checkpoint harmony

[sbp.git] / src / edu / berkeley / sbp / Parser.java

diff --git a/src/edu/berkeley/sbp/Parser.java b/src/edu/berkeley/sbp/Parser.java
index a54fa08..cbf9b47 100644 (file)
--- a/src/edu/berkeley/sbp/Parser.java
+++ b/src/edu/berkeley/sbp/Parser.java
@@ -11,44 +11,41 @@ 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> {
 
 /** 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 final Table pt;

 
 

-    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);
-    }
-
-    /**
-     *  create a parser to parse the grammar with start symbol <tt>u</tt>
-     */
+    /** 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)  { this.pt = new Table(u, top()); }
     protected Parser(Table pt) { this.pt = pt; }
 

-    public abstract Forest<R> shiftedToken(T t);
+    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 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 { return parse(input).expand1(); }
+    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;
+        }
+    }

 
     /** 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 {
         GSS gss = new GSS();
 
     /** 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 {
         GSS gss = new GSS();

-        GSS.Phase current = gss.new Phase(null, input.next());
-        current.newNode(null, null, pt.start, true, null);
+        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);
+        int count = 1;

         for(;;) {
         for(;;) {

-            GSS.Phase next = gss.new Phase(current, input.next());
+            loc = input.getLocation();

             current.reduce();
             current.reduce();

-            Forest forest = current.token==null ? null : shiftedToken((T)current.token);
-            current.shift(next, forest);
+            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);
+            count = next.hash.size();

             if (current.isDone()) return (Forest<R>)current.finalResult;
             if (current.isDone()) return (Forest<R>)current.finalResult;

-            current.checkFailure();

             current = next;
         }
     }
             current = next;
         }
     }

@@ -56,13 +53,13 @@ public abstract class Parser<T extends Token, R> {
 
     // Exceptions //////////////////////////////////////////////////////////////////////////////
 
 
     // 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; }
         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 {
     }
 
     public static class Ambiguous extends RuntimeException {

@@ -80,23 +77,12 @@ public abstract class Parser<T extends Token, R> {
 
     // Table //////////////////////////////////////////////////////////////////////////////
 
 
     // Table //////////////////////////////////////////////////////////////////////////////
 

-    static class Top extends Union { public Top() { super("0"); } }
-

     /** an SLR(1) parse table which may contain conflicts */
     /** an SLR(1) parse table which may contain conflicts */

-    static class Table {
+    static class Table extends Walk.Cache {

 
 

-        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; }
+        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;
         
         private void walk(Element e, HashSet<Element> hs) {
             if (e==null) return;

@@ -109,21 +95,6 @@ public abstract class Parser<T extends Token, R> {
                     walk(p.element(), hs);
             }
         }
                     walk(p.element(), hs);
             }
         }

-        public HashSet<Element> walk() {
-            HashSet<Element> ret = new HashSet<Element>();
-            walk(start0, ret);
-            return ret;
-        }
-
-        /*
-        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;
 
         /** the start state */
         public final State   start;

@@ -134,32 +105,48 @@ public abstract class Parser<T extends Token, R> {
         /** 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); }
         /** 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, null, null));
+
+            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>();
 
             // 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());
             for(Element e : all_elements)
                 cache.ys.put(e, new Walk.YieldSet(e, cache).walk());

-            this.start = new State(closure(), all_states, all_elements);
+            HashSet<Position> hp = new HashSet<Position>();
+            reachable(start0, hp);
+            this.start = new State(hp, all_states, all_elements);

 
             // for each state, fill in the corresponding "row" of the parse table
             for(State state : all_states.values())
                 for(Position p : state.hs) {
 
                     // the Grammar's designated "last position" is the only accepting state
 
             // for each state, fill in the corresponding "row" of the parse table
             for(State 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;
 
                         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);
                     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
                     }
 
                     // if the element following this position is an atom, copy the corresponding

@@ -167,31 +154,77 @@ public abstract class Parser<T extends Token, R> {
                     if (p.element() != null && p.element() instanceof Atom)
                         state.shifts.addAll(state.gotoSetTerminals.subset(((Atom)p.element())));
                 }
                     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 */
         }
 
         /** a single state in the LR table and the transitions possible from it */

-        public class State implements Comparable<Table.State>, Iterable<Position> {
+
+        public class State implements Comparable<Table.State>, IntegerMappable, Iterable<Position> {

         
         

