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 {
20 * Returns a Union representing the metagrammar (<tt>meta.g</tt>); the Tree produced by
21 * parsing with this Union should be provided to <tt>buildFromAST</tt>
23 public static Union getMetaGrammar() {
24 return buildFromAST(MetaGrammar.meta, "s", new File[0]);
28 * Create a grammar from a parse tree and binding resolver
30 * @param t a tree produced by parsing a grammar using the metagrammar
31 * @param s the name of the "start symbol"
32 * @param gbr a GrammarBindingResolver that resolves grammatical reductions into tree-node-heads
34 public static Union buildFromAST(Tree grammarAST, String startingNonterminal, File[] includes) {
35 return new GrammarAST(includes, "").buildGrammar(grammarAST, startingNonterminal);
38 private static Object illegalTag = ""; // this is the tag that should never appear in the non-dropped output FIXME
40 // Instance //////////////////////////////////////////////////////////////////////////////
42 private final String prefix;
43 private final File[] includes;
45 public GrammarAST(File[] includes, String prefix) {
47 this.includes = includes;
50 // Methods //////////////////////////////////////////////////////////////////////////////
52 private Union buildGrammar(Tree t, String rootNonTerminal) {
54 if (o instanceof Union) return (Union)o;
55 return ((GrammarAST.GrammarNode)o).build(rootNonTerminal);
58 private Object[] walkChildren(Tree t) {
59 Object[] ret = new Object[t.size()];
60 for(int i=0; i<ret.length; i++)
61 ret[i] = walk(t.child(i));
62 return Reflection.lub(ret);
64 private static String stringifyChildren(Tree t) {
65 StringBuffer sb = new StringBuffer();
66 for(int i=0; i<t.size(); i++) {
67 sb.append(t.child(i).head());
68 sb.append(stringifyChildren(t.child(i)));
72 private static String unescape(Tree t) {
73 StringBuffer sb = new StringBuffer();
74 for(int i=0; i<t.size(); i++)
75 sb.append(t.child(i).head()+stringifyChildren(t.child(i)));
79 private ElementNode walkElement(Tree t) { return (ElementNode)walk(t); }
80 private String walkString(Tree t) { return (String)walk(t); }
81 private Seq walkSeq(Tree t) { return (Seq)walk(t); }
82 private Object walk(Tree t) {
83 String head = (String)t.head();
84 while(head.indexOf('.') > 0)
85 head = head.substring(head.indexOf('.')+1);
86 if (head==null) throw new RuntimeException("head is null: " + t);
87 if (head.equals("|")) return walkChildren(t);
88 if (head.equals("RHS")) return walkChildren(t);
89 if (head.equals("Grammar")) return new GrammarNode(walkChildren(t));
90 if (head.equals("(")) return new UnionNode((Seq[][])walkChildren(t.child(0)));
91 if (head.equals("Word")) return stringifyChildren(t);
92 if (head.equals("Elements")) return new Seq((ElementNode[])Reflection.rebuild(walkChildren(t), ElementNode[].class));
93 if (head.equals("NonTerminalReference")) return new ReferenceNode(stringifyChildren(t.child(0)));
94 if (head.equals(")")) return new ReferenceNode(stringifyChildren(t.child(0)), true);
95 if (head.equals("::")) return walkSeq(t.child(1)).tag(walkString(t.child(0)));
96 if (head.equals("...")) return new DropNode(new RepeatNode(new TildeNode(new AtomNode()), null, true, true, false));
98 if (head.equals("++")) return new RepeatNode(walkElement(t.child(0)), null, false, true, true);
99 if (head.equals("+")) return new RepeatNode(walkElement(t.child(0)), null, false, true, false);
100 if (head.equals("++/")) return new RepeatNode(walkElement(t.child(0)), walkElement(t.child(1)), false, true, true);
