package edu.berkeley.sbp;
import edu.berkeley.sbp.*;
import edu.berkeley.sbp.util.*;
-import edu.berkeley.sbp.*;
import edu.berkeley.sbp.Sequence.Position;
-import edu.berkeley.sbp.*;
import java.io.*;
import java.util.*;
-import java.lang.reflect.*;
/** a parser which translates streams of Tokens of type T into a Forest<R> */
public abstract class Parser<T extends Token, R> {
- public final Table pt;
+ private final Table pt;
/** create a parser to parse the grammar with start symbol <tt>u</tt> */
protected Parser(Union u) { this.pt = new Table(u, top()); }
protected Parser(Table pt) { this.pt = pt; }
+ /** implement this method to create the output forest corresponding to a lone shifted input token */
public abstract Forest<R> shiftedToken(T t, Token.Location loc);
- public abstract Topology<T> top();
-
- /** parse <tt>input</tt> for a exactly one unique result, throwing <tt>Ambiguous</tt> if not unique or <tt>Failed</tt> if none */
- public Tree<R> parse1(Token.Stream<T> input) throws IOException, Failed, Ambiguous {
- Forest<R> ret = parse(input);
- try { return ret.expand1(); }
- catch (Ambiguous a) {
- System.out.println("while expanding:");
- System.out.println(ret);
- throw a;
- }
- }
+ /** this method must return an empty topology of the input token type */
+ public abstract Topology<T> top();
/** parse <tt>input</tt>, using the table <tt>pt</tt> to drive the parser */
- public Forest<R> parse(Token.Stream<T> input) throws IOException, Failed {
+ public Forest<R> parse(Token.Stream<T> input) throws IOException, ParseFailed {
GSS gss = new GSS();
Token.Location loc = input.getLocation();
GSS.Phase current = gss.new Phase(null, this, null, input.next(1, 0, 0), loc, null);
int count = 1;
for(;;) {
loc = input.getLocation();
- //current.checkFailure();
current.reduce();
Forest forest = current.token==null ? null : shiftedToken((T)current.token, loc);
GSS.Phase next = gss.new Phase(current, this, current, input.next(count, gss.resets, gss.waits), loc, forest);
current = next;
}
}
-
-
- // Exceptions //////////////////////////////////////////////////////////////////////////////
-
- public static class Failed extends RuntimeException {
- private final Token.Location location;
- private final String message;
- public Failed() { this("", null); }
- public Failed(String message, Token.Location loc) { this.location = loc; this.message = message; }
- public Token.Location getLocation() { return location; }
- public String toString() { return message/* + (location==null ? "" : (" at " + location))*/; }
- }
-
- public static class Ambiguous extends RuntimeException {
- public final Forest ambiguity;
- public Ambiguous(Forest ambiguity) { this.ambiguity = ambiguity; }
- public String toString() {
- StringBuffer sb = new StringBuffer();
- sb.append("unresolved ambiguity "/*"at " + ambiguity.getLocation() + ":"*/);
- for(Object result : ambiguity.expand(false))
- sb.append("\n " + result);
- return sb.toString();
- }
- }
-
// Table //////////////////////////////////////////////////////////////////////////////
public final Walk.Cache cache = this;
- public HashMapBag<Position,State> byPosition = new HashMapBag<Position,State>();
-
private void walk(Element e, HashSet<Element> hs) {
if (e==null) return;
if (hs.contains(e)) return;
HashSet<Element> all_elements = new HashSet<Element>();
walk(start0, all_elements);
for(Element e : all_elements)
- cache.ys.put(e, new Walk.YieldSet(e, cache).walk());
+ cache.ys.addAll(e, new Walk.YieldSet(e, cache).walk());
HashSet<Position> hp = new HashSet<Position>();
reachable(start0, hp);
this.start = new State(hp, all_states, all_elements);
if (start0.contains(p.owner()) && p.next()==null)
state.accept = true;
- if (p.isRightNullable(cache)) {
+ if (isRightNullable(p)) {
Walk.Follow wf = new Walk.Follow(top.empty(), p.owner(), all_elements, cache);
Reduction red = new Reduction(p);
Topology follow = wf.walk(p.owner());
- if (p.owner() instanceof Sequence.RewritingSequence &&
- (((Sequence.RewritingSequence)p.owner()).tag+"").equals("emailaddr")) {
- System.out.println("follow before: " + new edu.berkeley.sbp.misc.CharToken.CharRange(follow));
- }
for(Position p2 = p; p2 != null && p2.element() != null; p2 = p2.next())
follow = follow.intersect(new Walk.Follow(top.empty(), p2.element(), all_elements, cache).walk(p2.element()));
- if (p.owner() instanceof Sequence.RewritingSequence &&
- (((Sequence.RewritingSequence)p.owner()).tag+"").equals("emailaddr")) {
- System.out.println("follow after: " + new edu.berkeley.sbp.misc.CharToken.CharRange(follow));
- }
state.reductions.put(follow, red);
if (wf.includesEof()) state.eofReductions.add(red);
}
}
}
+ private boolean isRightNullable(Position p) {
+ if (p.isLast()) return true;
+ if (!p.element().possiblyEpsilon(this)) return false;
+ return isRightNullable(p.next());
+ }
+
/** a single state in the LR table and the transitions possible from it */
public class State implements Comparable<Table.State>, IntegerMappable, Iterable<Position> {
- public int toInt() { return idx; }
-
- public boolean lame() {
- for(Position p : this)
