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 abstract class Parser<Tok, Result> {
- public 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; }
-
- public abstract Forest<R> shiftedToken(T t, Token.Location loc);
- public abstract Topology<T> top();
+ protected Parser(Union u, Topology<Tok> top) { this.pt = new Table<Tok>(u, top); }
+ protected Parser(Table<Tok> pt) { this.pt = pt; }
-
- /** 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;
- }
- }
+ /** implement this method to create the output forest corresponding to a lone shifted input token */
+ public abstract Forest<Result> shiftToken(Tok t, Input.Location loc);
/** 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();
- 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, null), pt.start, true);
+ Input.Location loc = input.getLocation();
+ GSS.Phase current = gss.new Phase<Tok>(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();
- //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);
- count = next.hash.size();
- if (current.isDone()) return (Forest<R>)current.finalResult;
+ Forest forest = current.token==null ? null : shiftToken((Tok)current.token, loc);
+ GSS.Phase next = gss.new Phase<Tok>(current, this, current, input.next(count, gss.resets, gss.waits), loc, forest);
+ count = next.size();
+ if (current.isDone()) return (Forest<Result>)gss.finalResult;
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 //////////////////////////////////////////////////////////////////////////////
/** an SLR(1) parse table which may contain conflicts */
- static class Table extends Walk.Cache {
+ public static class Table<Tok> extends Walk.Cache {
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;
}
/** 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.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);
+ 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 (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);
+ 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 (!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> {
-
- /*
- 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 class State<Tok> implements Comparable<State<Tok>>, IntegerMappable, Iterable<Position> {
+
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 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 //////////////////////////////////////////////////////////////////////////////
- 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(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);
}
- 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);
+
+ 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;
// 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
// 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),
// "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) {
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 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 + "]"; }
-
- private Forest zero = null;
- public Forest zero() {
- if (zero != null) return zero;
- if (numPop > 0) throw new Error();
- return zero = position.rewrite(null);
+ public String toString() {
+ StringBuffer ret = new StringBuffer();
+ ret.append("state["+idx+"]: ");
+ for(Position p : this) ret.append("{"+p+"} ");
+ return ret.toString();
}
- public void reduce(GSS.Phase.Node parent) {
- if (numPop==0) finish(parent, zero(), parent.phase());
- else reduce(parent, numPop-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;
- 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 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 {
- for(GSS.Phase.Node child : parent.parents()) reduce(child, pos-1, target);
- }
- 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);
- }
+ 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(Element e, HashSet<Position> h) {
if (e instanceof Atom) return;
for(Sequence s : ((Union)e))
projects / sbp.git / blobdiff