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.*;
13 // FIXME: deal with question-marks
15 // FIXME: put back in associativity: ^")"
17 // means that the FIRST "A" cannot match the SAME sequence in
18 // which the (A) occurs
19 // -- and add a test case
21 // FIXME: make it so that we can have multi-result nonterminals
22 // so long as they always appear under double-colons?
23 // auto-insert the unwrap?
25 // FIXME: put associativity back in
27 // FIXME: "flat" sequences (no subtrees contain "::"s?)
29 // freezing problem is related to comments in grammar files
31 /** The java classes typically used to represent a parsed grammar AST; each inner class is a type of AST node. */
32 public class GrammarBuilder {
35 * Create a grammar from a parse tree and binding resolver
37 * @param t a tree produced by parsing a grammar using the metagrammar
38 * @param s the name of the "start symbol"
39 * @param gbr a GrammarBindingResolver that resolves grammatical reductions into tree-node-heads
41 public static Union buildFromAST(Tree grammarAST, String startingNonterminal, File[] includes) {
42 return new GrammarAST(includes, "").buildGrammar(grammarAST, startingNonterminal);
45 public static Object illegalTag = ""; // this is the tag that should never appear in the non-dropped output FIXME
47 private final String prefix;
48 private final File[] includes;
50 public GrammarAST(File[] includes, String prefix) {
52 this.includes = includes;
55 // Methods //////////////////////////////////////////////////////////////////////////////
57 private Union buildGrammar(Tree t, String rootNonTerminal) {
58 return ((GrammarAST.GrammarNode)walk(t)).build(rootNonTerminal);
61 public Object[] walkChildren(Tree t) {
62 Object[] ret = new Object[t.size()];
63 for(int i=0; i<ret.length; i++) {
64 ret[i] = walk(t.child(i));
65 if (ret[i] instanceof Object[])
66 ret[i] = Reflection.lub((Object[])ret[i]);
68 return Reflection.lub(ret);
70 public String stringifyChildren(Tree t) {
71 StringBuffer sb = new StringBuffer();
72 for(int i=0; i<t.size(); i++) {
73 sb.append(t.child(i).head());
74 sb.append(stringifyChildren(t.child(i)));
78 public String unescape(Tree t) {
79 StringBuffer sb = new StringBuffer();
80 for(int i=0; i<t.size(); i++)
81 sb.append(t.child(i).head()+stringifyChildren(t.child(i)));
85 public Object walk(Tree t) {
86 String head = (String)t.head();
87 while(head.indexOf('.') > 0)
88 head = head.substring(head.indexOf('.')+1);
89 if (head==null) throw new RuntimeException("head is null: " + t);
90 if (head.equals("|")) return walkChildren(t);
91 if (head.equals("RHS")) return walkChildren(t);
92 if (head.equals("Grammar")) return new GrammarNode(walkChildren(t));
93 if (head.equals("(")) return new UnionNode((Seq[][])walkChildren(t.child(0)));
94 if (head.equals("Word")) return stringifyChildren(t);
95 if (head.equals("Elements")) return seq2((ElementNode[])Reflection.rebuild(walkChildren(t), ElementNode[].class));
96 if (head.equals("NonTerminalReference")) return new ReferenceNode(stringifyChildren(t.child(0)));
97 //if (head.equals(")")) return new ReferenceNode(stringifyChildren(t.child(0)));
98 if (head.equals("{")) return new XTree((Seq)walk(t.child(0)));
99 if (head.equals("::")) return tag((String)walk(t.child(0)), (Seq)walk(t.child(1)));
100 if (head.equals("++")) return plusmax((ElementNode)walk(t.child(0)));
101 if (head.equals("...")) return star(new TildeNode(new AtomNode()));
102 if (head.equals("+")) return plus((ElementNode)walk(t.child(0)));
103 if (head.equals("++/")) return plusmaxfollow((ElementNode)walk(t.child(0)), (ElementNode)walk(t.child(1)));