101 if (head.equals("+/")) return new RepeatNode(walkElement(t.child(0)), walkElement(t.child(1)), false, true, false);
102 if (head.equals("**")) return new RepeatNode(walkElement(t.child(0)), null, true, true, true);
103 if (head.equals("*")) return new RepeatNode(walkElement(t.child(0)), null, true, true, false);
104 if (head.equals("**/")) return new RepeatNode(walkElement(t.child(0)), walkElement(t.child(1)), true, true, true);
105 if (head.equals("*/")) return new RepeatNode(walkElement(t.child(0)), walkElement(t.child(1)), true, true, false);
106 if (head.equals("?")) return new RepeatNode(walkElement(t.child(0)), null, true, false, false);
108 if (head.equals("!")) return new DropNode(walkElement(t.child(0)));
109 if (head.equals("^")) return new LiteralNode(walkString(t.child(0)), true);
110 if (head.equals("`")) return new BacktickNode(walkElement(t.child(0)));
111 if (head.equals("Quoted")) return stringifyChildren(t);
112 if (head.equals("Literal")) return new LiteralNode(walkString(t.child(0)));
113 if (head.equals("->")) return walkSeq(t.child(0)).follow(walkElement(t.child(1)));
114 if (head.equals("DropNT")) return new NonTerminalNode(walkString(t.child(0)), (Seq[][])walkChildren(t.child(1)), false, null, true);
115 if (head.equals("=")) return new NonTerminalNode(walkString(t.child(0)), (Seq[][])walk(t.child(2)),
116 true, t.size()==2 ? null : walkString(t.child(1)), false);
117 if (head.equals("&")) return walkSeq(t.child(0)).and(walkSeq(t.child(1)));
118 if (head.equals("&~")) return walkSeq(t.child(0)).andnot(walkSeq(t.child(1)));
119 if (head.equals("/")) return (walkSeq(t.child(0))).separate(walkElement(t.child(1)));
120 if (head.equals("()")) return new LiteralNode("");
121 if (head.equals("[")) return new AtomNode((char[][])Reflection.rebuild(walkChildren(t), char[][].class));
122 if (head.equals("\\{")) return new DropNode(new AtomNode(new char[] { CharAtom.left, CharAtom.left }));
123 if (head.equals("\\}")) return new DropNode(new AtomNode(new char[] { CharAtom.right, CharAtom.right }));
124 if (head.equals(">>")) return new DropNode(new AtomNode(new char[] { CharAtom.left, CharAtom.left }));
125 if (head.equals("<<")) return new DropNode(new AtomNode(new char[] { CharAtom.right, CharAtom.right }));
126 if (head.equals("~")) return new TildeNode(walkElement(t.child(0)));
127 if (head.equals("~~")) return new Seq(new RepeatNode(new TildeNode(new AtomNode()), null, true, true, false)).andnot(walkSeq(t.child(0)));
128 if (head.equals("Range")) {
129 if (t.size()==2 && ">".equals(t.child(0).head())) return new char[] { CharAtom.left, CharAtom.left };
130 if (t.size()==2 && "<".equals(t.child(0).head())) return new char[] { CharAtom.right, CharAtom.right };
131 if (t.size()==1) return new char[] { unescape(t).charAt(0), unescape(t).charAt(0) };
132 return new char[] { unescape(t).charAt(0), unescape(t).charAt(1) };
134 if (head.equals("\"\"")) return "";
135 if (head.equals("\n")) return "\n";
136 if (head.equals("\r")) return "\r";
137 if (head.equals("SubGrammar")) return GrammarAST.buildFromAST(t.child(0), "s", includes);
138 if (head.equals("NonTerminal"))
139 return new NonTerminalNode(walkString(t.child(0)),
140 (Seq[][])walkChildren(t.child(1)), false, null, false);
141 if (head.equals("Colons")) {
142 String tag = walkString(t.child(0));
143 Seq[][] seqs = (Seq[][])walk(t.child(1));
144 for(Seq[] seq : seqs)
145 for(int i=0; i<seq.length; i++)
146 seq[i] = seq[i].tag(tag);
147 return new NonTerminalNode(tag, seqs, false, null, false);
149 if (head.equals("#import")) {
150 String fileName = (String)stringifyChildren(t.child(0));
