1 // Copyright 2006-2007 all rights reserved; see LICENSE file for BSD-style license
3 package edu.berkeley.sbp.meta;
4 import edu.berkeley.sbp.util.*;
5 import edu.berkeley.sbp.*;
6 import edu.berkeley.sbp.chr.*;
7 import edu.berkeley.sbp.misc.*;
9 import java.lang.annotation.*;
10 import java.lang.reflect.*;
14 * The inner classes of this class represent nodes in the Abstract
15 * Syntax Tree of a grammar.
17 public class GrammarAST {
19 public static interface ImportResolver {
20 public InputStream getImportStream(String importname);
24 * Returns a Union representing the metagrammar (<tt>meta.g</tt>); the Tree produced by
25 * parsing with this Union should be provided to <tt>buildFromAST</tt>
27 public static Union getMetaGrammar() {
28 return buildFromAST(MetaGrammar.meta, "s", null);
32 * Create a grammar from a parse tree and binding resolver
34 * @param t a tree produced by parsing a grammar using the metagrammar
35 * @param s the name of the "start symbol"
36 * @param gbr a GrammarBindingResolver that resolves grammatical reductions into tree-node-heads
38 public static Union buildFromAST(Tree grammarAST, String startingNonterminal, ImportResolver resolver) {
39 return new GrammarAST(resolver, "").buildGrammar(grammarAST, startingNonterminal);
42 private static Object illegalTag = ""; // this is the tag that should never appear in the non-dropped output FIXME
44 // Instance //////////////////////////////////////////////////////////////////////////////
46 private final String prefix;
47 private final ImportResolver resolver;
49 public GrammarAST(ImportResolver resolver, String prefix) {
51 this.resolver = resolver;
54 // Methods //////////////////////////////////////////////////////////////////////////////
56 private Union buildGrammar(Tree t, String rootNonTerminal) {
58 if (o instanceof Union) return (Union)o;
59 return ((GrammarAST.GrammarNode)o).build(rootNonTerminal);
62 private Object[] walkChildren(Tree t) {
63 Object[] ret = new Object[t.size()];
64 for(int i=0; i<ret.length; i++)
65 ret[i] = walk(t.child(i));
66 return Reflection.lub(ret);
68 private static String stringifyChildren(Tree t) {
69 StringBuffer sb = new StringBuffer();
70 for(int i=0; i<t.size(); i++) {
71 sb.append(t.child(i).head());
72 sb.append(stringifyChildren(t.child(i)));
76 private static String unescape(Tree t) {
77 StringBuffer sb = new StringBuffer();
78 for(int i=0; i<t.size(); i++)
79 sb.append(t.child(i).head()+stringifyChildren(t.child(i)));
83 private ElementNode walkElement(Tree t) { return (ElementNode)walk(t); }
84 private String walkString(Tree t) { return (String)walk(t); }
85 private Seq walkSeq(Tree t) { return (Seq)walk(t); }
86 private Object walk(Tree t) {
87 String head = (String)t.head();
88 while(head.indexOf('.') > 0)
89 head = head.substring(head.indexOf('.')+1);
90 if (head==null) throw new RuntimeException("head is null: " + t);
91 if (head.equals("|")) return walkChildren(t);
92 if (head.equals("RHS")) return walkChildren(t);
93 if (head.equals("Grammar")) return new GrammarNode(walkChildren(t));
94 if (head.equals("(")) return new UnionNode((Seq[][])walkChildren(t.child(0)));
95 if (head.equals("Word")) return stringifyChildren(t);
96 if (head.equals("Elements")) return new Seq((ElementNode[])Reflection.rebuild(walkChildren(t), ElementNode[].class));
97 if (head.equals("NonTerminalReference")) return new ReferenceNode(stringifyChildren(t.child(0)));
98 if (head.equals(")")) return new ReferenceNode(stringifyChildren(t.child(0)), true);
99 if (head.equals("::")) return walkSeq(t.child(1)).tag(walkString(t.child(0)));
100 if (head.equals("...")) return new DropNode(new RepeatNode(new TildeNode(new AtomNode()), null, true, true, false));
