X-Git-Url: http://git.megacz.com/?p=sbp.git;a=blobdiff_plain;f=src%2Fedu%2Fberkeley%2Fsbp%2FParser.java;h=cbf9b479f5bfb7e7fd763c5e08a68dc03ac450a4;hp=3fad1ddacd8c1e7a257d17193b9fcff3973b24b4;hb=c366dacc334fe2e35835164f5a37d3eebb2ca6d5;hpb=72d5880c192b7fbd525c97ded4cf11e935c58cc8;ds=sidebyside diff --git a/src/edu/berkeley/sbp/Parser.java b/src/edu/berkeley/sbp/Parser.java index 3fad1dd..cbf9b47 100644 --- a/src/edu/berkeley/sbp/Parser.java +++ b/src/edu/berkeley/sbp/Parser.java @@ -11,7 +11,7 @@ import java.lang.reflect.*; /** a parser which translates streams of Tokens of type T into a Forest */ public abstract class Parser { - private final Table pt; + public final Table pt; /** create a parser to parse the grammar with start symbol u */ protected Parser(Union u) { this.pt = new Table(u, top()); } @@ -22,22 +22,30 @@ public abstract class Parser { /** parse input for a exactly one unique result, throwing Ambiguous if not unique or Failed if none */ - public Tree parse1(Token.Stream input) throws IOException, Failed, Ambiguous { return parse(input).expand1(); } + public Tree parse1(Token.Stream input) throws IOException, Failed, Ambiguous { + Forest ret = parse(input); + try { return ret.expand1(); } + catch (Ambiguous a) { + System.out.println("while expanding:"); + System.out.println(ret); + throw a; + } + } /** parse input, using the table pt to drive the parser */ public Forest parse(Token.Stream input) throws IOException, Failed { GSS gss = new GSS(); Token.Location loc = input.getLocation(); - GSS.Phase current = gss.new Phase(null, input.next(), loc); - current.newNode(null, null, pt.start, true, null); + 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); + int count = 1; for(;;) { loc = input.getLocation(); - GSS.Phase next = gss.new Phase(current, input.next(), loc); current.reduce(); Forest forest = current.token==null ? null : shiftedToken((T)current.token, loc); - current.shift(next, forest); + GSS.Phase next = gss.new Phase(current, this, current, input.next(count, gss.resets, gss.waits), loc, forest); + count = next.hash.size(); if (current.isDone()) return (Forest)current.finalResult; - current.checkFailure(); current = next; } } @@ -45,13 +53,13 @@ public abstract class Parser { // Exceptions ////////////////////////////////////////////////////////////////////////////// - public static class Failed extends Exception { + 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 String toString() { return message/* + (location==null ? "" : (" at " + location))*/; } } public static class Ambiguous extends RuntimeException { @@ -73,6 +81,8 @@ public abstract class Parser { static class Table extends Walk.Cache { public final Walk.Cache cache = this; + + public HashMapBag byPosition = new HashMapBag(); private void walk(Element e, HashSet hs) { if (e==null) return; @@ -120,13 +130,23 @@ public abstract class Parser { 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)) { 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()); + 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); } // if the element following this position is an atom, copy the corresponding @@ -134,31 +154,77 @@ public abstract class Parser { 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(); + } } /** a single state in the LR table and the transitions possible from it */ - public class State implements Comparable, Iterable { + + public class State implements Comparable, IntegerMappable, Iterable { - public final int idx = master_state_idx++; + 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 hs; private transient HashMap gotoSetNonTerminals = new HashMap(); private transient TopologicalBag gotoSetTerminals = new TopologicalBag(); private TopologicalBag reductions = new TopologicalBag(); - private FastSet eofReductions = new FastSet(); + private HashSet eofReductions = new HashSet(); private TopologicalBag shifts = new TopologicalBag(); private boolean accept = false; + private VisitableMap oshifts = null; + private VisitableMap oreductions = null; + // Interface Methods ////////////////////////////////////////////////////////////////////////////// - public boolean canShift(Token t) { return shifts.contains(t); } - public Iterable getShifts(Token t) { return shifts.get(t); } public boolean isAccepting() { return accept; } - public Iterable getReductions(Token t) { return reductions.get(t); } - public Iterable getEofReductions() { return eofReductions; } + + public boolean canShift(Token t) { return oshifts.contains(t); } + public boolean canReduce(Token t) { return t==null ? eofReductions.size()>0 : oreductions.contains(t); } + public Iterator iterator() { return hs.iterator(); } + public void invokeShifts(Token t, Invokable irbc, B b, C c) { + oshifts.invoke(t, irbc, b, c); + } + public void invokeReductions(Token t, Invokable irbc, B b, C c) { + if (t==null) for(Reduction r : eofReductions) irbc.invoke(r, b, c); + else oreductions.invoke(t, irbc, b, c); + } + // Constructor ////////////////////////////////////////////////////////////////////////////// /** @@ -191,6 +257,7 @@ public abstract class Parser { // 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 @@ -253,7 +320,13 @@ public abstract class Parser { } } - public String toString() { return "state["+idx+"]"; } + public String toString() { + //return "state["+idx+"]"; + StringBuffer ret = new StringBuffer(); + ret.append("state["+idx+"]: "); + for(Position p : this) ret.append("{"+p+"} "); + return ret.toString(); + } public int compareTo(Table.State s) { return idx==s.idx ? 0 : idx < s.idx ? -1 : 1; } } @@ -265,7 +338,7 @@ public abstract class Parser { 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 Position position; private final Forest[] holder; // to avoid constant reallocation public int hashCode() { return position.hashCode(); } public boolean equals(Object o) { @@ -281,33 +354,52 @@ public abstract class Parser { 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); + + private Forest zero = null; + public Forest zero() { + if (zero != null) return zero; + if (numPop > 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()); } - 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 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; + Forest old = holder[pos]; + holder[pos] = parent.pending(); + if (pos==0) { + System.arraycopy(holder, 0, position.holder, 0, holder.length); + finish(onlychild, position.rewrite(parent.phase().getLocation()), parent.phase()); + } else { + reduce(onlychild, pos-1, parent.phase()); + } + holder[pos] = old; } // 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) { + private void reduce(GSS.Phase.Node parent, int pos, GSS.Phase target) { + Forest old = holder[pos]; + holder[pos] = parent.pending(); + if (pos==0) { 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); + for(int i=0; i