-            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<Position> hs;
 
             private transient HashMap<Element,State>          gotoSetNonTerminals = new HashMap<Element,State>();
             private transient TopologicalBag<Token,State>     gotoSetTerminals    = new TopologicalBag<Token,State>();
 
             private           TopologicalBag<Token,Reduction> reductions          = new TopologicalBag<Token,Reduction>();
             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>();
 
             private           TopologicalBag<Token,Reduction> reductions          = new TopologicalBag<Token,Reduction>();

-            private           FastSet<Reduction>              eofReductions       = new FastSet<Reduction>();
+            private           HashSet<Reduction>              eofReductions       = new HashSet<Reduction>();

             private           TopologicalBag<Token,State>     shifts              = new TopologicalBag<Token,State>();
             private           boolean                         accept              = false;
 
             private           TopologicalBag<Token,State>     shifts              = new TopologicalBag<Token,State>();
             private           boolean                         accept              = false;
 

+            private VisitableMap<Token,State> oshifts = null;
+            private VisitableMap<Token,Reduction> oreductions = null;
+

             // Interface Methods //////////////////////////////////////////////////////////////////////////////
 
             // 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 boolean             isAccepting()               { return accept; }

-            public Iterable<Reduction> getReductions(Token t)      { return reductions.get(t); }
-            public Iterable<Reduction> 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<Position>  iterator()                  { return hs.iterator(); }
 
             public Iterator<Position>  iterator()                  { return hs.iterator(); }
 

+            public <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) {
+                if (t==null) for(Reduction r : eofReductions) irbc.invoke(r, b, c);
+                else         oreductions.invoke(t, irbc, b, c);
+            }
+

             // Constructor //////////////////////////////////////////////////////////////////////////////
 
             /**
             // Constructor //////////////////////////////////////////////////////////////////////////////
 
             /**

@@ -224,6 +257,7 @@ public abstract class Parser<T extends Token, R> {
                 // 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);
                 // 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
 
                 // 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

@@ -236,7 +270,7 @@ public abstract class Parser<T extends Token, R> {
                     Atom a = (Atom)position.element();
                     HashSet<Position> hp = new HashSet<Position>();
                     reachable(position.next(), hp);
                     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
                 }
 
                 // Step 1b: for each _minimal, contiguous_ set of characters having an identical next-position

@@ -286,7 +320,13 @@ public abstract class Parser<T extends Token, R> {
                 }
             }
 
                 }
             }
 

-            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; }
         }
 
             public int compareTo(Table.State s) { return idx==s.idx ? 0 : idx < s.idx ? -1 : 1; }
         }

@@ -298,7 +338,7 @@ public abstract class Parser<T extends Token, R> {
         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;
         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) {
             private final Forest[] holder;    // to avoid constant reallocation
             public int hashCode() { return position.hashCode(); }
             public boolean equals(Object o) {

@@ -314,36 +354,70 @@ public abstract class Parser<T extends Token, R> {
                 this.holder = new Forest[numPop];
             }
             public String toString() { return "[reduce " + position + "]"; }
                 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 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) {
+                if (numPop==0) finish(parent, zero(), parent.phase());
+                else           reduce(parent, numPop-1, parent.phase());

             }
 
             }
 

-            // 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) {
+            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);
                     System.arraycopy(holder, 0, position.holder, 0, holder.length);

-                    rex = position.rewrite(target.getLocation());
+                    finish(onlychild, position.rewrite(parent.phase().getLocation()), parent.phase());
+                } else {
+                    reduce(onlychild, pos-1, parent.phase());

                 }
                 }

-                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);
+                holder[pos] = old;
+            }
+
+            // FIXME: this could be more elegant and/or cleaner and/or somewhere else
+            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);
+                    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()) finish(child, rex, target);

                 } else {
                 } else {

-                    State state = parent.state.gotoSetNonTerminals.get(position.owner());
-                    if (state!=null)
-                        target.newNode(parent, rex, state, numPop<=0, parent.phase);
+                    for(GSS.Phase.Node child : parent.parents()) reduce(child, pos-1, target);

                 }
                 }

-                return rex;
+                holder[pos] = old;
+            }
+            private void finish(GSS.Phase.Node parent, Forest result, GSS.Phase target) {
+                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);

             }
         }
     }
 
     private static final Forest[] emptyForestArray = new Forest[0];
             }
         }
     }
 
     private static final Forest[] emptyForestArray = new Forest[0];

+
+
+    // Helpers //////////////////////////////////////////////////////////////////////////////
+
+    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);
+    }
+

 }
 }