151 for(File f : includes) {
152 File file = new File(f.getAbsolutePath()+File.separatorChar+fileName);
153 if (!file.exists()) continue;
155 String newPrefix = t.size()<2 ? "" : (walkString(t.child(1))+".");
156 FileInputStream fis = new FileInputStream(file);
157 Tree tr = new CharParser(getMetaGrammar()).parse(fis).expand1();
158 return (GrammarNode)new GrammarAST(includes, newPrefix).walk(tr);
159 } catch (Exception e) {
160 throw new RuntimeException("while parsing " + file, e);
163 throw new RuntimeException("unable to find #include file \""+fileName+"\"");
165 throw new RuntimeException("unknown head: \"" + head + "\" => " + (head.equals("...")));
169 // Nodes //////////////////////////////////////////////////////////////////////////////
171 /** Root node of a grammar's AST; a set of named nonterminals */
172 private class GrammarNode extends HashMap<String,NonTerminalNode> {
173 public GrammarNode(NonTerminalNode[] nonterminals) {
174 for(NonTerminalNode nt : nonterminals) {
175 if (nt==null) continue;
176 if (this.get(nt.name)!=null)
177 throw new RuntimeException("duplicate definition of nonterminal \""+nt.name+"\"");
178 this.put(nt.name, nt);
181 public GrammarNode(Object[] nt) { add(nt); }
182 private void add(Object o) {
184 else if (o instanceof Object[]) for(Object o2 : (Object[])o) add(o2);
185 else if (o instanceof NonTerminalNode) {
186 NonTerminalNode nt = (NonTerminalNode)o;
187 if (this.get(nt.name)!=null)
188 throw new RuntimeException("duplicate definition of nonterminal \""+nt.name+"\"");
189 this.put(nt.name, nt);
191 else if (o instanceof GrammarNode)
192 for(NonTerminalNode n : ((GrammarNode)o).values())
195 public String toString() {
197 for(NonTerminalNode nt : values()) ret += nt + ", ";
200 public Union build(String rootNonterminal) {
201 Context cx = new Context(this);
203 for(GrammarAST.NonTerminalNode nt : values())
204 if (nt.name.equals(rootNonterminal))
205 return (Union)cx.get(nt.name);
210 /** a node in the AST which is resolved into an Element */
211 private abstract class ElementNode {
212 public boolean isLifted() { return false; }
213 public boolean drop(Context cx) { return false; }
214 public Atom toAtom(Context cx) { throw new Error("can't convert a " + this.getClass().getName() + " to an atom: " + this); }
215 public abstract Element build(Context cx, NonTerminalNode cnt, boolean dropall);
218 private class UnionNode extends ElementNode {
219 public Seq[][] sequences;
220 public String sep = null;
222 public UnionNode(Seq seq) { this(new Seq[][] { new Seq[] { seq } }); }
223 public UnionNode(Seq[][] sequences) { this(sequences, false, null); }
224 public UnionNode(Seq[][] sequences, boolean rep, String sep) {
225 this.sequences = sequences;
229 public boolean drop(Context cx) {
230 for(Seq[] seqs : sequences)
236 public Atom toAtom(Context cx) {
238 for(Seq[] ss : sequences)
240 ret = ret==null ? s.toAtom(cx) : (Atom)ret.union(s.toAtom(cx));
243 public Element build(Context cx, NonTerminalNode cnt, boolean dropall) {
244 return buildIntoPreallocatedUnion(cx, cnt, dropall, new Union(null, false)); }
245 public Element buildIntoPreallocatedUnion(Context cx, NonTerminalNode cnt, boolean dropall, Union u) {
248 urep = new Union(null, false);
249 urep.add(Sequence.create(cnt.name, new Element[0]));
251 ? Sequence.create(new Element[] { u }, 0)
252 : Sequence.create(new Element[] { cx.get(sep), u }, 1));
254 HashSet<Sequence> bad2 = new HashSet<Sequence>();
255 for(int i=0; i<sequences.length; i++) {
256 Seq[] group = sequences[i];
