1 // Copyright 2006 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 * Create a grammar from a parse tree and binding resolver
22 * @param t a tree produced by parsing a grammar using the metagrammar
23 * @param s the name of the "start symbol"
24 * @param gbr a GrammarBindingResolver that resolves grammatical reductions into tree-node-heads
26 public static Union buildFromAST(Tree grammarAST, String startingNonterminal, File[] includes) {
27 return new GrammarAST(includes, "").buildGrammar(grammarAST, startingNonterminal);
30 public static Object illegalTag = ""; // this is the tag that should never appear in the non-dropped output FIXME
32 private final String prefix;
33 private final File[] includes;
35 public GrammarAST(File[] includes, String prefix) {
37 this.includes = includes;
40 // Methods //////////////////////////////////////////////////////////////////////////////
42 private Union buildGrammar(Tree t, String rootNonTerminal) {
43 return ((GrammarAST.GrammarNode)walk(t)).build(rootNonTerminal);
46 public Object[] walkChildren(Tree t) {
47 Object[] ret = new Object[t.size()];
48 for(int i=0; i<ret.length; i++) {
49 ret[i] = walk(t.child(i));
50 if (ret[i] instanceof Object[])
51 ret[i] = Reflection.lub((Object[])ret[i]);
53 return Reflection.lub(ret);
55 public String stringifyChildren(Tree t) {
56 StringBuffer sb = new StringBuffer();
57 for(int i=0; i<t.size(); i++) {
58 sb.append(t.child(i).head());
59 sb.append(stringifyChildren(t.child(i)));
63 public String unescape(Tree t) {
64 StringBuffer sb = new StringBuffer();
65 for(int i=0; i<t.size(); i++)
66 sb.append(t.child(i).head()+stringifyChildren(t.child(i)));
70 public Object walk(Tree t) {
71 String head = (String)t.head();
72 while(head.indexOf('.') > 0)
73 head = head.substring(head.indexOf('.')+1);
74 if (head==null) throw new RuntimeException("head is null: " + t);
75 if (head.equals("|")) return walkChildren(t);
76 if (head.equals("RHS")) return walkChildren(t);
77 if (head.equals("Grammar")) return new GrammarNode(walkChildren(t));
78 if (head.equals("(")) return new UnionNode((Seq[][])walkChildren(t.child(0)));
79 if (head.equals("Word")) return stringifyChildren(t);
80 if (head.equals("Elements")) return seq2((ElementNode[])Reflection.rebuild(walkChildren(t), ElementNode[].class));
81 if (head.equals("NonTerminalReference")) return new ReferenceNode(stringifyChildren(t.child(0)));
82 //if (head.equals(")")) return new ReferenceNode(stringifyChildren(t.child(0)));
83 if (head.equals("{")) return new XTree((Seq)walk(t.child(0)));
84 if (head.equals("::")) return tag((String)walk(t.child(0)), (Seq)walk(t.child(1)));
85 if (head.equals("++")) return plusmax((ElementNode)walk(t.child(0)));
86 if (head.equals("...")) return star(new TildeNode(new AtomNode()));
87 if (head.equals("+")) return plus((ElementNode)walk(t.child(0)));
88 if (head.equals("++/")) return plusmaxfollow((ElementNode)walk(t.child(0)), (ElementNode)walk(t.child(1)));
89 if (head.equals("+/")) return plusfollow((ElementNode)walk(t.child(0)), (ElementNode)walk(t.child(1)));
90 if (head.equals("**")) return starmax((ElementNode)walk(t.child(0)));
91 if (head.equals("*")) return star((ElementNode)walk(t.child(0)));
92 if (head.equals("**/")) return starmaxfollow((ElementNode)walk(t.child(0)), (ElementNode)walk(t.child(1)));
93 if (head.equals("*/")) return starfollow((ElementNode)walk(t.child(0)), (ElementNode)walk(t.child(1)));
94 if (head.equals("?")) return question((ElementNode)walk(t.child(0)));
