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> {
-
- private final Table pt;
+public abstract class Parser<Tok, Result> {
- private static void reachable(Element e, HashSet<Position> h) {
- if (e instanceof Atom) return;
- for(Sequence s : ((Union)e))
- reachable(s.firstp(), h);
- }
- private static void reachable(Position p, HashSet<Position> h) {
- if (h.contains(p)) return;
- h.add(p);
- if (p.element() != null) reachable(p.element(), h);
- }
+ protected final Table<Tok> 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; }
+ /** create a parser to parse the grammar with start symbol <tt>u</tt> */
+ protected Parser(Union u, Topology<Tok> top) { this.pt = new Table<Tok>(u, top); }
+ protected Parser(Table<Tok> pt) { this.pt = pt; }
- public abstract Forest<R> shiftedToken(T t);
- public abstract Topology<T> top();
+ /** implement this method to create the output forest corresponding to a lone shifted input token */
+ protected abstract Forest<Result> shiftToken(Tok t, Input.Location newloc);
+ boolean helpgc = true;
- /** 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 { return parse(input).expand1(); }
+ public String toString() { return pt.toString(); }
/** 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<Result> parse(Input<Tok> input) throws IOException, ParseFailed {
GSS gss = new GSS();
- GSS.Phase current = gss.new Phase(null, input.next());
- current.newNode(null, null, pt.start, true, null);
- for(;;) {
- GSS.Phase next = gss.new Phase(current, input.next());
+ Input.Location loc = input.getLocation();
+ GSS.Phase current = gss.new Phase<Tok>(null, this, null, input.next(), loc, null);
+ current.newNode(null, Forest.create(null, null, null, false), pt.start, true);
+ int count = 1;
+ for(int idx=0;;idx++) {
+ Input.Location oldloc = loc;
+ loc = input.getLocation();
current.reduce();
- Forest forest = current.token==null ? null : shiftedToken((T)current.token);
- current.shift(next, forest);
- if (current.isDone()) return (Forest<R>)current.finalResult;
- current.checkFailure();
+ Forest forest = current.token==null ? null : shiftToken((Tok)current.token, loc);
+ GSS.Phase next = gss.new Phase<Tok>(current, this, current, input.next(), loc, forest);
+ if (!helpgc) {
+ FileOutputStream fos = new FileOutputStream("out-"+idx+".dot");
+ PrintWriter p = new PrintWriter(new OutputStreamWriter(fos));
+ GraphViz gv = new GraphViz();
+ for(Object n : next)
+ ((GSS.Phase.Node)n).toGraphViz(gv);
+ gv.dump(p);
+ p.flush();
+ p.close();
+ }
+ count = next.size();
+ if (current.isDone()) return (Forest<Result>)gss.finalResult;
current = next;
}
}
-
- // Exceptions //////////////////////////////////////////////////////////////////////////////
+ // Table //////////////////////////////////////////////////////////////////////////////
- public static class Failed extends Exception {
- 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)); }
- }
+ /** an SLR(1) parse table which may contain conflicts */
+ static class Table<Tok> extends Walk.Cache {
- 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);
+ sb.append("parse table");
+ for(State<Tok> state : all_states.values()) {
+ sb.append(" " + state + "\n");
+ for(Topology<Tok> t : state.shifts) {
+ sb.append(" shift \""+
+ new edu.berkeley.sbp.chr.CharTopology((IntegerTopology<Character>)t)+"\" => ");
+ for(State st : state.shifts.getAll(t))
+ sb.append(st.idx+" ");
+ sb.append("\n");
+ }
+ for(Topology<Tok> t : state.reductions)
+ sb.append(" reduce \""+
+ new edu.berkeley.sbp.chr.CharTopology((IntegerTopology<Character>)t)+"\" => " +