257 Union u2 = new Union(null, false);
258 if (sequences.length==1) u2 = u;
259 for(int j=0; j<group.length; j++)
261 group[j].build(cx, u2, cnt, dropall);
263 Union u3 = new Union(null, false);
264 group[j].build(cx, u3, cnt, dropall);
265 Sequence s = Sequence.create(cnt.name,
266 new Element[] { u3, urep },
267 new boolean[] { false, false },
268 new boolean[] { false, true});
271 if (sequences.length==1) break;
272 Sequence seq = Sequence.create(u2);
273 for(Sequence s : bad2) seq = seq.andnot(s);
275 bad2.add(Sequence.create(u2));
281 private class NonTerminalNode extends UnionNode {
282 public boolean alwaysDrop;
283 public String name = null;
284 public boolean drop(Context cx) { return alwaysDrop; }
285 public NonTerminalNode(String name, Seq[][] sequences, boolean rep, String sep, boolean alwaysDrop) {
286 super(sequences, rep, sep==null?null:(prefix + sep));
287 this.name = prefix + name;
288 this.alwaysDrop = alwaysDrop;
290 public Element build(Context cx, NonTerminalNode cnt, boolean dropall) { return cx.get(name); }
294 public boolean alwaysDrop = false;
295 public boolean drop(Context cx) {
296 if (alwaysDrop) return true;
297 if (tag!=null) return false;
298 for(int i=0; i<elements.length; i++)
299 if (!elements[i].drop(cx))
303 HashSet<Seq> and = new HashSet<Seq>();
304 HashSet<Seq> not = new HashSet<Seq>();
305 ElementNode[] elements;
308 public Seq(ElementNode e) { this(new ElementNode[] { e }); }
309 public Seq(ElementNode[] elements) { this(elements, true); }
310 public Seq(ElementNode[] el, boolean check) {
311 this.elements = new ElementNode[el.length];
312 System.arraycopy(el, 0, elements, 0, el.length);
313 for(int i=0; i<elements.length; i++) {
314 if (elements[i]==null)
315 throw new RuntimeException();
317 // FIXME: this whole mechanism is sketchy
319 for(int i=0; i<elements.length; i++) {
320 if ((elements[i] instanceof ReferenceNode) && ((ReferenceNode)elements[i]).parenthesized) {
321 ReferenceNode rn = (ReferenceNode)elements[i];
322 ElementNode replace = null;
323 for(int j=0; j<elements.length; j++) {
324 if (!(elements[j] instanceof ReferenceNode)) continue;
325 ReferenceNode rn2 = (ReferenceNode)elements[j];
326 if (rn2.nonTerminal.equals(rn.nonTerminal) && !rn2.parenthesized) {
327 if (replace == null) {
328 replace = new UnionNode(new Seq(rn2).andnot(new Seq(elements, false)));
330 elements[j] = replace;
336 public Atom toAtom(Context cx) {
337 if (elements.length != 1)
338 throw new Error("you attempted to use ->, **, ++, or a similar character-class"+
339 " operator on a [potentially] multicharacter production");
340 return elements[0].toAtom(cx);
342 public Seq tag(String tag) { this.tag = tag; return this; }
343 public Seq follow(ElementNode follow) { this.follow = follow; return this; }
344 public Seq and(Seq s) { and.add(s); s.alwaysDrop = true; return this; }
345 public Seq andnot(Seq s) { not.add(s); s.alwaysDrop = true; return this; }
346 public Seq separate(ElementNode sep) {
347 ElementNode[] elements = new ElementNode[this.elements.length * 2 - 1];
348 for(int i=0; i<this.elements.length; i++) {
349 elements[i*2] = this.elements[i];
350 if (i<this.elements.length-1)
351 elements[i*2+1] = new DropNode(sep);
353 this.elements = elements;
356 public Sequence build(Context cx, Union u, NonTerminalNode cnt, boolean dropall) {
357 Sequence ret = build0(cx, cnt, dropall);
358 for(Seq s : and) ret = ret.and(s.build(cx, null, cnt, true));
359 for(Seq s : not) ret = ret.andnot(s.build(cx, null, cnt, true));
360 if (u!=null) u.add(ret);