95 if (head.equals("!")) return new DropNode((ElementNode)walk(t.child(0)));
96 if (head.equals("^")) return new LiteralNode((String)walk(t.child(0)), true);
97 if (head.equals("`")) {
98 ElementNode ret = (ElementNode)walk(t.child(0));
102 if (head.equals("Quoted")) return stringifyChildren(t);
103 if (head.equals("Literal")) return new LiteralNode((String)walk(t.child(0)));
104 if (head.equals("->")) return arrow((Seq)walk(t.child(0)), (ElementNode)walk(t.child(1)));
105 if (head.equals("DropNT")) return new NonTerminalNode((String)walk(t.child(0)), (Seq[][])walkChildren(t.child(1)), false, null, true);
106 if (head.equals("=") && t.size()==2) return new NonTerminalNode((String)walk(t.child(0)), (Seq[][])walk(t.child(1)), true, null, false);
107 if (head.equals("=")) return new NonTerminalNode((String)walk(t.child(0)), (Seq[][])walk(t.child(2)), true, (String)walk(t.child(1)), false);
108 if (head.equals("&")) return and2((Seq)walk(t.child(0)), (Seq)walk(t.child(1)));
109 if (head.equals("&~")) return andnot2((Seq)walk(t.child(0)), (Seq)walk(t.child(1)));
110 if (head.equals("/")) return ((Seq)walk(t.child(0))).separate((ElementNode)walk(t.child(1)));
111 if (head.equals("()")) return epsilon;
112 if (head.equals("[")) return new AtomNode((AtomNodeRange[])Reflection.rebuild(walkChildren(t), AtomNodeRange[].class));
113 if (head.equals("\\{")) return new DropNode(new AtomNode(new AtomNodeRange(CharAtom.left, CharAtom.left)));
114 if (head.equals("\\}")) return new DropNode(new AtomNode(new AtomNodeRange(CharAtom.right, CharAtom.right)));
115 if (head.equals("~")) return new TildeNode((ElementNode)walk(t.child(0)));
116 if (head.equals("~~")) {
117 Seq seq = new Seq(star(new TildeNode(new AtomNode())));
118 return seq.andnot((Seq)walk(t.child(0)));
120 if (head.equals("Range") && t.size()==1) return new AtomNodeRange(unescape(t).charAt(0));
121 if (head.equals("Range")) return new AtomNodeRange(unescape(t).charAt(0), unescape(t).charAt(1));
122 if (head.equals("\"\"")) return "";
123 if (head.equals("\n")) return "\n";
124 if (head.equals("\r")) return "\r";
125 if (head.equals("grammar.Grammar")) return walkChildren(t);
126 if (head.equals("SubGrammar")) return GrammarBuilder.buildFromAST(t.child(0), "s", includes);
127 if (head.equals("NonTerminal"))
128 return new NonTerminalNode((String)walk(t.child(0)),
129 (Seq[][])walkChildren(t.child(1)), false, null, false);
130 if (head.equals("Colons")) {
131 String tag = (String)walk(t.child(0));
132 Seq[][] seqs = (Seq[][])walk(t.child(1));
133 for(Seq[] seq : seqs)
134 for(int i=0; i<seq.length; i++)
135 seq[i] = tag(tag, seq[i]);
136 return new NonTerminalNode(tag, seqs, false, null, false);
138 if (head.equals("TestCase"))
139 return new RegressionTests.TestCase(walkString(t.child(0)),
140 walkString(t.child(1)),
141 (String[])Reflection.coerce(walkChildren(t.child(2)), String[].class),
142 (Union)walk(t.child(3)),
145 if (head.equals("#import")) {
146 String fileName = (String)stringifyChildren(t.child(0));
147 for(File f : includes) {
148 File file = new File(f.getAbsolutePath()+File.separatorChar+fileName);
149 if (!file.exists()) continue;
151 String newPrefix = t.size()<2 ? "" : ((String)walk(t.child(1))+".");
152 FileInputStream fis = new FileInputStream(file);
153 Tree tr = new CharParser(MetaGrammar.newInstance()).parse(fis).expand1();
154 return (GrammarNode)new GrammarBuilder(includes, newPrefix).walk(tr);
155 } catch (Exception e) {
