+// 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.*;
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> {
+/**
+ * <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 final Tree<T> expand1() throws Parser.Ambiguous, Parser.Failed {
- Iterator<Tree<T>> it = expand(true).iterator();
- if (!it.hasNext()) throw new Parser.Failed();
- return it.next();
- }
+ public abstract Tree<NodeType> expand1() throws Ambiguous;
/** expand this forest into a set of trees */
- public abstract HashSet<Tree<T>> expand(boolean toss);
+ public Iterable<Tree<NodeType>> expand() {
+ HashSet<Tree<NodeType>> ht = new HashSet<Tree<NodeType>>();
+ expand(ht, new HashSet<Forest<NodeType>>(), null);
+ return ht;
+ }
- abstract boolean valid();
+ /** returns the input Region which this Forest was parsed from */
+ public abstract Input.Region getRegion();
- static <T> Forest<T> singleton(Token.Location loc, Sequence creator) { return create(loc, null, new Forest[] { }, creator, false, true); }
- static <T> Forest<T> singleton(Token.Location loc, Forest<T> body, Sequence creator) { return create(loc, null, new Forest[] { body }, creator, false, true); }
- static <T> Forest<T> leaf(Token.Location loc, T tag, Sequence creator) { return create(loc, tag, null, creator, false, false); }
- static <T> Forest<T> create(Token.Location loc, T tag, Forest<T>[] tokens, Sequence creator, boolean unwrap, boolean singleton) {
- return new MultiForest<T>(loc, tag, tokens, creator, unwrap, singleton);
+ // 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);
}
- // 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.Node n);
+ boolean ambiguous() { return false; }
+
+ // One //////////////////////////////////////////////////////////////////////////////
- protected static class Body<T> {
+ /** A "single" forest with a head and child subforests */
+ private static class One<NodeType> extends Forest<NodeType> {
- private final Token.Location location;
- private final T tag;
- private final Forest<T>[] tokens;
- private final Sequence creator;
- private final boolean unwrap;
- private final boolean singleton;
+ private final Input.Region location;
+ private final NodeType head;
+ private final Forest<NodeType>[] children;
- private Body(Token.Location loc, T tag, Forest<T>[] tokens, Sequence creator, boolean unwrap, boolean singleton) {
- 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);
- this.creator = creator;
- this.unwrap = unwrap;
- this.singleton = singleton;
- }
+ /** if true, the last child's children are considered children of this node */
+ private final boolean[] lifts;
- private HashSet<Tree<T>> expand(boolean toss, ArrayList<Tree<T>> toks, int i, HashSet<Tree<T>> h) {
- if (singleton) {
- for(Body<T> b : (IterableForest<T>)tokens[0]) b.expand(toss, toks, i, h);
+ public Input.Region getRegion() { return location; }
- } else if (i==tokens.length) {
- h.add(new Tree<T>(null, tag, toks.toArray(tree_hint)));
+ private One(Input.Region loc, NodeType head, Forest<NodeType>[] children, boolean[] lifts) {
+ this.location = loc;
+ 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;
+ }
- } else if (unwrap && i==tokens.length-1) {
- if (tokens[i] != null)
- for(Body b : (IterableForest<T>)tokens[i])
- b.expand(toss, toks, 0, h);
+ 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);
+ }
+ void gather(HashSet<Forest<NodeType>> hf) {
+ hf.add(this);
+ for(Forest<NodeType> f : children) f.gather(hf);
+ }
+ 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[children.length], ht, ignore, bogus);
+ }
+ 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 {
- if (tokens[i]!=null) {
- HashSet<Tree<T>> exp = tokens[i].expand(toss);
- if (exp != null)
- for(Tree<T> r : exp) {
- int old = toks.size();
- toks.add(r);
- expand(toss, toks, i+1, h);