363 public Sequence build0(Context cx, NonTerminalNode cnt, boolean dropall) {
364 boolean[] drops = new boolean[elements.length];
365 Element[] els = new Element[elements.length];
367 for(int i=0; i<elements.length; i++) {
368 if (dropall) drops[i] = true;
369 else drops[i] = elements[i].drop(cx);
370 if (elements[i] instanceof LiteralNode && ((LiteralNode)elements[i]).caret) {
371 if (tag != null) throw new RuntimeException("cannot have multiple tags in a sequence: " + this);
372 tag = ((LiteralNode)elements[i]).getLiteralTag();
377 boolean multiNonDrop = false;
378 for(int i=0; i<drops.length; i++)
380 if (idx==-1) idx = i;
381 else multiNonDrop = true;
382 for(int i=0; i<elements.length; i++) {
383 if (!multiNonDrop && i==idx && tag!=null && elements[i] instanceof RepeatNode) {
384 els[i] = ((RepeatNode)elements[i]).build(cx, cnt, dropall, tag);
387 els[i] = elements[i].build(cx, cnt, dropall);
389 if (tag==null && multiNonDrop)
390 throw new RuntimeException("multiple non-dropped elements in sequence: " + Sequence.create("", els));
391 boolean[] lifts = new boolean[elements.length];
392 for(int i=0; i<elements.length; i++)
393 lifts[i] = elements[i].isLifted();
397 ? Sequence.create(illegalTag, els)
398 : Sequence.create(tag, els, drops, lifts);
399 else if (tag==null) ret = Sequence.create(els, idx);
400 else ret = Sequence.create(tag, els, drops, lifts);
403 ret = Sequence.create(tag, els, drops, lifts);
404 if (this.follow != null)
405 ret = ret.followedBy(this.follow.toAtom(cx));
410 private class ReferenceNode extends ElementNode {
411 public String nonTerminal;
412 public boolean parenthesized;
413 public ReferenceNode() { }
414 public ReferenceNode(String nonTerminal) { this(nonTerminal, false); }
415 public ReferenceNode(String nonTerminal, boolean parenthesized) {
416 this.nonTerminal = nonTerminal.indexOf('.')==-1 ? (prefix + nonTerminal) : nonTerminal;
417 this.parenthesized = parenthesized;
419 public NonTerminalNode resolve(Context cx) {
420 NonTerminalNode ret = cx.grammar.get(nonTerminal);
421 if (ret==null) throw new RuntimeException("undefined nonterminal: " + nonTerminal);
424 public Atom toAtom(Context cx) {
425 ElementNode ret = cx.grammar.get(nonTerminal);
426 if (ret == null) throw new RuntimeException("unknown nonterminal \""+nonTerminal+"\"");
427 return ret.toAtom(cx);
429 public boolean drop(Context cx) { return resolve(cx).drop(cx); }
430 public Element build(Context cx, NonTerminalNode cnt, boolean dropall) {
431 Element ret = cx.get(nonTerminal);
432 if (ret == null) throw new RuntimeException("unknown nonterminal \""+nonTerminal+"\"");
437 private class LiteralNode extends ElementNode {
438 private String string;
439 private final String thePrefix = prefix;
440 private boolean caret;
441 public LiteralNode(String string) { this(string, false); }
442 public LiteralNode(String string, boolean caret) {
443 this.string = string;
446 public String getLiteralTag() { return caret ? thePrefix+string : null; }
447 public String toString() { return "\""+string+"\""; }
448 public boolean drop(Context cx) { return true; }
449 public Atom toAtom(Context cx) {
450 if (string.length()!=1) return super.toAtom(cx);
451 Range.Set set = new Range.Set();
452 set.add(string.charAt(0), string.charAt(0));
453 return CharAtom.set(set);
455 public Element build(Context cx, NonTerminalNode cnt, boolean dropall) { return CharAtom.string(string); }
458 private class AtomNode extends ElementNode {
460 public AtomNode() { this(new char[0][]); }
461 public AtomNode(char[][] ranges) { this.ranges = ranges; }