156 throw new RuntimeException("while parsing " + file, e);
159 throw new RuntimeException("unable to find #include file \""+fileName+"\"");
161 throw new RuntimeException("unknown head: \"" + head + "\" => " + (head.equals("...")));
164 // Nodes //////////////////////////////////////////////////////////////////////////////
166 /** Root node of a grammar's AST; a set of named nonterminals */
167 private class GrammarNode extends HashMap<String,NonTerminalNode> {
168 public GrammarNode(NonTerminalNode[] nonterminals) {
169 for(NonTerminalNode nt : nonterminals) {
170 if (nt==null) continue;
171 if (this.get(nt.name)!=null)
172 throw new RuntimeException("duplicate definition of nonterminal \""+nt.name+"\"");
173 this.put(nt.name, nt);
176 public GrammarNode(Object[] nt) { add(nt); }
177 private void add(Object o) {
179 else if (o instanceof Object[]) for(Object o2 : (Object[])o) add(o2);
180 else if (o instanceof NonTerminalNode) {
181 NonTerminalNode nt = (NonTerminalNode)o;
182 if (this.get(nt.name)!=null)
183 throw new RuntimeException("duplicate definition of nonterminal \""+nt.name+"\"");
184 this.put(nt.name, nt);
186 else if (o instanceof GrammarNode)
187 for(NonTerminalNode n : ((GrammarNode)o).values())
190 public String toString() {
192 for(NonTerminalNode nt : values()) ret += nt + ", ";
195 public Union build(String rootNonterminal) {
196 Context cx = new Context(this);
198 for(GrammarBuilder.NonTerminalNode nt : values())
199 if (nt.name.equals(rootNonterminal))
200 return (Union)cx.get(nt.name);
205 public class UnionNode extends ElementNode {
206 public Seq[][] sequences;
207 public String sep = null;
209 public UnionNode(Seq[][] sequences) { this(sequences, false, null); }
210 public UnionNode(Seq[][] sequences, boolean rep, String sep) {
211 this.sequences = sequences;
215 public Atom toAtom(Context cx) {
217 for(Seq[] ss : sequences)
219 ret = ret==null ? s.toAtom(cx) : (Atom)ret.union(s.toAtom(cx));
222 public Element build(Context cx, NonTerminalNode cnt, boolean dropall) {
223 return buildIntoPreallocatedUnion(cx, cnt, dropall, new Union(null, false)); }
224 public Element buildIntoPreallocatedUnion(Context cx, NonTerminalNode cnt, boolean dropall, Union u) {
227 urep = new Union(null, false);
228 urep.add(Sequence.create(new Element[0], cnt.name));
230 ? Sequence.create(new Element[] { u }, 0)
231 : Sequence.create(new Element[] { cx.get(sep), u }, 1));
233 HashSet<Sequence> bad2 = new HashSet<Sequence>();
234 for(int i=0; i<sequences.length; i++) {
235 Seq[] group = sequences[i];
236 Union u2 = new Union(null, false);
237 if (sequences.length==1) u2 = u;
238 for(int j=0; j<group.length; j++)
240 group[j].build(cx, u2, cnt, dropall);
242 Union u3 = new Union(null, false);
243 group[j].build(cx, u3, cnt, dropall);
244 Sequence s = Sequence.create(cnt.name,
245 new Element[] { u3, urep },
246 new boolean[] { false, false },
250 if (sequences.length==1) break;
251 Sequence seq = Sequence.create(u2);
252 for(Sequence s : bad2) seq = seq.andnot(s);
254 bad2.add(Sequence.create(u2));
260 public class NonTerminalNode extends UnionNode {
261 public boolean alwaysDrop;
262 public String name = null;
263 public boolean drop(Context cx) { return alwaysDrop; }
264 public NonTerminalNode(String name, Seq[][] sequences, boolean rep, String sep, boolean alwaysDrop) {
265 super(sequences, rep, sep==null?null:(prefix + sep));
266 this.name = prefix + name;
267 this.alwaysDrop = alwaysDrop;
269 public Element build(Context cx, NonTerminalNode cnt, boolean dropall) { return cx.get(name); }
