+// Copyright 2006 all rights reserved; see LICENSE file for BSD-style license
+
package edu.berkeley.sbp;
import edu.berkeley.sbp.*;
import edu.berkeley.sbp.util.*;
import java.io.*;
import java.util.*;
-/** a parser which translates streams of Tokens of type T into a Forest<R> */
-public abstract class Parser<Tok, Result> {
-
- protected final Table<Tok> pt;
+/** a parser which translates an Input<Token> into a Forest<NodeType> */
+public abstract class Parser<Token, NodeType> {
+ protected final Table<Token> 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 Parser(Union u, Topology<Token> top) { this.pt = new Table<Token>(u, top); }
+ Parser(Table<Token> pt) { this.pt = pt; }
/** 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);
+ public abstract Forest<NodeType> shiftToken(Token t, Input.Location newloc);
+
+ boolean helpgc = true;
- /** parse <tt>input</tt>, using the table <tt>pt</tt> to drive the parser */
- public Forest<Result> parse(Input<Tok> input) throws IOException, ParseFailed {
- GSS gss = new GSS();
+ public String toString() { return pt.toString(); }
+
+ /** parse <tt>input</tt>, and return the shared packed parse forest (or throw an exception) */
+ public Forest<NodeType> parse(Input<Token> input) throws IOException, ParseFailed {
+ GSS gss = new GSS(input);
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, null), pt.start, true);
+ Token tok = input.next();
+ GSS.Phase current = gss.new Phase<Token>(null, null, tok, loc, input.getLocation(), null);
+ current.newNode(new Result(Forest.create(loc.createRegion(loc), null, null, false), null, null), pt.start, true);
int count = 1;
- for(;;) {
- loc = input.getLocation();
+ for(int idx=0;;idx++) {
+ Input.Location oldloc = loc;
current.reduce();
- 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);
+ Forest forest = current.token==null ? null : shiftToken((Token)current.token, loc);
+ loc = input.getLocation();
+ Token nextToken = input.next();
+ GSS.Phase next = gss.new Phase<Token>(current, current, nextToken, loc, input.getLocation(), forest);
+
+ /*
+ FileOutputStream fos = new FileOutputStream("out-"+idx+".dot");
+ PrintWriter p = new PrintWriter(new OutputStreamWriter(fos));
+ GraphViz gv = new GraphViz();
+ for(Object n : current)
+ ((Node)n).toGraphViz(gv);
+ gv.dump(p);
+ p.flush();
+ p.close();
+ */
+
count = next.size();
- if (current.isDone()) return (Forest<Result>)gss.finalResult;
+ if (current.isDone()) return (Forest<NodeType>)gss.finalResult;
current = next;
}
}
// Table //////////////////////////////////////////////////////////////////////////////
/** an SLR(1) parse table which may contain conflicts */
- public static class Table<Tok> extends Walk.Cache {
+ static class Table<Token> extends Walk.Cache {
+
+ public String toString() {
+ StringBuffer sb = new StringBuffer();
+ sb.append("parse table");
+ for(State<Token> state : all_states.values()) {
+ sb.append(" " + state + "\n");
+ for(Topology<Token> 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<Token> t : state.reductions)
+ sb.append(" reduce \""+
+ new edu.berkeley.sbp.chr.CharTopology((IntegerTopology<Character>)t)+"\" => " +
+ state.reductions.getAll(t) + "\n");
+ for(Sequence s : state.gotoSetNonTerminals.keySet())
+ sb.append(" goto "+state.gotoSetNonTerminals.get(s)+" from " + s + "\n");
+ }
+ return sb.toString();
+ }
public final Walk.Cache cache = this;
- private void walk(Element e, HashSet<Element> hs) {
+ private void walk(Element e, HashSet<SequenceOrElement> 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);
+ }
+ private void walk(Sequence s, HashSet<SequenceOrElement> 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<Tok> start;
+ public final State<Token> start;
+
+ /** the state from which no reductions can be done */
+ private final State<Token> 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>>();
+ HashMap<HashSet<Position>,State<Token>> all_states = new HashMap<HashSet<Position>,State<Token>>();
+ HashSet<SequenceOrElement> all_elements = new HashSet<SequenceOrElement>();
/** construct a parse table for the given grammar */
public Table(Topology top) { this("s", top); }
cache.eof.put(start0, true);
// construct the set of states
- HashSet<Element> all_elements = new HashSet<Element>();
walk(start0, all_elements);
- for(Element e : all_elements)
+ for(SequenceOrElement e : all_elements)
cache.ys.addAll(e, new Walk.YieldSet(e, cache).walk());
+ for(SequenceOrElement e : all_elements)
+ cache.ys2.addAll(e, new Walk.YieldSet2(e, cache).walk());
HashSet<Position> hp = new HashSet<Position>();