104 if (head.equals("+/")) return plusfollow((ElementNode)walk(t.child(0)), (ElementNode)walk(t.child(1)));
105 if (head.equals("**")) return starmax((ElementNode)walk(t.child(0)));
106 if (head.equals("*")) return star((ElementNode)walk(t.child(0)));
107 if (head.equals("**/")) return starmaxfollow((ElementNode)walk(t.child(0)), (ElementNode)walk(t.child(1)));
108 if (head.equals("*/")) return starfollow((ElementNode)walk(t.child(0)), (ElementNode)walk(t.child(1)));
109 if (head.equals("?")) return question((ElementNode)walk(t.child(0)));
110 if (head.equals("!")) return new DropNode((ElementNode)walk(t.child(0)));
111 if (head.equals("^")) return new LiteralNode((String)walk(t.child(0)), true);
112 if (head.equals("`")) {
113 ElementNode ret = (ElementNode)walk(t.child(0));
117 if (head.equals("Quoted")) return stringifyChildren(t);
118 if (head.equals("Literal")) return new LiteralNode((String)walk(t.child(0)));
119 if (head.equals("->")) return arrow((Seq)walk(t.child(0)), (ElementNode)walk(t.child(1)));
120 if (head.equals("DropNT")) return new NonTerminalNode((String)walk(t.child(0)), (Seq[][])walkChildren(t.child(1)), false, null, true);
121 if (head.equals("=") && t.size()==2) return new NonTerminalNode((String)walk(t.child(0)), (Seq[][])walk(t.child(1)), true, null, false);
122 if (head.equals("=")) return new NonTerminalNode((String)walk(t.child(0)), (Seq[][])walk(t.child(2)), true, (String)walk(t.child(1)), false);
123 if (head.equals("&")) return and2((Seq)walk(t.child(0)), (Seq)walk(t.child(1)));
124 if (head.equals("&~")) return andnot2((Seq)walk(t.child(0)), (Seq)walk(t.child(1)));
125 if (head.equals("/")) return ((Seq)walk(t.child(0))).separate((ElementNode)walk(t.child(1)));
126 if (head.equals("()")) return epsilon;
127 if (head.equals("[")) return new AtomNode((AtomNodeRange[])Reflection.rebuild(walkChildren(t), AtomNodeRange[].class));
128 if (head.equals("\\{")) return new DropNode(new AtomNode(new AtomNodeRange(CharAtom.left, CharAtom.left)));
129 if (head.equals("\\}")) return new DropNode(new AtomNode(new AtomNodeRange(CharAtom.right, CharAtom.right)));
130 if (head.equals("~")) return new TildeNode((ElementNode)walk(t.child(0)));
131 if (head.equals("~~")) {
132 Seq seq = new Seq(star(new TildeNode(new AtomNode())));
133 return seq.andnot((Seq)walk(t.child(0)));
135 if (head.equals("Range") && t.size()==1) return new AtomNodeRange(unescape(t).charAt(0));
136 if (head.equals("Range")) return new AtomNodeRange(unescape(t).charAt(0), unescape(t).charAt(1));
137 if (head.equals("\"\"")) return "";
138 if (head.equals("\n")) return "\n";
139 if (head.equals("\r")) return "\r";
140 if (head.equals("grammar.Grammar")) return walkChildren(t);
141 if (head.equals("SubGrammar")) return GrammarBuilder.buildFromAST(t.child(0), "s", includes);
142 if (head.equals("NonTerminal"))
143 return new NonTerminalNode((String)walk(t.child(0)),
144 (Seq[][])walkChildren(t.child(1)), false, null, false);
145 if (head.equals("Colons")) {
146 String tag = (String)walk(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] = tag(tag, seq[i]);
151 return new NonTerminalNode(tag, seqs, false, null, false);
153 if (head.equals("TestCase"))
154 return new RegressionTests.TestCase(walkString(t.child(0)),
155 walkString(t.child(1)),
156 (String[])Reflection.coerce(walkChildren(t.child(2)), String[].class),
157 (Union)walk(t.child(3)),
160 if (head.equals("#import")) {
161 String fileName = (String)stringifyChildren(t.child(0));
162 for(File f : includes) {
163 File file = new File(f.getAbsolutePath()+File.separatorChar+fileName);