273 public boolean alwaysDrop = false;
274 public boolean drop(Context cx) { return alwaysDrop; }
275 HashSet<Seq> and = new HashSet<Seq>();
276 HashSet<Seq> not = new HashSet<Seq>();
277 ElementNode[] elements;
280 public Seq(ElementNode e) { this(new ElementNode[] { e }); }
281 public Seq(ElementNode[] elements) {
282 this.elements = elements;
283 for(int i=0; i<elements.length; i++)
284 if (elements[i]==null)
285 throw new RuntimeException();
287 public Atom toAtom(Context cx) {
288 if (elements.length != 1)
289 throw new Error("you attempted to use ->, **, ++, or a similar character-class"+
290 " operator on a [potentially] multicharacter production");
291 return elements[0].toAtom(cx);
293 public Seq tag(String tag) { this.tag = tag; return this; }
294 public Seq follow(ElementNode follow) { this.follow = follow; return this; }
295 public Seq and(Seq s) { and.add(s); return this; }
296 public Seq andnot(Seq s) { not.add(s); return this; }
298 Seq ret = new Seq(elements);
305 public Seq separate(ElementNode sep) {
306 ElementNode[] elements = new ElementNode[this.elements.length * 2 - 1];
307 for(int i=0; i<this.elements.length; i++) {
308 elements[i*2] = this.elements[i];
309 if (i<this.elements.length-1)
310 elements[i*2+1] = new DropNode(sep);
312 this.elements = elements;
315 public Sequence build(Context cx, Union u, NonTerminalNode cnt, boolean dropall) {
316 Sequence ret = build0(cx, cnt, dropall);
317 for(Seq s : and) ret = ret.and(s.build(cx, null, cnt, true));
318 for(Seq s : not) ret = ret.andnot(s.build(cx, null, cnt, true));
319 if (u!=null) u.add(ret);
322 public Sequence build0(Context cx, NonTerminalNode cnt, boolean dropall) {
323 boolean[] drops = new boolean[elements.length];
324 Element[] els = new Element[elements.length];
326 for(int i=0; i<elements.length; i++) {
327 if (dropall) drops[i] = true;
328 else drops[i] = elements[i].drop(cx);
329 if (elements[i].getOwnerTag() != null)
330 tag = elements[i].getOwnerTag();
334 boolean multiNonDrop = false;
335 for(int i=0; i<drops.length; i++)
337 if (idx==-1) idx = i;
338 else multiNonDrop = true;
339 for(int i=0; i<elements.length; i++) {
340 if (!multiNonDrop && i==idx && tag!=null && elements[i] instanceof RepeatNode) {
341 els[i] = ((RepeatNode)elements[i]).build(cx, cnt, dropall, tag);
344 els[i] = elements[i].build(cx, cnt, dropall);
346 if (tag==null && multiNonDrop)
347 throw new RuntimeException("multiple non-dropped elements in sequence: " + Sequence.create(els, ""));
348 boolean lift = false;
349 if (elements.length > 0 && elements[0].lifted)
354 ? Sequence.create(els, illegalTag)
355 : Sequence.createLeft(tag, els, drops, lift);
356 else if (tag==null) ret = Sequence.create(els, idx);
357 else ret = Sequence.createLeft(tag, els, drops, lift);
360 ret = Sequence.createLeft(tag, els, drops, lift);
361 if (this.follow != null)
362 ret = ret.followedBy(this.follow.toAtom(cx));
367 public class ReferenceNode extends ElementNode {
368 public String nonTerminal;
369 public ReferenceNode() { }
370 public NonTerminalNode resolve(Context cx) {
371 NonTerminalNode ret = cx.grammar.get(nonTerminal);
372 if (ret==null) throw new RuntimeException("undefined nonterminal: " + nonTerminal);
375 public ReferenceNode(String nonTerminal) { this.nonTerminal = prefix + nonTerminal; }
376 public Atom toAtom(Context cx) { return cx.grammar.get(nonTerminal).toAtom(cx); }
377 public boolean drop(Context cx) { return resolve(cx).drop(cx); }