reachable(start0, hp);
- this.start = new State<Tok>(hp, all_states, all_elements);
+
+ this.dead_state = new State<Token>(new HashSet<Position>());
+ this.start = new State<Token>(hp);
// for each state, fill in the corresponding "row" of the parse table
- for(State<Tok> state : all_states.values())
+ for(State<Token> 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);
Topology follow = wf.walk(p.owner());
- for(Position p2 = p; p2 != null && p2.element() != null; p2 = p2.next())
+ for(Position p2 = p; p2 != null && p2.element() != null; p2 = p2.next()) {
+ Atom set = new Walk.EpsilonFollowSet(new edu.berkeley.sbp.chr.CharAtom(top.empty().complement()),
+ new edu.berkeley.sbp.chr.CharAtom(top.empty()),
+ cache).walk(p2.element());
follow = follow.intersect(new Walk.Follow(top.empty(), p2.element(), all_elements, cache).walk(p2.element()));
+ if (set != null) follow = follow.intersect(set.getTokenTopology());
+ }
state.reductions.put(follow, p);
if (wf.includesEof()) state.eofReductions.add(p);
}
// if the element following this position is an atom, copy the corresponding
// set of rows out of the "master" goto table and into this state's shift table
if (p.element() != null && p.element() instanceof Atom)
- state.shifts.addAll(state.gotoSetTerminals.subset(((Atom)p.element())));
+ state.shifts.addAll(state.gotoSetTerminals.subset(((Atom)p.element()).getTokenTopology()));
}
if (top instanceof IntegerTopology)
- for(State<Tok> state : all_states.values()) {
+ for(State<Token> state : all_states.values()) {
state.oreductions = state.reductions.optimize(((IntegerTopology)top).functor());
state.oshifts = state.shifts.optimize(((IntegerTopology)top).functor());
}
/** a single state in the LR table and the transitions possible from it */
- public class State<Tok> implements Comparable<State<Tok>>, IntegerMappable, Iterable<Position> {
+ class State<Token> implements IntegerMappable, Iterable<Position> {
public final int idx = master_state_idx++;
private final HashSet<Position> hs;
+ public HashSet<State<Token>> also = new HashSet<State<Token>>();
- public transient HashMap<Element,State<Tok>> gotoSetNonTerminals = new HashMap<Element,State<Tok>>();
- private transient TopologicalBag<Tok,State<Tok>> gotoSetTerminals = new TopologicalBag<Tok,State<Tok>>();
+ public transient HashMap<Sequence,State<Token>> gotoSetNonTerminals = new HashMap<Sequence,State<Token>>();
+ private transient TopologicalBag<Token,State<Token>> gotoSetTerminals = new TopologicalBag<Token,State<Token>>();
- private TopologicalBag<Tok,Position> reductions = new TopologicalBag<Tok,Position>();
+ private TopologicalBag<Token,Position> reductions = new TopologicalBag<Token,Position>();
private HashSet<Position> eofReductions = new HashSet<Position>();
- private TopologicalBag<Tok,State<Tok>> shifts = new TopologicalBag<Tok,State<Tok>>();
+ private TopologicalBag<Token,State<Token>> shifts = new TopologicalBag<Token,State<Token>>();
private boolean accept = false;
- private VisitableMap<Tok,State<Tok>> oshifts = null;
- private VisitableMap<Tok,Position> oreductions = null;
+ private VisitableMap<Token,State<Token>> oshifts = null;
+ private VisitableMap<Token,Position> oreductions = null;
// Interface Methods //////////////////////////////////////////////////////////////////////////////
- boolean isAccepting() { return accept; }
- public Iterator<Position> iterator() { return hs.iterator(); }
+ boolean isAccepting() { return accept; }
+ public Iterator<Position> iterator() { return hs.iterator(); }
- boolean canShift(Tok t) { return oshifts.contains(t); }
- <B,C> void invokeShifts(Tok t, Invokable<State<Tok>,B,C> irbc, B b, C c) {
+ boolean canShift(Token t) { return oshifts!=null && oshifts.contains(t); }
+ <B,C> void invokeShifts(Token t, Invokable<State<Token>,B,C> irbc, B b, C c) {
oshifts.invoke(t, irbc, 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) {
+ boolean canReduce(Token t) { return oreductions != null && (t==null ? eofReductions.size()>0 : oreductions.contains(t)); }
+ <B,C> void invokeReductions(Token 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);
}
/**
* create a new state consisting of all the <tt>Position</tt>s in <tt>hs</tt>
* @param hs the set of <tt>Position</tt>s comprising this <tt>State</tt>
- * @param all_states the set of states already constructed (to avoid recreating states)
* @param all_elements the set of all elements (Atom instances need not be included)
*
* In principle these two steps could be merged, but they
* for non-Atom Elements.