- for(Position p2 = p; p2!=null; p2=p2.next())
- if (p2.isLast() && !p2.owner().lame)
- return false;
- return true;
- }
- /*
- public boolean isResolvable(Token t) {
- boolean found = false;
- for(Reduction r : getReductions(t)) {
- Position p = r.position;
- if (!p.isRightNullable(cache)) continue;
- if (p.owner().firstp()==p) continue;
- if (found) {
- // found two items meeting criteria #1
- return false;
- } else {
- found = true;
- continue;
- }
- if (p.element()==null) continue;
- Topology first = new Walk.First(top(), cache).walk(p.element());
- if (first.contains(t))
- }
- }
- */
-
public final int idx = master_state_idx++;
private final HashSet<Position> hs;
// Interface Methods //////////////////////////////////////////////////////////////////////////////
- public boolean isAccepting() { return accept; }
-
- public boolean canShift(Token t) { return oshifts.contains(t); }
- public boolean canReduce(Token t) { return t==null ? eofReductions.size()>0 : oreductions.contains(t); }
-
+ boolean isAccepting() { return accept; }
public Iterator<Position> iterator() { return hs.iterator(); }
- public <B,C> void invokeShifts(Token t, Invokable<State,B,C> irbc, B b, C c) {
+ boolean canShift(Token t) { return oshifts.contains(t); }
+ <B,C> void invokeShifts(Token t, Invokable<State,B,C> irbc, B b, C c) {
oshifts.invoke(t, irbc, b, c);
}
- public <B,C> void invokeReductions(Token t, Invokable<Reduction,B,C> irbc, B b, C c) {
+
+ boolean canReduce(Token t) { return t==null ? eofReductions.size()>0 : oreductions.contains(t); }
+ <B,C> void invokeReductions(Token t, Invokable<Reduction,B,C> irbc, B b, C c) {
if (t==null) for(Reduction r : eofReductions) irbc.invoke(r, b, c);
else oreductions.invoke(t, irbc, b, c);
}
// register ourselves in the all_states hash so that no
// two states are ever created with an identical position set
all_states.put(hs, this);
- for(Position p : hs) byPosition.add(p,this);
// Step 1a: examine all Position's in this state and compute the mappings from
// sets of follow tokens (tokens which could follow this position) to sets
// "yields" [in one or more step] is used instead of "produces" [in exactly one step]
// to avoid having to iteratively construct our set of States as shown in most
// expositions of the algorithm (ie "keep doing XYZ until things stop changing").
- /*
- for(Element e : all_elements) {
- if (e instanceof Atom) continue;
- HashSet<Position> h = new Walk.Closure(null, g.cache).closure(e, hs);
- State s = all_states.get(h) == null ? new State(h, all_states, all_elements) : all_states.get(h);
- if (gotoSetNonTerminals.get(e) != null)
- throw new Error("this should not happen");
- gotoSetNonTerminals.put(e, s);
- }
- */
HashMapBag<Element,Position> move = new HashMapBag<Element,Position>();
for(Position p : hs) {
Element e = p.element();
if (e==null) continue;
- HashSet<Element> ys = cache.ys.get(e);
- if (ys != null) {
- for(Element y : ys) {
- HashSet<Position> hp = new HashSet<Position>();
- reachable(p.next(), hp);
- move.addAll(y, hp);
- }
+ for(Element y : cache.ys.getAll(e)) {
+ HashSet<Position> hp = new HashSet<Position>();
+ reachable(p.next(), hp);
+ move.addAll(y, hp);
}
}
for(Element y : move) {
}
public String toString() {
- //return "state["+idx+"]";
StringBuffer ret = new StringBuffer();
ret.append("state["+idx+"]: ");
for(Position p : this) ret.append("{"+p+"} ");
}
public int compareTo(Table.State s) { return idx==s.idx ? 0 : idx < s.idx ? -1 : 1; }
+ public int toInt() { return idx; }
}
/**
*/
public class Reduction {
// FIXME: cleanup; almost everything in here could go in either Sequence.Position.getRewrite() or else in GSS.Reduct
- public final int numPop;
/*private*/ final Position position;
private final Forest[] holder; // to avoid constant reallocation
public int hashCode() { return position.hashCode(); }
}
public Reduction(Position p) {
this.position = p;
- this.numPop = p.pos;
- this.holder = new Forest[numPop];
+ this.holder = new Forest[position.pos];
}
public String toString() { return "[reduce " + position + "]"; }
private Forest zero = null;
public Forest zero() {
if (zero != null) return zero;
- if (numPop > 0) throw new Error();
+ if (position.pos > 0) throw new Error();
return zero = position.rewrite(null);
}
public void reduce(GSS.Phase.Node parent) {
- if (numPop==0) finish(parent, zero(), parent.phase());
- else reduce(parent, numPop-1, parent.phase());
+ if (position.pos==0) finish(parent, zero(), parent.phase());
+ else reduce(parent, position.pos-1, parent.phase());
}
public void reduce(GSS.Phase.Node parent, GSS.Phase.Node onlychild) {
- if (numPop<=0) throw new Error("called wrong form of reduce()");
- int pos = numPop-1;
+ if (position.pos<=0) throw new Error("called wrong form of reduce()");
+ int pos = position.pos-1;
Forest old = holder[pos];
holder[pos] = parent.pending();
if (pos==0) {
State state = parent.state.gotoSetNonTerminals.get(position.owner());
if (result==null) throw new Error();
if (state!=null)
- target.newNode(parent, result, state, numPop<=0, this);
+ target.newNode(parent, result, state, position.pos<=0, this);
}
}
}