- while(toks.size() > old) toks.remove(toks.size()-1);
- }
+ 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;
}
}
- return h;
- }
- void addTo(HashSet<Body> h) {
- if (!singleton) h.add(this);
- else for(Body b : (IterableForest<T>)tokens[0]) b.addTo(h);
- }
- void addTo(FastSet<Body> h) {
- if (!singleton) h.add(this, true);
- else for(Body b : (IterableForest<T>)tokens[0]) b.addTo(h);
- }
-
- private boolean kcache = false;
- private boolean keep = false;
- public boolean keep() {
- if (kcache) return keep;
- kcache = true;
- for(Forest<T> token : tokens) if (!token.valid()) return keep = false;
- return keep = creator==null || (creator.needs.size()==0 && creator.hates.size()==0);
- }
- public boolean keep(Iterable<Body<T>> h) {
- if (keep()) return true;
- for(Forest<T> token : tokens) if (!token.valid()) return false;
- int needs = 0;
- for(Body<T> b : h) {
- if (creator.hates.contains(b.creator) && b.keep(h)) return false;
- if (creator.needs.contains(b.creator) && b.keep(h)) needs--;
- }
- return needs <= -1 * creator.needs.size();
}
-
- public String toString() {
- StringBuffer ret = new StringBuffer();
- for(int i=0; i<tokens.length; i++) {
- String q = tokens[i]==null ? "null" : tokens[i].toString();
- if (q.length() > 0) {
- ret.append(q);
- ret.append(" ");
+ // GraphViz, ToInt //////////////////////////////////////////////////////////////////////////////
+
+ public boolean isTransparent() { return false; }
+ public boolean isHidden() { return false; }
+ public GraphViz.Node toGraphViz(GraphViz gv) {
+ if (gv.hasNode(this)) return gv.createNode(this);
+ GraphViz.Node n = gv.createNode(this);
+ n.label = headToString()==null?"":headToString();
+ n.directed = true;
+ edges(n);
+ return n;
+ }
+ boolean edges = false; // FIXME ??
+ public void edges(GraphViz.Node n) {
+ if (edges) return;
+ edges = true;
+ for(int i=0; i<children.length; i++) {
+ if (lifts[i] && !children[i].ambiguous()) {
+ children[i].edges(n);
+ } else {
+ n.edge(children[i], null);
}
}
- String tail = ret.toString().trim();
- String head = (tag!=null && !tag.toString().equals("")) ? (tail.length() > 0 ? tag+":" : tag+"") : "";
- if (tail.length() > 0) tail = "{" + tail + "}";
- return head + tail;
}
+
+ protected String headToString() { return head==null?null:head.toString(); }
+ protected String headToJava() { return "null"; }
+ protected String left() { return "{"; }
+ protected String right() { return "}"; }
+ protected boolean ignoreSingleton() { return false; }
}
- // Ref //////////////////////////////////////////////////////////////////////////////
+ // Many //////////////////////////////////////////////////////////////////////////////
- static abstract class IterableForest<T> extends Forest<T> implements Iterable<Forest.Body<T>> {
- public abstract Iterator<Forest.Body<T>> iterator();
- }
+ /** An "ambiguity node"; this is immutable once it has been "looked at" */
+ static class Many<NodeType> extends Forest<NodeType> {
- /**
- * 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 IterableForest<T> {
- private FastSet<Forest> hp = new FastSet<Forest>();
- private Forest res = null;
- public boolean valid = false;
- public Ref() { }
- public void merge(Forest p) {
- if (res != null) throw new Error("already resolved!");
- if (p==null) throw new Error();
- if (p!=this) hp.add(p);
- }
- public Iterator<Body<T>> iterator() { return ((IterableForest<T>)resolve()).iterator(); }
- public HashSet<Tree<T>> expand(boolean toss) { return resolve().expand(toss); }
- public boolean valid() { resolve(); return valid; }
- public String toString() { return resolve().toString(); }
- public Forest resolve() {
- if (hp==null) return res;
- HashSet<Body> results = null;
- FastSet<Body> nh = new FastSet<Body>();
- for(Forest<?> p : hp)