* </ul>
*/
- public State(HashSet<Position> hs,
- HashMap<HashSet<Position>,State<Tok>> all_states,
- HashSet<Element> all_elements) {
+ public State(HashSet<Position> hs) { this(hs, false); }
+ public boolean special;
+ public State(HashSet<Position> hs, boolean special) {
this.hs = hs;
+ this.special = special;
// 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);
+ ((HashMap)all_states).put(hs, this);
+
+ for(Position p : hs) {
+ if (!p.isFirst()) continue;
+ for(Sequence s : p.owner().needs()) {
+ if (hs.contains(s.firstp())) continue;
+ HashSet<Position> h2 = new HashSet<Position>();
+ reachable(s.firstp(), h2);
+ also.add((State<Token>)(all_states.get(h2) == null ? (State)new State<Token>(h2,true) : (State)all_states.get(h2)));
+ }
+ for(Sequence s : p.owner().hates()) {
+ if (hs.contains(s.firstp())) continue;
+ HashSet<Position> h2 = new HashSet<Position>();
+ reachable(s, h2);
+ also.add((State<Token>)(all_states.get(h2) == null ? (State)new State<Token>(h2,true) : (State)all_states.get(h2)));
+ }
+ }
// 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<Tok,Position> bag0 = new TopologicalBag<Tok,Position>();
+ TopologicalBag<Token,Position> bag0 = new TopologicalBag<Token,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, hp);
+ bag0.addAll(a.getTokenTopology(), 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<Tok> r : bag0) {
+ for(Topology<Token> 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<Tok>(h, all_states, all_elements) : all_states.get(h));
+ ((TopologicalBag)gotoSetTerminals).put(r, all_states.get(h) == null
+ ? new State<Token>(h) : 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").
- HashMapBag<Element,Position> move = new HashMapBag<Element,Position>();
+
+ HashMapBag<SequenceOrElement,Position> move = new HashMapBag<SequenceOrElement,Position>();
for(Position p : hs) {
Element e = p.element();
if (e==null) continue;
- for(Element y : cache.ys.getAll(e)) {
+ for(SequenceOrElement y : cache.ys2.getAll(e)) {
+ //System.out.println(e + " yields " + y);
HashSet<Position> hp = new HashSet<Position>();
reachable(p.next(), hp);
move.addAll(y, hp);
}
}
- for(Element y : move) {
+ OUTER: for(SequenceOrElement y : move) {
HashSet<Position> h = move.getAll(y);
- State<Tok> s = all_states.get(h) == null ? new State<Tok>(h, all_states, all_elements) : all_states.get(h);
- gotoSetNonTerminals.put(y, s);
+ State<Token> s = all_states.get(h) == null ? (State)new State<Token>(h) : (State)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((Sequence)y, dead_state);
+ continue OUTER;
+ }
+ }
+ }
+ gotoSetNonTerminals.put((Sequence)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(State<Tok> s) { return idx==s.idx ? 0 : idx < s.idx ? -1 : 1; }
+ public Walk.Cache cache() { return cache; }
public int toInt() { return idx; }
}
}
// 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