import java.util.*;
/** a parser which translates streams of Tokens of type T into a Forest<R> */
-public abstract class Parser<T extends Token, R> {
+public abstract class Parser<Tok, Result> {
- private final Table pt;
+ 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; }
+ protected Parser(Union u, Topology<Tok> top) { this.pt = new Table<Tok>(u, top); }
+ protected Parser(Table<Tok> 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);
+ protected abstract Forest<Result> shiftToken(Tok t, Input.Location newloc);
- /** this method must return an empty topology of the input token type */
- public abstract Topology<T> top();
+ boolean helpgc = true;
+
+ 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, ParseFailed {
+ public Forest<Result> parse(Input<Tok> 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);
- current.newNode(null, Forest.leaf(null, null), pt.start, true);
+ 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(;;) {
+ for(int idx=0;;idx++) {
+ Input.Location oldloc = loc;
loc = input.getLocation();
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);
+ 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<R>)gss.finalResult;
+ if (current.isDone()) return (Forest<Result>)gss.finalResult;
current = next;
}
}
// Table //////////////////////////////////////////////////////////////////////////////
/** an SLR(1) parse table which may contain conflicts */
- static class Table extends Walk.Cache {
+ static class Table<Tok> extends Walk.Cache {
+
+ public String toString() {
+ StringBuffer sb = new StringBuffer();
+ 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;
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);
+ }
+ 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);
}
/** the start state */
- public final State start;
+ public final State<Tok> start;
/** 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 ux, Topology top) {
Union start0 = new Union("0");
- start0.add(new Sequence.Singleton(ux, null, null));
+ 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 = new HashSet<Element>();
+ HashSet<Element> all_elements = new HashSet<Element>();
walk(start0, all_elements);
for(Element e : all_elements)
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);
+ 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 (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());
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, red);
- if (wf.includesEof()) state.eofReductions.add(red);
+ 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())));
}
- for(State state : all_states.values()) {
- state.oreductions = state.reductions.optimize();
- state.oshifts = state.shifts.optimize();
- }
+ 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 (!p.element().possiblyEpsilon(this)) return false;
+ 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>, IntegerMappable, Iterable<Position> {
+ class State<Tok> implements Comparable<State<Tok>>, IntegerMappable, Iterable<Position> {
public final int idx = master_state_idx++;
private final HashSet<Position> hs;
- public 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 HashSet<Reduction> eofReductions = new HashSet<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<Token,State> oshifts = null;
- private VisitableMap<Token,Reduction> oreductions = null;
+ private VisitableMap<Tok,State<Tok>> oshifts = null;
+ private VisitableMap<Tok,Position> oreductions = null;
// Interface Methods //////////////////////////////////////////////////////////////////////////////
boolean isAccepting() { return accept; }
public Iterator<Position> iterator() { return hs.iterator(); }
- boolean canShift(Token t) { return oshifts.contains(t); }
- <B,C> void invokeShifts(Token t, Invokable<State,B,C> irbc, B b, C c) {
+ boolean canShift(Tok t) { return 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(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);
+ boolean canReduce(Tok t) { return 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);
}
* </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();
// 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),
}
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);
gotoSetNonTerminals.put(y, s);
}
}
+ 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 String toString() {
StringBuffer ret = new StringBuffer();
ret.append("state["+idx+"]: ");
return ret.toString();
}
- public int compareTo(Table.State s) { return idx==s.idx ? 0 : idx < s.idx ? -1 : 1; }
+ public int compareTo(State<Tok> s) { return idx==s.idx ? 0 : idx < s.idx ? -1 : 1; }
public int toInt() { return idx; }
}
-
- /**
- * 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
- /*private*/ final Position position;
- 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;
- }
- public String toString() { return "[reduce " + position + "]"; }
-
- private Forest zero = null;
- public Forest zero() {
- if (zero != null) return zero;
- if (position.pos > 0) throw new Error();
- return zero = position.rewrite(null);
- }
-
- // FIXME: this could be more elegant and/or cleaner and/or somewhere else
- public void reduce(GSS.Phase.Node parent, int pos, GSS.Phase target, Forest[] holder) {
- Forest old = holder[pos];
- holder[pos] = parent.pending();
- if (pos==0) {
- System.arraycopy(holder, 0, position.holder, 0, holder.length);
- for(int i=0; i<position.pos; i++) if (position.holder[i]==null) throw new Error("realbad");
- Forest rex = position.rewrite(target.getLocation());
- for(GSS.Phase.Node child : parent.parents()) child.finish(this, rex, target, holder);
- } else {
- for(GSS.Phase.Node child : parent.parents()) reduce(child, pos-1, target, holder);
- }
- holder[pos] = old;
- }
- }
}
- 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.firstp(), h);
+ reachable(s, h);
}
private static void reachable(Position p, HashSet<Position> h) {
if (h.contains(p)) return;
projects / sbp.git / blobdiff