164 if (!file.exists()) continue;
166 String newPrefix = t.size()<2 ? "" : ((String)walk(t.child(1))+".");
167 FileInputStream fis = new FileInputStream(file);
168 Tree tr = new CharParser(MetaGrammar.newInstance()).parse(fis).expand1();
169 return (GrammarNode)new GrammarBuilder(includes, newPrefix).walk(tr);
170 } catch (Exception e) {
171 throw new RuntimeException("while parsing " + file, e);
174 throw new RuntimeException("unable to find #include file \""+fileName+"\"");
176 throw new RuntimeException("unknown head: \"" + head + "\" => " + (head.equals("...")));
179 // Nodes //////////////////////////////////////////////////////////////////////////////
181 /** Root node of a grammar's AST; a set of named nonterminals */
182 private class GrammarNode extends HashMap<String,NonTerminalNode> {
183 public GrammarNode(NonTerminalNode[] nonterminals) {
184 for(NonTerminalNode nt : nonterminals) {
185 if (nt==null) continue;
186 if (this.get(nt.name)!=null)
187 throw new RuntimeException("duplicate definition of nonterminal \""+nt.name+"\"");
188 this.put(nt.name, nt);
191 public GrammarNode(Object[] nt) { add(nt); }
192 private void add(Object o) {
194 else if (o instanceof Object[]) for(Object o2 : (Object[])o) add(o2);
195 else if (o instanceof NonTerminalNode) {
196 NonTerminalNode nt = (NonTerminalNode)o;
197 if (this.get(nt.name)!=null)
198 throw new RuntimeException("duplicate definition of nonterminal \""+nt.name+"\"");
199 this.put(nt.name, nt);
201 else if (o instanceof GrammarNode)
202 for(NonTerminalNode n : ((GrammarNode)o).values())
205 public String toString() {
207 for(NonTerminalNode nt : values()) ret += nt + ", ";
210 public Union build(String rootNonterminal) {
211 Context cx = new Context(this);
213 for(GrammarBuilder.NonTerminalNode nt : values())
214 if (nt.name.equals(rootNonterminal))
215 return (Union)cx.get(nt.name);
220 public class UnionNode extends ElementNode {
221 public Seq[][] sequences;
222 public String sep = null;
224 public UnionNode(Seq[][] sequences) { this(sequences, false, null); }
225 public UnionNode(Seq[][] sequences, boolean rep, String sep) {
226 this.sequences = sequences;
230 public Atom toAtom(Context cx) {
232 for(Seq[] ss : sequences)
234 ret = ret==null ? s.toAtom(cx) : (Atom)ret.union(s.toAtom(cx));
237 public Element build(Context cx, NonTerminalNode cnt, boolean dropall) {
238 return buildIntoPreallocatedUnion(cx, cnt, dropall, new Union(null, false)); }
239 public Element buildIntoPreallocatedUnion(Context cx, NonTerminalNode cnt, boolean dropall, Union u) {
242 urep = new Union(null, false);
243 urep.add(Sequence.create(new Element[0], cnt.name));
245 ? Sequence.create(new Element[] { u }, 0)
246 : Sequence.create(new Element[] { cx.get(sep), u }, 1));
248 HashSet<Sequence> bad2 = new HashSet<Sequence>();
249 for(int i=0; i<sequences.length; i++) {
250 Seq[] group = sequences[i];
251 Union u2 = new Union(null, false);
252 if (sequences.length==1) u2 = u;
253 for(int j=0; j<group.length; j++)
255 group[j].build(cx, u2, cnt, dropall);
257 Union u3 = new Union(null, false);
258 group[j].build(cx, u3, cnt, dropall);
259 Sequence s = Sequence.create(cnt.name,
260 new Element[] { u3, urep },
261 new boolean[] { false, false },
265 if (sequences.length==1) break;
266 Sequence seq = Sequence.create(u2);
267 for(Sequence s : bad2) seq = seq.andnot(s);
269 bad2.add(Sequence.create(u2));
275 public class NonTerminalNode extends UnionNode {
276 public boolean alwaysDrop;
277 public String name = null;
278 public boolean drop(Context cx) { return alwaysDrop; }