102 if (head.equals("++")) return new RepeatNode(walkElement(t.child(0)), null, false, true, true);
103 if (head.equals("+")) return new RepeatNode(walkElement(t.child(0)), null, false, true, false);
104 if (head.equals("++/")) return new RepeatNode(walkElement(t.child(0)), walkElement(t.child(1)), false, true, true);
105 if (head.equals("+/")) return new RepeatNode(walkElement(t.child(0)), walkElement(t.child(1)), false, true, false);
106 if (head.equals("**")) return new RepeatNode(walkElement(t.child(0)), null, true, true, true);
107 if (head.equals("*")) return new RepeatNode(walkElement(t.child(0)), null, true, true, false);
108 if (head.equals("**/")) return new RepeatNode(walkElement(t.child(0)), walkElement(t.child(1)), true, true, true);
109 if (head.equals("*/")) return new RepeatNode(walkElement(t.child(0)), walkElement(t.child(1)), true, true, false);
110 if (head.equals("?")) return new RepeatNode(walkElement(t.child(0)), null, true, false, false);
112 if (head.equals("!")) return new DropNode(walkElement(t.child(0)));
113 if (head.equals("^")) return new LiteralNode(walkString(t.child(0)), true);
114 if (head.equals("`")) return new BacktickNode(walkElement(t.child(0)));
115 if (head.equals("Quoted")) return stringifyChildren(t);
116 if (head.equals("Literal")) return new LiteralNode(walkString(t.child(0)));
117 if (head.equals("->")) return walkSeq(t.child(0)).follow(walkElement(t.child(1)));
118 if (head.equals("DropNT")) return new NonTerminalNode(walkString(t.child(0)), (Seq[][])walkChildren(t.child(1)), false, null, true);
119 if (head.equals("=")) return new NonTerminalNode(walkString(t.child(0)), (Seq[][])walk(t.child(2)),
120 true, t.size()==2 ? null : walkString(t.child(1)), false);
121 if (head.equals("&")) return walkSeq(t.child(0)).and(walkSeq(t.child(1)));
122 if (head.equals("&~")) return walkSeq(t.child(0)).andnot(walkSeq(t.child(1)));
123 if (head.equals("/")) return (walkSeq(t.child(0))).separate(walkElement(t.child(1)));
124 if (head.equals("()")) return new LiteralNode("");
125 if (head.equals("[")) return new AtomNode((char[][])Reflection.rebuild(walkChildren(t), char[][].class));
126 if (head.equals("\\{")) return new DropNode(new AtomNode(new char[] { CharAtom.left, CharAtom.left }));
127 if (head.equals("\\}")) return new DropNode(new AtomNode(new char[] { CharAtom.right, CharAtom.right }));
128 if (head.equals(">>")) return new DropNode(new AtomNode(new char[] { CharAtom.left, CharAtom.left }));
129 if (head.equals("<<")) return new DropNode(new AtomNode(new char[] { CharAtom.right, CharAtom.right }));
130 if (head.equals("~")) return new TildeNode(walkElement(t.child(0)));
131 if (head.equals("~~")) return new Seq(new RepeatNode(new TildeNode(new AtomNode()), null, true, true, false)).andnot(walkSeq(t.child(0)));
132 if (head.equals("Range")) {
133 if (t.size()==2 && ">".equals(t.child(0).head())) return new char[] { CharAtom.left, CharAtom.left };
134 if (t.size()==2 && "<".equals(t.child(0).head())) return new char[] { CharAtom.right, CharAtom.right };
135 if (t.size()==1) return new char[] { unescape(t).charAt(0), unescape(t).charAt(0) };
136 return new char[] { unescape(t).charAt(0), unescape(t).charAt(1) };
138 if (head.equals("\"\"")) return "";
139 if (head.equals("\n")) return "\n";
140 if (head.equals("\r")) return "\r";
141 if (head.equals("SubGrammar")) return GrammarAST.buildFromAST(t.child(0), "s", resolver);
142 if (head.equals("NonTerminal"))
143 return new NonTerminalNode(walkString(t.child(0)),
144 (Seq[][])walkChildren(t.child(1)), false, null, false);
145 if (head.equals("Colons")) {