+ state.reductions.getAll(t) + "\n");
+ }
return sb.toString();
}
- }
+ public final Walk.Cache cache = this;
- // Table //////////////////////////////////////////////////////////////////////////////
-
- static class Top extends Union { public Top() { super("0"); } }
-
- /** an SLR(1) parse table which may contain conflicts */
- static class Table {
-
- private final Union start0 = new Top();
- private final Sequence start0seq;
-
- public final Walk.Cache cache = new Walk.Cache();
-
- public HashSet<Position> closure() {
- HashSet<Position> hp = new HashSet<Position>();
- reachable(start0, hp);
- return hp;
- }
- public Position firstPosition() { return start0seq.firstp(); }
- public Position lastPosition() { Position ret = start0seq.firstp(); while(!ret.isLast()) ret = ret.next(); return ret; }
-
private void walk(Element e, HashSet<Element> hs) {
if (e==null) return;
if (hs.contains(e)) return;
hs.add(e);
if (e instanceof Atom) return;
- for(Sequence s : (Union)e) {
- hs.add(s);
- for(Position p = s.firstp(); p != null; p = p.next())
- walk(p.element(), hs);
- }
+ for(Sequence s : (Union)e)
+ walk(s, hs);
}
- public HashSet<Element> walk() {
- HashSet<Element> ret = new HashSet<Element>();
- walk(start0, ret);
- return ret;
+ private void walk(Sequence s, HashSet<Element> hs) {
+ hs.add(s);
+ for(Position p = s.firstp(); p != null; p = p.next())
+ walk(p.element(), hs);
+ for(Sequence ss : s.needs()) walk(ss, hs);
+ for(Sequence ss : s.hates()) walk(ss, hs);
}
- /*
- public String toString() {
- StringBuffer sb = new StringBuffer();
- for(Element e : walk())
- if (e instanceof Union)
- ((Union)e).toString(sb);
- return sb.toString();
- }
- */
-
/** the start state */
- public final State start;
+ public final State<Tok> start;
+
+ /** the state from which no reductions can be done */
+ private final State<Tok> dead_state;
/** used to generate unique values for State.idx */
private int master_state_idx = 0;
+ HashMap<HashSet<Position>,State<Tok>> all_states = new HashMap<HashSet<Position>,State<Tok>>();
/** construct a parse table for the given grammar */
public Table(Topology top) { this("s", top); }
public Table(String startSymbol, Topology top) { this(new Union(startSymbol), top); }
- public Table(Union u, Topology top) {
- start0seq = new Sequence.Singleton(u, null, null);
- start0.add(start0seq);
+ public Table(Union ux, Topology top) {
+ Union start0 = new Union("0");
+ start0.add(new Sequence.Singleton(ux));
+
+ for(Sequence s : start0) cache.eof.put(s, true);
+ cache.eof.put(start0, true);
// construct the set of states
- HashMap<HashSet<Position>,State> all_states = new HashMap<HashSet<Position>,State>();
- HashSet<Element> all_elements = walk();
+ 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());
- this.start = new State(closure(), all_states, all_elements);
+ cache.ys.addAll(e, new Walk.YieldSet(e, cache).walk());
+ HashSet<Position> hp = new HashSet<Position>();
+ reachable(start0, hp);
+
+ this.dead_state = new State<Tok>(new HashSet<Position>(), all_states, all_elements);
+ this.start = new State<Tok>(hp, all_states, all_elements);
// for each state, fill in the corresponding "row" of the parse table
- for(State state : all_states.values())
+ for(State<Tok> state : all_states.values())
for(Position p : state.hs) {
// the Grammar's designated "last position" is the only accepting state
- if (p==lastPosition())
+ if (start0.contains(p.owner()) && p.next()==null)
state.accept = true;
- // FIXME: how does right-nullability interact with follow restrictions?