279 public NonTerminalNode(String name, Seq[][] sequences, boolean rep, String sep, boolean alwaysDrop) {
280 super(sequences, rep, sep==null?null:(prefix + sep));
281 this.name = prefix + name;
282 this.alwaysDrop = alwaysDrop;
284 public Element build(Context cx, NonTerminalNode cnt, boolean dropall) { return cx.get(name); }
288 public boolean alwaysDrop = false;
289 public boolean drop(Context cx) { return alwaysDrop; }
290 HashSet<Seq> and = new HashSet<Seq>();
291 HashSet<Seq> not = new HashSet<Seq>();
292 ElementNode[] elements;
295 public Seq(ElementNode e) { this(new ElementNode[] { e }); }
296 public Seq(ElementNode[] elements) {
297 this.elements = elements;
298 for(int i=0; i<elements.length; i++)
299 if (elements[i]==null)
300 throw new RuntimeException();
302 public Atom toAtom(Context cx) {
303 if (elements.length != 1)
304 throw new Error("you attempted to use ->, **, ++, or a similar character-class"+
305 " operator on a [potentially] multicharacter production");
306 return elements[0].toAtom(cx);
308 public Seq tag(String tag) { this.tag = tag; return this; }
309 public Seq follow(ElementNode follow) { this.follow = follow; return this; }
310 public Seq and(Seq s) { and.add(s); return this; }
311 public Seq andnot(Seq s) { not.add(s); return this; }
313 Seq ret = new Seq(elements);
320 public Seq separate(ElementNode sep) {
321 ElementNode[] elements = new ElementNode[this.elements.length * 2 - 1];
322 for(int i=0; i<this.elements.length; i++) {
323 elements[i*2] = this.elements[i];
324 if (i<this.elements.length-1)
325 elements[i*2+1] = new DropNode(sep);
327 this.elements = elements;
330 public Sequence build(Context cx, Union u, NonTerminalNode cnt, boolean dropall) {
331 Sequence ret = build0(cx, cnt, dropall);
332 for(Seq s : and) ret = ret.and(s.build(cx, null, cnt, true));
333 for(Seq s : not) ret = ret.andnot(s.build(cx, null, cnt, true));
334 if (u!=null) u.add(ret);
337 public Sequence build0(Context cx, NonTerminalNode cnt, boolean dropall) {
338 boolean[] drops = new boolean[elements.length];
339 Element[] els = new Element[elements.length];
341 for(int i=0; i<elements.length; i++) {
342 if (dropall) drops[i] = true;
343 else drops[i] = elements[i].drop(cx);
344 if (elements[i].getOwnerTag() != null)
345 tag = elements[i].getOwnerTag();
349 boolean multiNonDrop = false;
350 for(int i=0; i<drops.length; i++)
352 if (idx==-1) idx = i;
353 else multiNonDrop = true;
354 for(int i=0; i<elements.length; i++) {
355 if (!multiNonDrop && i==idx && tag!=null && elements[i] instanceof RepeatNode) {
356 els[i] = ((RepeatNode)elements[i]).build(cx, cnt, dropall, tag);
359 els[i] = elements[i].build(cx, cnt, dropall);
361 if (tag==null && multiNonDrop)
362 throw new RuntimeException("multiple non-dropped elements in sequence: " + Sequence.create(els, ""));
363 boolean lift = false;
364 if (elements.length > 0 && elements[0].lifted)
369 ? Sequence.create(els, illegalTag)
370 : Sequence.createLeft(tag, els, drops, lift);
371 else if (tag==null) ret = Sequence.create(els, idx);
372 else ret = Sequence.createLeft(tag, els, drops, lift);
375 ret = Sequence.createLeft(tag, els, drops, lift);
376 if (this.follow != null)
377 ret = ret.followedBy(this.follow.toAtom(cx));
382 public class ReferenceNode extends ElementNode {
383 public String nonTerminal;
384 public ReferenceNode() { }
385 public NonTerminalNode resolve(Context cx) {
386 NonTerminalNode ret = cx.grammar.get(nonTerminal);
387 if (ret==null) throw new RuntimeException("undefined nonterminal: " + nonTerminal);