462 public AtomNode(char[] range) { this.ranges = new char[][] { range }; }
463 public Element build(Context cx, NonTerminalNode cnt, boolean dropall) { return toAtom(cx); }
464 public Atom toAtom(Context cx) {
465 Range.Set set = new Range.Set();
466 for(char[] r : ranges) set.add(r[0], r[1]);
467 return CharAtom.set(set);
471 private class RepeatNode extends ElementNode {
472 public ElementNode e, sep;
473 public final boolean zero, many, max;
474 public RepeatNode(ElementNode e, ElementNode sep, boolean zero, boolean many, boolean max) {
475 this.e = e; this.sep = sep; this.zero = zero; this.many = many; this.max = max;
477 public Atom toAtom(Context cx) { return sep==null ? e.toAtom(cx) : super.toAtom(cx); }
478 public boolean drop(Context cx) { return e.drop(cx); }
479 public Element build(Context cx, NonTerminalNode cnt, boolean dropall) {
480 Element ret = build(cx, cnt, dropall, illegalTag);
481 String must = "must be tagged unless they appear within a dropped expression or their contents are dropped: ";
482 if (!dropall && !drop(cx) && !e.drop(cx))
483 if (!many) throw new RuntimeException("options (?) " + must + ret);
484 else if (zero) throw new RuntimeException("zero-or-more repetitions (*) " + must + ret);
485 else throw new RuntimeException("one-or-more repetitions (+) " + must + ret);
488 public Element build(Context cx, NonTerminalNode cnt, boolean dropall, Object repeatTag) {
490 ? Repeat.repeat(e.build(cx, null, dropall), zero, many, sep==null ? null : sep.build(cx, null, dropall), repeatTag)
492 ? Repeat.repeatMaximal(e.toAtom(cx), zero, many, repeatTag)
493 : Repeat.repeatMaximal(e.build(cx, null, dropall), zero, many, sep.toAtom(cx), repeatTag);
497 private abstract class ElementNodeWrapper extends ElementNode {
498 protected ElementNode _e;
499 public ElementNodeWrapper(ElementNode e) { this._e = e; }
500 public boolean drop(Context cx) { return _e.drop(cx); }
501 public Atom toAtom(Context cx) { return _e.toAtom(cx); }
502 public Element build(Context cx, NonTerminalNode cnt, boolean dropall) { return _e.build(cx, cnt, dropall); }
505 private class BacktickNode extends ElementNodeWrapper {
506 public BacktickNode(ElementNode e) { super(e); }
507 public boolean isLifted() { return true; }
510 private class TildeNode extends ElementNodeWrapper {
511 public TildeNode(ElementNode e) { super(e); }
512 public Atom toAtom(Context cx) { return (Atom)((Topology<Character>)_e.toAtom(cx).complement()); }
513 public Element build(Context cx, NonTerminalNode cnt, boolean dropall) { return toAtom(cx); }
516 private class DropNode extends ElementNodeWrapper {
517 public DropNode(ElementNode e) { super(e); }
518 public boolean drop(Context cx) { return true; }
521 //////////////////////////////////////////////////////////////////////////////
523 public class Context {
524 public HashMap<String,Union> map = new HashMap<String,Union>();
525 public GrammarNode grammar;
526 public Context(Tree t) { }
527 public Context(GrammarNode g) { this.grammar = g; }
528 public Union build() {
530 for(NonTerminalNode nt : grammar.values()) {
531 Union u = get(nt.name);
532 if ("s".equals(nt.name))
537 public Union peek(String name) { return map.get(name); }
538 public void put(String name, Union u) { map.put(name, u); }
539 public Union get(String name) {
540 Union ret = map.get(name);
541 if (ret != null) return ret;
542 NonTerminalNode nt = grammar.get(name);
544 throw new Error("warning could not find " + name);
546 ret = new Union(name, false);
548 nt.buildIntoPreallocatedUnion(this, nt, nt.drop(this), ret);