378 public Element build(Context cx, NonTerminalNode cnt, boolean dropall) {
379 if (!this.nonTerminal.startsWith(prefix)) nonTerminal = prefix + nonTerminal;
380 Element ret = cx.get(nonTerminal);
381 if (ret == null) throw new RuntimeException("unknown nonterminal \""+nonTerminal+"\"");
386 public class LiteralNode extends ElementNode {
387 private String string;
388 private final String thePrefix = prefix;
389 private boolean caret;
390 public LiteralNode(String string) { this(string, false); }
391 public LiteralNode(String string, boolean caret) {
392 this.string = string;
395 public String getOwnerTag() { return caret ? thePrefix+string : super.getOwnerTag(); }
396 public String toString() { return "\""+string+"\""; }
397 public boolean drop(Context cx) { return true; }
398 public Atom toAtom(Context cx) {
399 if (string.length()!=1) return super.toAtom(cx);
400 edu.berkeley.sbp.util.Range.Set set = new edu.berkeley.sbp.util.Range.Set();
401 set.add(string.charAt(0), string.charAt(0));
402 return CharAtom.set(set);
404 public Element build(Context cx, NonTerminalNode cnt, boolean dropall) { return CharAtom.string(string); }
407 public class AtomNode extends ElementNode {
408 AtomNodeRange[] ranges;
409 public AtomNode() { this(new AtomNodeRange[0]); }
410 public AtomNode(AtomNodeRange[] ranges) { this.ranges = ranges; }
411 public AtomNode(AtomNodeRange range) { this.ranges = new AtomNodeRange[] { range }; }
412 public Element build(Context cx, NonTerminalNode cnt, boolean dropall) { return toAtom(cx); }
413 public Atom toAtom(Context cx) {
414 edu.berkeley.sbp.util.Range.Set set = new edu.berkeley.sbp.util.Range.Set();
415 for(AtomNodeRange r : ranges) set.add(r.first, r.last);
416 return CharAtom.set(set);
419 public class AtomNodeRange {
422 public AtomNodeRange(char only) { first = only; last = only; }
423 public AtomNodeRange(char first, char last) { this.first = first; this.last = last; }
426 public class RepeatNode extends ElementNode {
427 public ElementNode e, sep;
428 public final boolean zero, many, max;
429 public RepeatNode(ElementNode e, ElementNode sep, boolean zero, boolean many, boolean max) {
430 this.e = e; this.sep = sep; this.zero = zero; this.many = many; this.max = max;
432 public Atom toAtom(Context cx) { return sep==null ? e.toAtom(cx) : super.toAtom(cx); }
433 public boolean drop(Context cx) { return e.drop(cx); }
434 public Element build(Context cx, NonTerminalNode cnt, boolean dropall) {
435 Element ret = build(cx, cnt, dropall, illegalTag);
436 String must = "must be tagged unless they appear within a dropped expression or their contents are dropped: ";
437 if (!dropall && !drop(cx) && !e.drop(cx))
438 if (!many) throw new RuntimeException("options (?) " + must + ret);
439 else if (zero) throw new RuntimeException("zero-or-more repetitions (*) " + must + ret);
440 else throw new RuntimeException("one-or-more repetitions (+) " + must + ret);
443 public Element build(Context cx, NonTerminalNode cnt, boolean dropall, Object repeatTag) {
445 ? Repeat.repeat(e.build(cx, null, dropall), zero, many, sep==null ? null : sep.build(cx, null, dropall), repeatTag)
447 ? Repeat.repeatMaximal(e.toAtom(cx), zero, many, repeatTag)
448 : Repeat.repeatMaximal(e.build(cx, null, dropall), zero, many, sep.toAtom(cx), repeatTag);
452 public abstract class ElementNode {
453 public boolean lifted = false;
454 public String getOwnerTag() { return null; }
455 public boolean drop(Context cx) { return false; }