- // all right-nullable rules get a reduction [Johnstone 2000]
- 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);
- state.reductions.put(wf.walk(p.owner()), red);
- if (wf.includesEof()) state.eofReductions.add(red, true);
+ Topology follow = wf.walk(p.owner());
+ 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()));
+ state.reductions.put(follow, p);
+ if (wf.includesEof()) state.eofReductions.add(p);
}
// if the element following this position is an atom, copy the corresponding
if (p.element() != null && p.element() instanceof Atom)
state.shifts.addAll(state.gotoSetTerminals.subset(((Atom)p.element())));
}
+ if (top instanceof IntegerTopology)
+ for(State<Tok> state : all_states.values()) {
+ state.oreductions = state.reductions.optimize(((IntegerTopology)top).functor());
+ state.oshifts = state.shifts.optimize(((IntegerTopology)top).functor());
+ }
+ }
+
+ private boolean isRightNullable(Position p) {
+ if (p.isLast()) return true;
+ if (!possiblyEpsilon(p.element())) 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>, Iterable<Position> {
+
+ class State<Tok> implements Comparable<State<Tok>>, IntegerMappable, Iterable<Position> {
- public final int idx = master_state_idx++;
+ public final int idx = master_state_idx++;
private final HashSet<Position> hs;
- private transient HashMap<Element,State> gotoSetNonTerminals = new HashMap<Element,State>();
- private transient TopologicalBag<Token,State> gotoSetTerminals = new TopologicalBag<Token,State>();
+ public transient HashMap<Element,State<Tok>> gotoSetNonTerminals = new HashMap<Element,State<Tok>>();
+ private transient TopologicalBag<Tok,State<Tok>> gotoSetTerminals = new TopologicalBag<Tok,State<Tok>>();
- private TopologicalBag<Token,Reduction> reductions = new TopologicalBag<Token,Reduction>();
- private FastSet<Reduction> eofReductions = new FastSet<Reduction>();
- private TopologicalBag<Token,State> shifts = new TopologicalBag<Token,State>();
+ private TopologicalBag<Tok,Position> reductions = new TopologicalBag<Tok,Position>();
+ private HashSet<Position> eofReductions = new HashSet<Position>();
+ private TopologicalBag<Tok,State<Tok>> shifts = new TopologicalBag<Tok,State<Tok>>();
private boolean accept = false;
+ private VisitableMap<Tok,State<Tok>> oshifts = null;
+ private VisitableMap<Tok,Position> oreductions = null;
+
// Interface Methods //////////////////////////////////////////////////////////////////////////////
- public boolean canShift(Token t) { return shifts.contains(t); }
- public Iterable<State> getShifts(Token t) { return shifts.get(t); }
- public boolean isAccepting() { return accept; }
- public Iterable<Reduction> getReductions(Token t) { return reductions.get(t); }
- public Iterable<Reduction> getEofReductions() { return eofReductions; }
- public Iterator<Position> iterator() { return hs.iterator(); }
+ boolean isAccepting() { return accept; }
+ public Iterator<Position> iterator() { return hs.iterator(); }
+
+ boolean canShift(Tok t) { return oshifts!=null && oshifts.contains(t); }
+ <B,C> void invokeShifts(Tok t, Invokable<State<Tok>,B,C> irbc, B b, C c) {
+ oshifts.invoke(t, irbc, b, c);
+ }
+
+ boolean canReduce(Tok t) { return oreductions != null && (t==null ? eofReductions.size()>0 : oreductions.contains(t)); }
+ <B,C> void invokeReductions(Tok t, Invokable<Position,B,C> irbc, B b, C c) {
+ if (t==null) for(Position r : eofReductions) irbc.invoke(r, b, c);
+ else oreductions.invoke(t, irbc, b, c);
+ }
// Constructor //////////////////////////////////////////////////////////////////////////////
* </ul>
*/
public State(HashSet<Position> hs,
- HashMap<HashSet<Position>,State> all_states,
+ HashMap<HashSet<Position>,State<Tok>> all_states,
HashSet<Element> all_elements) {
this.hs = hs;
// of _new_ positions (positions after shifting). These mappings are
// collectively known as the _closure_
- TopologicalBag<Token,Position> bag0 = new TopologicalBag<Token,Position>();
+ TopologicalBag<Tok,Position> bag0 = new TopologicalBag<Tok,Position>();
for(Position position : hs) {
if (position.isLast() || !(position.element() instanceof Atom)) continue;
Atom a = (Atom)position.element();
HashSet<Position> hp = new HashSet<Position>();
reachable(position.next(), hp);
- bag0.addAll(a, /*clo.walk()*/hp);
+ bag0.addAll(a, hp);
}
// Step 1b: for each _minimal, contiguous_ set of characters having an identical next-position
// set, add that character set to the goto table (with the State corresponding to the
// computed next-position set).