146 String tag = walkString(t.child(0));
147 Seq[][] seqs = (Seq[][])walk(t.child(1));
148 for(Seq[] seq : seqs)
149 for(int i=0; i<seq.length; i++)
150 seq[i] = seq[i].tag(tag);
151 return new NonTerminalNode(tag, seqs, false, null, false);
153 if (head.equals("#import")) {
154 if (resolver != null) {
155 String fileName = (String)stringifyChildren(t.child(0));
157 String newPrefix = t.size()<2 ? "" : (walkString(t.child(1))+".");
158 InputStream fis = resolver.getImportStream(fileName);
160 throw new RuntimeException("unable to find #include file \""+fileName+"\"");
161 Tree tr = new CharParser(getMetaGrammar()).parse(fis).expand1();
162 return (GrammarNode)new GrammarAST(resolver, newPrefix).walk(tr);
163 } catch (Exception e) {
164 throw new RuntimeException("while parsing " + fileName, e);
167 throw new RuntimeException("no resolver given");
170 throw new RuntimeException("unknown head: \"" + head + "\" => " + (head.equals("...")));
174 // Nodes //////////////////////////////////////////////////////////////////////////////
176 /** Root node of a grammar's AST; a set of named nonterminals */
177 private class GrammarNode extends HashMap<String,NonTerminalNode> {
178 public GrammarNode(NonTerminalNode[] nonterminals) {
179 for(NonTerminalNode nt : nonterminals) {
180 if (nt==null) continue;
181 if (this.get(nt.name)!=null)
182 throw new RuntimeException("duplicate definition of nonterminal \""+nt.name+"\"");
183 this.put(nt.name, nt);
186 public GrammarNode(Object[] nt) { add(nt); }
187 private void add(Object o) {
189 else if (o instanceof Object[]) for(Object o2 : (Object[])o) add(o2);
190 else if (o instanceof NonTerminalNode) {
191 NonTerminalNode nt = (NonTerminalNode)o;
192 if (this.get(nt.name)!=null)
193 throw new RuntimeException("duplicate definition of nonterminal \""+nt.name+"\"");
194 this.put(nt.name, nt);
196 else if (o instanceof GrammarNode)
197 for(NonTerminalNode n : ((GrammarNode)o).values())
200 public String toString() {
202 for(NonTerminalNode nt : values()) ret += nt + ", ";
205 public Union build(String rootNonterminal) {
206 Context cx = new Context(this);
208 for(GrammarAST.NonTerminalNode nt : values())
209 if (nt.name.equals(rootNonterminal))
210 return (Union)cx.get(nt.name);
215 /** a node in the AST which is resolved into an Element */
216 private abstract class ElementNode {
217 public boolean isLifted() { return false; }
218 public boolean isDropped(Context cx) { return false; }
219 public Atom toAtom(Context cx) { throw new Error("can't convert a " + this.getClass().getName() + " to an atom: " + this); }
220 public abstract Element build(Context cx, NonTerminalNode cnt, boolean dropall);
223 private class UnionNode extends ElementNode {
224 public Seq[][] sequences;
225 public String sep = null;
227 public UnionNode(Seq seq) { this(new Seq[][] { new Seq[] { seq } }); }
228 public UnionNode(Seq[][] sequences) { this(sequences, false, null); }
229 public UnionNode(Seq[][] sequences, boolean rep, String sep) {
230 this.sequences = sequences;
234 public boolean isDropped(Context cx) {
235 for(Seq[] seqs : sequences)
237 if (!seq.isDropped(cx))
241 public Atom toAtom(Context cx) {
243 for(Seq[] ss : sequences)
245 ret = ret==null ? s.toAtom(cx) : (Atom)ret.union(s.toAtom(cx));
248 public Element build(Context cx, NonTerminalNode cnt, boolean dropall) {
249 return buildIntoPreallocatedUnion(cx, cnt, dropall, new Union(null, false)); }
250 public Element buildIntoPreallocatedUnion(Context cx, NonTerminalNode cnt, boolean dropall, Union u) {
253 urep = new Union(null, false);
254 urep.add(Sequence.create(cnt.name, new Element[0]));
256 ? Sequence.create(new Element[] { u }, 0)