456 public Atom toAtom(Context cx) { throw new Error("can't convert a " + this.getClass().getName() + " to an atom: " + this); }
457 public abstract Element build(Context cx, NonTerminalNode cnt, boolean dropall);
460 public abstract class ElementNodeWrapper extends ElementNode {
461 protected ElementNode _e;
462 public ElementNodeWrapper(ElementNode e) { this._e = e; }
463 public String getOwnerTag() { return _e.getOwnerTag(); }
464 public boolean drop(Context cx) { return _e.drop(cx); }
465 public Atom toAtom(Context cx) { return _e.toAtom(cx); }
466 public Element build(Context cx, NonTerminalNode cnt, boolean dropall) { return _e.build(cx, cnt, dropall); }
469 public class TildeNode extends ElementNodeWrapper {
470 public TildeNode(ElementNode e) { super(e); }
471 public Atom toAtom(Context cx) { return (Atom)((Topology<Character>)_e.toAtom(cx).complement()); }
472 public Element build(Context cx, NonTerminalNode cnt, boolean dropall) { return toAtom(cx); }
475 public class DropNode extends ElementNodeWrapper {
476 public DropNode(ElementNode e) { super(e); }
477 public boolean drop(Context cx) { return true; }
480 public Seq and2(Seq s, Seq a) { a.alwaysDrop = true; return s.and(a); }
481 public Seq andnot2(Seq s, Seq a) { a.alwaysDrop = true; return s.andnot(a); }
482 public Seq arrow(Seq s, ElementNode e) { return s.follow(e); }
483 public Seq tag(String tagname, Seq s) { return s.tag(tagname); }
484 public Seq seq(ElementNode[] elements) { return new Seq(elements); }
485 public Seq seq2(ElementNode[] elements) { return new Seq(elements); }
486 public ElementNode plusmax(final ElementNode e) { return new RepeatNode(e, null, false, true, true); }
487 public ElementNode plus(final ElementNode e) { return new RepeatNode(e, null, false, true, false); }
488 public ElementNode plusmaxfollow(final ElementNode e, final ElementNode sep) { return new RepeatNode(e, sep, false, true, true); }
489 public ElementNode plusfollow(final ElementNode e, final ElementNode sep) { return new RepeatNode(e, sep, false, true, false); }
490 public ElementNode starmax(final ElementNode e) { return new RepeatNode(e, null, true, true, true); }
491 public ElementNode star(final ElementNode e) { return new RepeatNode(e, null, true, true, false); }
492 public ElementNode starmaxfollow(final ElementNode e, final ElementNode sep) { return new RepeatNode(e, sep, true, true, true); }
493 public ElementNode starfollow(final ElementNode e, final ElementNode sep) { return new RepeatNode(e, sep, true, true, false); }
494 public ElementNode question(final ElementNode e) { return new RepeatNode(e, null, true, false, false); }
496 //////////////////////////////////////////////////////////////////////////////
498 public class Context {
499 public HashMap<String,Union> map = new HashMap<String,Union>();
500 public GrammarNode grammar;
501 public Context(Tree t) { }
502 public Context(GrammarNode g) { this.grammar = g; }
503 public Union build() {
505 for(NonTerminalNode nt : grammar.values()) {
506 Union u = get(nt.name);
507 if ("s".equals(nt.name))
512 public Union peek(String name) { return map.get(name); }
513 public void put(String name, Union u) { map.put(name, u); }
514 public Union get(String name) {
515 Union ret = map.get(name);
516 if (ret != null) return ret;
517 NonTerminalNode nt = grammar.get(name);
519 throw new Error("warning could not find " + name);
521 ret = new Union(name, false);
523 nt.buildIntoPreallocatedUnion(this, nt, nt.drop(this), ret);