
+        void gather(HashSet<Forest<NodeType>> hf) {
             hf.add(this);
-            for(Forest<T> f : tokens) f.gather(hf);
+            for(Forest<NodeType> f : children) f.gather(hf);
         }
-        public void expand(HashSet<Tree<T>> ht, HashSet<Forest<T>> ignore, Tree<T> bogus) {
+        void expand(HashSet<Tree<NodeType>> ht, HashSet<Forest<NodeType>> ignore, Tree<NodeType> bogus) {
             if (ignore.contains(this)) { ht.add(bogus); return; }
-            expand(0, new Tree[tokens.length], ht, ignore, bogus);
+            expand(0, new Tree[children.length], ht, ignore, bogus);
         }
-        public void expand(final int i, Tree<T>[] ta, HashSet<Tree<T>> ht, HashSet<Forest<T>> ignore, Tree<T> bogus) {
-            if (i==tokens.length) {
-                ht.add(new Tree<T>(location, tag, ta, unwrap));
+        private void expand(final int i, Tree<NodeType>[] ta, HashSet<Tree<NodeType>> ht, HashSet<Forest<NodeType>> ignore,
+                            Tree<NodeType> bogus) {
+            if (i==children.length) {
+                ht.add(new Tree<NodeType>(location, head, ta, lifts));
             } else {
-                HashSet<Tree<T>> ht2 = new HashSet<Tree<T>>();
-                tokens[i].expand(ht2, ignore, bogus);
-                for(Tree<T> tc : ht2) {
+                HashSet<Tree<NodeType>> ht2 = new HashSet<Tree<NodeType>>();
+                children[i].expand(ht2, ignore, bogus);
+                for(Tree<NodeType> tc : ht2) {
                     ta[i] = tc;
                     expand(i+1, ta, ht, ignore, bogus);
                     ta[i] = null;
@@ -84,28 +97,28 @@ public abstract class Forest<T> /*extends PrintableTree<Forest.Body<T>>*/
 
         public boolean isTransparent() { return false; }
         public boolean isHidden() { return false; }
-        public GraphViz.Node toGraphViz(GraphViz gv) {
+        public GraphViz.StateNode toGraphViz(GraphViz gv) {
             if (gv.hasNode(this)) return gv.createNode(this);
-            GraphViz.Node n = gv.createNode(this);
+            GraphViz.StateNode n = gv.createNode(this);
             n.label = headToString()==null?"":headToString();
             n.directed = true;
             edges(n);
             return n;
         }
-        boolean edges = false;
-        public void edges(GraphViz.Node n) {
+        boolean edges = false; // FIXME ??
+        public void edges(GraphViz.StateNode n) {
             if (edges) return;
             edges = true;
-            for(int i=0; i<tokens.length; i++) {
-                if (i==tokens.length-1 && unwrap && !tokens[i].ambiguous()) {
-                    tokens[i].edges(n);
+            for(int i=0; i<children.length; i++) {
+                if (lifts[i] && !children[i].ambiguous()) {
+                    children[i].edges(n);
                 } else {
-                    n.edge(tokens[i], null);
+                    n.edge(children[i], null);
                 }
             }
         }
 
-        protected String  headToString()         { return tag==null?null:tag.toString(); }
+        protected String  headToString()         { return head==null?null:head.toString(); }
         protected String  headToJava()           { return "null"; }
         protected String  left()                 { return "{"; }
         protected String  right()                { return "}"; }
@@ -113,59 +126,98 @@ public abstract class Forest<T> /*extends PrintableTree<Forest.Body<T>>*/
     }
 
 
-    // Ref //////////////////////////////////////////////////////////////////////////////
+    // Many //////////////////////////////////////////////////////////////////////////////
+
+    /** An "ambiguity node"; this is immutable once it has been "looked at" */
+    static class Many<NodeType> extends Forest<NodeType> {
+
+        private FastSet<Forest<NodeType>> hp = new FastSet<Forest<NodeType>>();
+        private boolean touched = false;
 
-    /**
-     *  This class represents a partially complete collection of
-     *  forests to be viewed as a forest at some later date; once
-     *  viewed, it becomes immutable
-     */
-    static class Ref<T> extends Forest<T> {
-        public HashSet<GSS.Phase.Node> parents = new HashSet<GSS.Phase.Node>();
-        private FastSet<Forest<T>> hp = new FastSet<Forest<T>>();
+        public Many() { }
 
-        public Ref() { }
+        public Input.Region getRegion() { return hp.iterator().next().getRegion(); } // all should be identical
 