- for(Body<?> b : (IterableForest<?>)p) {
- if (b.keep() && (b.creator==null || !b.creator.lame)) { valid = true; b.addTo(nh); }
- else results = new HashSet<Body>();
- }
- if (results != null) {
- for(Forest<?> p : hp) for(Body<?> b : (IterableForest<?>)p) results.add(b);
- for(Body b : results) {
- if (b.keep() && (b.creator==null || !b.creator.lame)) continue;
- if (!b.keep(results)) continue;
- if (b.creator!=null && b.creator.lame) continue;
- valid = true;
- b.addTo(nh);
+ private FastSet<Forest<NodeType>> hp = new FastSet<Forest<NodeType>>();
+ private boolean touched = false;
+
+ public Many() { }
+
+ public Input.Region getRegion() { return hp.iterator().next().getRegion(); } // all should be identical
+
+ 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);
}
- hp = null;
- return res = new MultiForest(nh, valid);
+ return hp.iterator().next().expand1();
}
- }
+
+ 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;
+ }
- // Implementations //////////////////////////////////////////////////////////////////////////////
+ ht.add(this);
+ for(Forest<NodeType> f : hp) f.gather(ht);
+ }
- private static class MultiForest<T> extends IterableForest<T> {
- private final FastSet<Body<T>> results;
- private boolean valid;
- public boolean valid() { return valid; }
- private MultiForest(FastSet<Body<T>> results, boolean valid) { this.results = results; this.valid = valid; }
- public MultiForest(Token.Location loc, T tag, Forest<T>[] tokens, Sequence creator, boolean unwrap, boolean singleton) {
- this.results = new FastSet<Body<T>>(new Body(loc, tag, tokens, creator, unwrap, singleton));
- this.valid = true;
+ 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);
+ }
+ boolean ambiguous() {
+ touched();
+ if (hp.size()==0) return false;
+ if (hp.size()==1) return hp.iterator().next().ambiguous();
+ return true;
}
- public Iterator<Body<T>> iterator() { return results.iterator(); }
- public HashSet<Tree<T>> expand(boolean toss) {
- HashSet<Tree<T>> ret = new HashSet<Tree<T>>();
- for(Body<T> b : results)
- ret.addAll(b.expand(toss, new ArrayList<Tree<T>>(), 0, new HashSet<Tree<T>>()));
- if (toss && ret.size() > 1) throw new Parser.Ambiguous(this);
- return ret;
+ 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);
}
-
- // Display //////////////////////////////////////////////////////////////////////////////
-
- private String toString = null;
- public String toString() {
- if (toString != null) return toString;
- StringBuffer ret = new StringBuffer();
- ret.append("<?");
- boolean first = true;
- for(Forest.Body<T> r : results) {
- if (!first) ret.append(' ');
- first = false;
- ret.append(r);
- }
- ret.append("?>");
- return toString = ret.toString();
+
+
+ // GraphViz, ToInt //////////////////////////////////////////////////////////////////////////////
+
+ public boolean isTransparent() { return hp.size()==1; }
+ public boolean isHidden() { return hp.size()==0; }
+ public void edges(GraphViz.Node 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()==1) return hp.iterator().next().toGraphViz(gv);
+ if (gv.hasNode(this)) return gv.createNode(this);
+ GraphViz.Node n = gv.createNode(this);
+ n.label = "?";
+ n.color = "red";
+ for(Forest f : hp) n.edge(f, null);
+ return n;
}
}
// Statics //////////////////////////////////////////////////////////////////////////////
private static Tree[] tree_hint = new Tree[0];
- private static Body[] body_hint = new Body[0];
+ private static String[] string_hint = new String[0];
private static final Forest[] emptyForestArray = new Forest[0];
+
+ protected String headToString() { return null; }
+ protected String headToJava() { return "null"; }
+ protected String left() { return "<?"; }
+ protected String right() { return "?>"; }
+ protected boolean ignoreSingleton() { return true; }
}