257 : Sequence.create(new Element[] { cx.get(sep), u }, 1));
259 HashSet<Sequence> bad2 = new HashSet<Sequence>();
260 for(int i=0; i<sequences.length; i++) {
261 Seq[] group = sequences[i];
262 Union u2 = new Union(null, false);
263 if (sequences.length==1) u2 = u;
264 for(int j=0; j<group.length; j++)
266 group[j].build(cx, u2, cnt, dropall);
268 Union u3 = new Union(null, false);
269 group[j].build(cx, u3, cnt, dropall);
270 Sequence s = Sequence.create(cnt.name,
271 new Element[] { u3, urep },
272 new boolean[] { false, false },
273 new boolean[] { false, true});
276 if (sequences.length==1) break;
277 Sequence seq = Sequence.create(u2);
278 for(Sequence s : bad2) seq = seq.andnot(s);
280 bad2.add(Sequence.create(u2));
286 private class NonTerminalNode extends UnionNode {
287 public boolean alwaysDrop;
288 public String name = null;
289 public boolean isDropped(Context cx) { return alwaysDrop; }
290 public NonTerminalNode(String name, Seq[][] sequences, boolean rep, String sep, boolean alwaysDrop) {
291 super(sequences, rep, sep==null?null:(prefix + sep));
292 this.name = prefix + name;
293 this.alwaysDrop = alwaysDrop;
295 public Element build(Context cx, NonTerminalNode cnt, boolean dropall) { return cx.get(name); }
299 public boolean alwaysDrop = false;
300 public boolean isDropped(Context cx) {
301 if (alwaysDrop) return true;
302 if (tag!=null) return false;
303 for(int i=0; i<elements.length; i++)
304 if (!elements[i].isDropped(cx))
308 HashSet<Seq> and = new HashSet<Seq>();
309 HashSet<Seq> not = new HashSet<Seq>();
310 ElementNode[] elements;
313 public Seq(ElementNode e) { this(new ElementNode[] { e }); }
314 public Seq(ElementNode[] elements) { this(elements, true); }
315 public Seq(ElementNode[] el, boolean check) {
316 this.elements = new ElementNode[el.length];
317 System.arraycopy(el, 0, elements, 0, el.length);
318 for(int i=0; i<elements.length; i++) {
319 if (elements[i]==null)
320 throw new RuntimeException();
322 // FIXME: this whole mechanism is sketchy
324 for(int i=0; i<elements.length; i++) {
325 if ((elements[i] instanceof ReferenceNode) && ((ReferenceNode)elements[i]).parenthesized) {
326 ReferenceNode rn = (ReferenceNode)elements[i];
327 ElementNode replace = null;
328 for(int j=0; j<elements.length; j++) {
329 if (!(elements[j] instanceof ReferenceNode)) continue;
330 ReferenceNode rn2 = (ReferenceNode)elements[j];
331 if (rn2.nonTerminal.equals(rn.nonTerminal) && !rn2.parenthesized) {
332 if (replace == null) {
333 replace = new UnionNode(new Seq(rn2).andnot(new Seq(elements, false)));
335 elements[j] = replace;
341 public Atom toAtom(Context cx) {
342 if (elements.length != 1)
343 throw new Error("you attempted to use ->, **, ++, or a similar character-class"+
344 " operator on a [potentially] multicharacter production");
345 return elements[0].toAtom(cx);
347 public Seq tag(String tag) { this.tag = tag; return this; }
348 public Seq follow(ElementNode follow) { this.follow = follow; return this; }
349 public Seq and(Seq s) { and.add(s); s.alwaysDrop = true; return this; }
350 public Seq andnot(Seq s) { not.add(s); s.alwaysDrop = true; return this; }
351 public Seq separate(ElementNode sep) {
352 ElementNode[] elements = new ElementNode[this.elements.length * 2 - 1];
353 for(int i=0; i<this.elements.length; i++) {
354 elements[i*2] = this.elements[i];
355 if (i<this.elements.length-1)
356 elements[i*2+1] = new DropNode(sep);
358 this.elements = elements;
361 public Sequence build(Context cx, Union u, NonTerminalNode cnt, boolean dropall) {
362 Sequence ret = build0(cx, cnt, dropall);
363 for(Seq s : and) ret = ret.and(s.build(cx, null, cnt, true));