390 public ReferenceNode(String nonTerminal) { this.nonTerminal = prefix + nonTerminal; }
391 public Atom toAtom(Context cx) { return cx.grammar.get(nonTerminal).toAtom(cx); }
392 public boolean drop(Context cx) { return resolve(cx).drop(cx); }
393 public Element build(Context cx, NonTerminalNode cnt, boolean dropall) {
394 if (!this.nonTerminal.startsWith(prefix)) nonTerminal = prefix + nonTerminal;
395 Element ret = cx.get(nonTerminal);
396 if (ret == null) throw new RuntimeException("unknown nonterminal \""+nonTerminal+"\"");
401 public class LiteralNode extends ElementNode {
402 private String string;
403 private final String thePrefix = prefix;
404 private boolean caret;
405 public LiteralNode(String string) { this(string, false); }
406 public LiteralNode(String string, boolean caret) {
407 this.string = string;
410 public String getOwnerTag() { return caret ? thePrefix+string : super.getOwnerTag(); }
411 public String toString() { return "\""+string+"\""; }
412 public boolean drop(Context cx) { return true; }
413 public Atom toAtom(Context cx) {
414 if (string.length()!=1) return super.toAtom(cx);
415 edu.berkeley.sbp.util.Range.Set set = new edu.berkeley.sbp.util.Range.Set();
416 set.add(string.charAt(0), string.charAt(0));
417 return CharAtom.set(set);
419 public Element build(Context cx, NonTerminalNode cnt, boolean dropall) { return CharAtom.string(string); }
422 public class AtomNode extends ElementNode {
423 AtomNodeRange[] ranges;
424 public AtomNode() { this(new AtomNodeRange[0]); }
425 public AtomNode(AtomNodeRange[] ranges) { this.ranges = ranges; }
426 public AtomNode(AtomNodeRange range) { this.ranges = new AtomNodeRange[] { range }; }
427 public Element build(Context cx, NonTerminalNode cnt, boolean dropall) { return toAtom(cx); }
428 public Atom toAtom(Context cx) {
429 edu.berkeley.sbp.util.Range.Set set = new edu.berkeley.sbp.util.Range.Set();
430 for(AtomNodeRange r : ranges) set.add(r.first, r.last);
431 return CharAtom.set(set);
434 public class AtomNodeRange {
437 public AtomNodeRange(char only) { first = only; last = only; }
438 public AtomNodeRange(char first, char last) { this.first = first; this.last = last; }
441 public class RepeatNode extends ElementNode {
442 public ElementNode e, sep;
443 public final boolean zero, many, max;
444 public RepeatNode(ElementNode e, ElementNode sep, boolean zero, boolean many, boolean max) {
445 this.e = e; this.sep = sep; this.zero = zero; this.many = many; this.max = max;
447 public Atom toAtom(Context cx) { return sep==null ? e.toAtom(cx) : super.toAtom(cx); }
448 public boolean drop(Context cx) { return e.drop(cx); }
449 public Element build(Context cx, NonTerminalNode cnt, boolean dropall) {
450 Element ret = build(cx, cnt, dropall, illegalTag);
451 String must = "must be tagged unless they appear within a dropped expression or their contents are dropped: ";
452 if (!dropall && !drop(cx) && !e.drop(cx))
453 if (!many) throw new RuntimeException("options (?) " + must + ret);
454 else if (zero) throw new RuntimeException("zero-or-more repetitions (*) " + must + ret);
455 else throw new RuntimeException("one-or-more repetitions (+) " + must + ret);
458 public Element build(Context cx, NonTerminalNode cnt, boolean dropall, Object repeatTag) {
460 ? Repeat.repeat(e.build(cx, null, dropall), zero, many, sep==null ? null : sep.build(cx, null, dropall), repeatTag)
462 ? Repeat.repeatMaximal(e.toAtom(cx), zero, many, repeatTag)
463 : Repeat.repeatMaximal(e.build(cx, null, dropall), zero, many, sep.toAtom(cx), repeatTag);
467 public abstract class ElementNode {
468 public boolean lifted = false;