- for(Topology<Token> r : bag0) {
+ for(Topology<Tok> r : bag0) {
HashSet<Position> h = new HashSet<Position>();
for(Position p : bag0.getAll(r)) h.add(p);
- gotoSetTerminals.put(r, all_states.get(h) == null ? new State(h, all_states, all_elements) : all_states.get(h));
+ gotoSetTerminals.put(r, all_states.get(h) == null ? new State<Tok>(h, all_states, all_elements) : all_states.get(h));
}
// Step 2: for every non-Atom element (ie every Element which has a corresponding reduction),
// "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) {
+ OUTER: for(Element y : move) {
HashSet<Position> h = move.getAll(y);
- State s = all_states.get(h) == null ? new State(h, all_states, all_elements) : all_states.get(h);
+ State<Tok> s = all_states.get(h) == null ? new State<Tok>(h, all_states, all_elements) : all_states.get(h);
+ // if a reduction is "lame", it should wind up in the dead_state after reducing
+ if (y instanceof Sequence) {
+ for(Position p : hs) {
+ if (p.element() != null && (p.element() instanceof Union)) {
+ Union u = (Union)p.element();
+ for(Sequence seq : u)
+ if (seq.needs.contains((Sequence)y) || seq.hates.contains((Sequence)y)) {
+ // FIXME: what if there are two "routes" to get to the sequence?
+ ((HashMap)gotoSetNonTerminals).put(y, dead_state);
+ continue OUTER;
+ }
+ }
+ }
+ }
gotoSetNonTerminals.put(y, s);
}
}
- public String toString() { return "state["+idx+"]"; }
-
- public int compareTo(Table.State s) { return idx==s.idx ? 0 : idx < s.idx ? -1 : 1; }
- }
-
- /**
- * the information needed to perform a reduction; copied here to
- * avoid keeping references to <tt>Element</tt> objects in a Table
- */
- 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 boolean equals(Object o) {
- if (o==null) return false;
- if (o==this) return true;
- if (!(o instanceof Reduction)) return false;
- Reduction r = (Reduction)o;
- return r.position == position;
- }
- public Reduction(Position p) {
- this.position = p;
- this.numPop = p.pos;
- this.holder = new Forest[numPop];
- }
- public String toString() { return "[reduce " + position + "]"; }
- public Forest reduce(Forest f, GSS.Phase.Node parent, GSS.Phase.Node onlychild, GSS.Phase target, Forest rex) {
- holder[numPop-1] = f;
- return reduce(parent, numPop-2, rex, onlychild, target);
+ public String toStringx() {
+ StringBuffer st = new StringBuffer();
+ for(Position p : this) {
+ if (st.length() > 0) st.append("\n");
+ st.append(p);
+ }
+ return st.toString();
}
- public Forest reduce(GSS.Phase.Node parent, GSS.Phase.Node onlychild, GSS.Phase target, Forest rex) {
- return reduce(parent, numPop-1, rex, onlychild, target);
+ public String toString() {
+ StringBuffer ret = new StringBuffer();
+ ret.append("state["+idx+"]: ");
+ for(Position p : this) ret.append("{"+p+"} ");
+ return ret.toString();
}
- // FIXME: this could be more elegant and/or cleaner and/or somewhere else
- private Forest reduce(GSS.Phase.Node parent, int pos, Forest rex, GSS.Phase.Node onlychild, GSS.Phase target) {
- if (pos>=0) holder[pos] = parent.pending();
- if (pos<=0 && rex==null) {
- System.arraycopy(holder, 0, position.holder, 0, holder.length);
- rex = position.rewrite(target.getLocation());
- }
- if (pos >=0) {
- if (onlychild != null)
- reduce(onlychild, pos-1, rex, null, target);
- else
- for(GSS.Phase.Node child : parent.parents())
- reduce(child, pos-1, rex, null, target);
- } else {
- State state = parent.state.gotoSetNonTerminals.get(position.owner());
- if (state!=null)
- target.newNode(parent, rex, state, numPop<=0, parent.phase);
- }
- return rex;
- }
+ public int compareTo(State<Tok> s) { return idx==s.idx ? 0 : idx < s.idx ? -1 : 1; }
+ public int toInt() { return idx; }
}
}
- private static final Forest[] emptyForestArray = new Forest[0];
+ // Helpers //////////////////////////////////////////////////////////////////////////////
+
+ private static void reachable(Sequence s, HashSet<Position> h) {
+ reachable(s.firstp(), h);
+ for(Sequence ss : s.needs()) reachable(ss, h);
+ for(Sequence ss : s.hates()) reachable(ss, h);
+ }
+ private static void reachable(Element e, HashSet<Position> h) {
+ if (e instanceof Atom) return;
+ for(Sequence s : ((Union)e))
+ reachable(s, h);
+ }
+ private static void reachable(Position p, HashSet<Position> h) {
+ if (h.contains(p)) return;
+ h.add(p);
+ if (p.element() != null) reachable(p.element(), h);
+ }
+
}