364 for(Seq s : not) ret = ret.andnot(s.build(cx, null, cnt, true));
365 if (u!=null) u.add(ret);
368 public Sequence build0(Context cx, NonTerminalNode cnt, boolean dropall) {
369 boolean[] drops = new boolean[elements.length];
370 Element[] els = new Element[elements.length];
371 dropall |= isDropped(cx);
372 for(int i=0; i<elements.length; i++) {
373 if (dropall) drops[i] = true;
374 else drops[i] = elements[i].isDropped(cx);
375 if (elements[i] instanceof LiteralNode && ((LiteralNode)elements[i]).caret) {
376 if (tag != null) throw new RuntimeException("cannot have multiple tags in a sequence: " + this);
377 tag = ((LiteralNode)elements[i]).getLiteralTag();
382 boolean multiNonDrop = false;
383 for(int i=0; i<drops.length; i++)
385 if (idx==-1) idx = i;
386 else multiNonDrop = true;
387 for(int i=0; i<elements.length; i++) {
388 if (!multiNonDrop && i==idx && tag!=null && elements[i] instanceof RepeatNode) {
389 els[i] = ((RepeatNode)elements[i]).build(cx, cnt, dropall, tag);
392 els[i] = elements[i].build(cx, cnt, dropall);
394 if (tag==null && multiNonDrop)
395 throw new RuntimeException("multiple non-dropped elements in sequence: " + Sequence.create("", els));
396 boolean[] lifts = new boolean[elements.length];
397 for(int i=0; i<elements.length; i++)
398 lifts[i] = elements[i].isLifted();
402 ? Sequence.create(illegalTag, els)
403 : Sequence.create(tag, els, drops, lifts);
404 else if (tag==null) ret = Sequence.create(els, idx);
405 else ret = Sequence.create(tag, els, drops, lifts);
408 ret = Sequence.create(tag, els, drops, lifts);
409 if (this.follow != null)
410 ret = ret.followedBy(this.follow.toAtom(cx));
415 private class ReferenceNode extends ElementNode {
416 public String nonTerminal;
417 public boolean parenthesized;
418 public ReferenceNode() { }
419 public ReferenceNode(String nonTerminal) { this(nonTerminal, false); }
420 public ReferenceNode(String nonTerminal, boolean parenthesized) {
421 this.nonTerminal = nonTerminal.indexOf('.')==-1 ? (prefix + nonTerminal) : nonTerminal;
422 this.parenthesized = parenthesized;
424 public NonTerminalNode resolve(Context cx) {
425 NonTerminalNode ret = cx.grammar.get(nonTerminal);
426 if (ret==null) throw new RuntimeException("undefined nonterminal: " + nonTerminal);
429 public Atom toAtom(Context cx) {
430 ElementNode ret = cx.grammar.get(nonTerminal);
431 if (ret == null) throw new RuntimeException("unknown nonterminal \""+nonTerminal+"\"");
432 return ret.toAtom(cx);
434 public boolean isDropped(Context cx) { return resolve(cx).isDropped(cx); }
435 public Element build(Context cx, NonTerminalNode cnt, boolean dropall) {
436 Element ret = cx.get(nonTerminal);
437 if (ret == null) throw new RuntimeException("unknown nonterminal \""+nonTerminal+"\"");
442 private class LiteralNode extends ElementNode {
443 private String string;
444 private final String thePrefix = prefix;
445 private boolean caret;
446 public LiteralNode(String string) { this(string, false); }
447 public LiteralNode(String string, boolean caret) {
448 this.string = string;
451 public String getLiteralTag() { return caret ? thePrefix+string : null; }
452 public String toString() { return "\""+string+"\""; }
453 public boolean isDropped(Context cx) { return true; }
454 public Atom toAtom(Context cx) {
455 if (string.length()!=1) return super.toAtom(cx);
456 Range.Set set = new Range.Set();
457 set.add(string.charAt(0), string.charAt(0));
458 return CharAtom.set(set);
460 public Element build(Context cx, NonTerminalNode cnt, boolean dropall) { return CharAtom.string(string); }
463 private class AtomNode extends ElementNode {
465 public AtomNode() { this(new char[0][]); }