-        public Tree<T> expand1() throws Ambiguous {
-            if (hp.size() > 1) throw new Ambiguous(this);
+        public Tree<NodeType> expand1() throws Ambiguous {
+            touched();
+            if (hp.size() > 1) {
+                HashSet<Forest<NodeType>> hf0 = new HashSet<Forest<NodeType>>();
+                Iterator<Forest<NodeType>> ih = hp.iterator();
+                ih.next().gather(hf0);
+                for(Forest<NodeType> f : hp) {
+                    HashSet<Forest<NodeType>> hf1 = new HashSet<Forest<NodeType>>();
+                    f.gather(hf1);
+                    hf0.retainAll(hf1);
+                }
+                HashSet<Tree<NodeType>> ht = new HashSet<Tree<NodeType>>();
+                expand(ht, hf0, new Tree(null, "*"));
+                throw new Ambiguous((Forest<?>)this,
+                                    (HashSet<Tree<?>>)(Object)ht);
+            }
             return hp.iterator().next().expand1();
         }
         
-        public void gather(HashSet<Forest<T>> ht) {
+        void gather(HashSet<Forest<NodeType>> ht) {
+            touched();
+
+            // FIXME: do something more sensible here
+            if (ht.contains(this)) {
+                System.err.println("WARNING: grammar produced a circular forest\n" + this);
+                //throw new Error("grammar produced a circular forest:\n" + this);
+                return;
+            }
+
             ht.add(this);
-            for(Forest<T> f : hp) f.gather(ht);
+            for(Forest<NodeType> f : hp) f.gather(ht);
         }
 
-        public Forest resolve() { return this; }
-        public void expand(HashSet<Tree<T>> ht, HashSet<Forest<T>> ignore, Tree<T> bogus) {
-            if (ignore.contains(this)) { ht.add(bogus); return; }
-            for (Forest<T> f : hp) f.expand(ht, ignore, bogus);
+        private void touched() {
+            if (touched) return;
+            touched = true;
+            /*
+            FastSet<Forest<NodeType>> f2 = new FastSet<Forest<NodeType>>();
+            for(Forest f : hp)
+                if (f instanceof Forest.One) f2.add(f);
+                else for(Forest ff : ((Forest.Many<NodeType>)f))
+                    f2.add(ff);
+            hp = f2;
+            */
+        }
+        public boolean contains(Forest f) {
+            touched();
+            return hp.contains(f);
+        }
+        public void merge(Forest p) { 
+            if (touched) throw new RuntimeException("attempt to merge() on a Forest.Many that has already been examined");
+            if (p==this) throw new RuntimeException("attempt to merge() a Forest.Many to itself!");
+            hp.add(p, true);
         }
-        public boolean contains(Forest f) { return hp.contains(f); }
-        public void merge(Forest p) { if (p!=this) hp.add(p, true); }
-        public boolean ambiguous() {
+        boolean ambiguous() {
+            touched();
             if (hp.size()==0) return false;
             if (hp.size()==1) return hp.iterator().next().ambiguous();
             return true;
         }
 
+        void expand(HashSet<Tree<NodeType>> ht, HashSet<Forest<NodeType>> ignore, Tree<NodeType> bogus) {
+            touched();
+            if (ignore.contains(this)) { ht.add(bogus); return; }
+            for (Forest<NodeType> f : hp) f.expand(ht, ignore, bogus);
+        }
+
+
         // GraphViz, ToInt //////////////////////////////////////////////////////////////////////////////
 
-        public int toInt() {
-            if (hp.size()==1) return hp.iterator().next().toInt();
-            return super.toInt();
-        }
         public boolean isTransparent() { return hp.size()==1; }
         public boolean isHidden() { return hp.size()==0; }
-        public void edges(GraphViz.Node n) {
+        public void edges(GraphViz.StateNode n) {
             if (hp.size()==1) { hp.iterator().next().edges(n); return; }
             for(Forest f : hp) f.edges(n);
         }
-        public GraphViz.Node toGraphViz(GraphViz gv) {
-            //if (hp.size()==0) return null;
+        public GraphViz.StateNode toGraphViz(GraphViz gv) {
             if (hp.size()==1) return hp.iterator().next().toGraphViz(gv);
             if (gv.hasNode(this)) return gv.createNode(this);
-            GraphViz.Node n = gv.createNode(this);
+            GraphViz.StateNode n = gv.createNode(this);
             n.label = "?";
             n.color = "red";
             for(Forest f : hp) n.edge(f, null);