469 public String getOwnerTag() { return null; }
470 public boolean drop(Context cx) { return false; }
471 public Atom toAtom(Context cx) { throw new Error("can't convert a " + this.getClass().getName() + " to an atom: " + this); }
472 public abstract Element build(Context cx, NonTerminalNode cnt, boolean dropall);
475 public abstract class ElementNodeWrapper extends ElementNode {
476 protected ElementNode _e;
477 public ElementNodeWrapper(ElementNode e) { this._e = e; }
478 public String getOwnerTag() { return _e.getOwnerTag(); }
479 public boolean drop(Context cx) { return _e.drop(cx); }
480 public Atom toAtom(Context cx) { return _e.toAtom(cx); }
481 public Element build(Context cx, NonTerminalNode cnt, boolean dropall) { return _e.build(cx, cnt, dropall); }
484 public class TildeNode extends ElementNodeWrapper {
485 public TildeNode(ElementNode e) { super(e); }
486 public Atom toAtom(Context cx) { return (Atom)((Topology<Character>)_e.toAtom(cx).complement()); }
487 public Element build(Context cx, NonTerminalNode cnt, boolean dropall) { return toAtom(cx); }
490 public class DropNode extends ElementNodeWrapper {
491 public DropNode(ElementNode e) { super(e); }
492 public boolean drop(Context cx) { return true; }
495 public Seq and2(Seq s, Seq a) { a.alwaysDrop = true; return s.and(a); }
496 public Seq andnot2(Seq s, Seq a) { a.alwaysDrop = true; return s.andnot(a); }
497 public Seq arrow(Seq s, ElementNode e) { return s.follow(e); }
498 public Seq tag(String tagname, Seq s) { return s.tag(tagname); }
499 public Seq seq(ElementNode[] elements) { return new Seq(elements); }
500 public Seq seq2(ElementNode[] elements) { return new Seq(elements); }
501 public ElementNode plusmax(final ElementNode e) { return new RepeatNode(e, null, false, true, true); }
502 public ElementNode plus(final ElementNode e) { return new RepeatNode(e, null, false, true, false); }
503 public ElementNode plusmaxfollow(final ElementNode e, final ElementNode sep) { return new RepeatNode(e, sep, false, true, true); }
504 public ElementNode plusfollow(final ElementNode e, final ElementNode sep) { return new RepeatNode(e, sep, false, true, false); }
505 public ElementNode starmax(final ElementNode e) { return new RepeatNode(e, null, true, true, true); }
506 public ElementNode star(final ElementNode e) { return new RepeatNode(e, null, true, true, false); }
507 public ElementNode starmaxfollow(final ElementNode e, final ElementNode sep) { return new RepeatNode(e, sep, true, true, true); }
508 public ElementNode starfollow(final ElementNode e, final ElementNode sep) { return new RepeatNode(e, sep, true, true, false); }
509 public ElementNode question(final ElementNode e) { return new RepeatNode(e, null, true, false, false); }
511 //////////////////////////////////////////////////////////////////////////////
513 public class Context {
514 public HashMap<String,Union> map = new HashMap<String,Union>();
515 public GrammarNode grammar;
516 public Context(Tree t) { }
517 public Context(GrammarNode g) { this.grammar = g; }
518 public Union build() {
520 for(NonTerminalNode nt : grammar.values()) {
521 Union u = get(nt.name);
522 if ("s".equals(nt.name))
527 public Union peek(String name) { return map.get(name); }
528 public void put(String name, Union u) { map.put(name, u); }
529 public Union get(String name) {
530 Union ret = map.get(name);
531 if (ret != null) return ret;
532 NonTerminalNode nt = grammar.get(name);
534 throw new Error("warning could not find " + name);
536 ret = new Union(name, false);
538 nt.buildIntoPreallocatedUnion(this, nt, nt.drop(this), ret);