466 public AtomNode(char[][] ranges) { this.ranges = ranges; }
467 public AtomNode(char[] range) { this.ranges = new char[][] { range }; }
468 public Element build(Context cx, NonTerminalNode cnt, boolean dropall) { return toAtom(cx); }
469 public Atom toAtom(Context cx) {
470 Range.Set set = new Range.Set();
471 for(char[] r : ranges) set.add(r[0], r[1]);
472 return CharAtom.set(set);
476 private class RepeatNode extends ElementNode {
477 public ElementNode e, sep;
478 public final boolean zero, many, max;
479 public RepeatNode(ElementNode e, ElementNode sep, boolean zero, boolean many, boolean max) {
480 this.e = e; this.sep = sep; this.zero = zero; this.many = many; this.max = max;
482 public Atom toAtom(Context cx) { return sep==null ? e.toAtom(cx) : super.toAtom(cx); }
483 public boolean isDropped(Context cx) { return e.isDropped(cx); }
484 public Element build(Context cx, NonTerminalNode cnt, boolean dropall) {
485 Element ret = build(cx, cnt, dropall, illegalTag);
486 String must = "must be tagged unless they appear within a dropped expression or their contents are dropped: ";
487 if (!dropall && !isDropped(cx) && !e.isDropped(cx))
488 if (!many) throw new RuntimeException("options (?) " + must + ret);
489 else if (zero) throw new RuntimeException("zero-or-more repetitions (*) " + must + ret);
490 else throw new RuntimeException("one-or-more repetitions (+) " + must + ret);
493 public Element build(Context cx, NonTerminalNode cnt, boolean dropall, Object repeatTag) {
495 ? Repeat.repeat(e.build(cx, null, dropall), zero, many, sep==null ? null : sep.build(cx, null, dropall), repeatTag)
497 ? Repeat.repeatMaximal(e.toAtom(cx), zero, many, repeatTag)
498 : Repeat.repeatMaximal(e.build(cx, null, dropall), zero, many, sep.toAtom(cx), repeatTag);
502 private abstract class ElementNodeWrapper extends ElementNode {
503 protected ElementNode _e;
504 public ElementNodeWrapper(ElementNode e) { this._e = e; }
505 public boolean isDropped(Context cx) { return _e.isDropped(cx); }
506 public Atom toAtom(Context cx) { return _e.toAtom(cx); }
507 public Element build(Context cx, NonTerminalNode cnt, boolean dropall) { return _e.build(cx, cnt, dropall); }
510 private class BacktickNode extends ElementNodeWrapper {
511 public BacktickNode(ElementNode e) { super(e); }
512 public boolean isLifted() { return true; }
515 private class TildeNode extends ElementNodeWrapper {
516 public TildeNode(ElementNode e) { super(e); }
517 public Atom toAtom(Context cx) { return (Atom)((Topology<Character>)_e.toAtom(cx).complement()); }
518 public Element build(Context cx, NonTerminalNode cnt, boolean dropall) { return toAtom(cx); }
521 private class DropNode extends ElementNodeWrapper {
522 public DropNode(ElementNode e) { super(e); }
523 public boolean isDropped(Context cx) { return true; }
526 //////////////////////////////////////////////////////////////////////////////
528 public class Context {
529 public HashMap<String,Union> map = new HashMap<String,Union>();
530 public GrammarNode grammar;
531 public Context(Tree t) { }
532 public Context(GrammarNode g) { this.grammar = g; }
533 public Union build() {
535 for(NonTerminalNode nt : grammar.values()) {
536 Union u = get(nt.name);
537 if ("s".equals(nt.name))
542 public Union peek(String name) { return map.get(name); }
543 public void put(String name, Union u) { map.put(name, u); }
544 public Union get(String name) {
545 Union ret = map.get(name);
546 if (ret != null) return ret;
547 NonTerminalNode nt = grammar.get(name);
549 throw new Error("warning could not find " + name);
551 ret = new Union(name, false);
553 nt.buildIntoPreallocatedUnion(this, nt, nt.isDropped(this), ret);