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 GrammarBuilder(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 GrammarBuilder(String path) { this(path, ""); }
51 public GrammarBuilder(File[] includes, String prefix) {
53 this.includes = includes;
56 public Union buildGrammar(Tree t, String rootNonTerminal) {
57 return ((GrammarBuilder.GrammarNode)walk(t)).build(rootNonTerminal);
60 private ElementNode epsilon = new LiteralNode("");
62 public 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 if (ret[i] instanceof Object[])
67 ret[i] = Reflection.lub((Object[])ret[i]);
69 return Reflection.lub(ret);
71 public String stringifyChildren(Tree t) {
72 StringBuffer sb = new StringBuffer();
73 for(int i=0; i<t.size(); i++) {
74 sb.append(t.child(i).head());
75 sb.append(stringifyChildren(t.child(i)));
79 public String unescape(Tree t) {
80 StringBuffer sb = new StringBuffer();
81 for(int i=0; i<t.size(); i++)
82 sb.append(t.child(i).head()+stringifyChildren(t.child(i)));
86 public Object walk(Tree t) {
87 String head = (String)t.head();
88 while(head.indexOf('.') != -1) 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 plus((ElementNode)walk(t.child(0)));
102 if (head.equals("++/")) return plusmaxfollow((ElementNode)walk(t.child(0)), (ElementNode)walk(t.child(1)));
103 if (head.equals("+/")) return plusfollow((ElementNode)walk(t.child(0)), (ElementNode)walk(t.child(1)));
104 if (head.equals("**")) return starmax((ElementNode)walk(t.child(0)));
105 if (head.equals("*")) return star((ElementNode)walk(t.child(0)));
106 if (head.equals("**/")) return starmaxfollow((ElementNode)walk(t.child(0)), (ElementNode)walk(t.child(1)));
107 if (head.equals("*/")) return starfollow((ElementNode)walk(t.child(0)), (ElementNode)walk(t.child(1)));
108 if (head.equals("?")) return question((ElementNode)walk(t.child(0)));
109 if (head.equals("!")) return new DropNode((ElementNode)walk(t.child(0)));
110 if (head.equals("^")) return new LiteralNode((String)walk(t.child(0)), true);
111 if (head.equals("Quoted")) return stringifyChildren(t);
112 if (head.equals("Literal")) return new LiteralNode((String)walk(t.child(0)));
113 if (head.equals("->")) return arrow((Seq)walk(t.child(0)), (ElementNode)walk(t.child(1)));
114 if (head.equals("DropNT")) return new NonTerminalNode((String)walk(t.child(0)), (Seq[][])walkChildren(t.child(1)), false, null, true);
115 if (head.equals("=") && t.size()==2) return new NonTerminalNode((String)walk(t.child(0)), (Seq[][])walk(t.child(1)), true, null, false);
116 if (head.equals("=")) return new NonTerminalNode((String)walk(t.child(0)), (Seq[][])walk(t.child(2)), true, (String)walk(t.child(1)), false);
117 if (head.equals("&")) return and2((Seq)walk(t.child(0)), (Seq)walk(t.child(1)));
118 if (head.equals("&~")) return andnot2((Seq)walk(t.child(0)), (Seq)walk(t.child(1)));
119 if (head.equals("/")) return ((Seq)walk(t.child(0))).separate((ElementNode)walk(t.child(1)));
120 if (head.equals("()")) return epsilon;
121 if (head.equals("[")) return new AtomNode((AtomNodeRange[])Reflection.rebuild(walkChildren(t), AtomNodeRange[].class));
122 if (head.equals("\\{")) return new DropNode(new AtomNode(new AtomNodeRange(CharAtom.left, CharAtom.left)));
123 if (head.equals("\\}")) return new DropNode(new AtomNode(new AtomNodeRange(CharAtom.right, CharAtom.right)));
124 if (head.equals("~")) return new TildeNode((ElementNode)walk(t.child(0)));
125 if (head.equals("Range") && t.size()==1) return new AtomNodeRange(unescape(t).charAt(0));
126 if (head.equals("Range")) return new AtomNodeRange(unescape(t).charAt(0), unescape(t).charAt(1));
127 if (head.equals("\"\"")) return "";
128 if (head.equals("\n")) return "\n";
129 if (head.equals("\r")) return "\r";
130 if (head.equals("grammar.Grammar")) return walkChildren(t);
131 if (head.equals("SubGrammar")) return GrammarBuilder.buildFromAST(t.child(0), "s", includes);
132 if (head.equals("NonTerminal"))
133 return new NonTerminalNode((String)walk(t.child(0)),
134 (Seq[][])walkChildren(t.child(1)), false, null, false);
135 if (head.equals("Colons")) {
136 String tag = (String)walk(t.child(0));
137 Seq[][] seqs = (Seq[][])walk(t.child(1));
138 for(Seq[] seq : seqs)
139 for(int i=0; i<seq.length; i++)
140 seq[i] = tag(tag, seq[i]);
141 return new NonTerminalNode(tag, seqs, false, null, false);
143 if (head.equals("TestCase"))
144 return new RegressionTests.TestCase((String)walk(t.child(0)),
145 (String)walk(t.child(1)),
146 (String[])Reflection.coerce(walkChildren(t.child(2)), String[].class),
147 (Union)walk(t.child(3)),
150 if (head.equals("#import")) {
151 String fileName = (String)stringifyChildren(t.child(0));
152 for(File f : includes) {
153 File file = new File(f.getAbsolutePath()+File.separatorChar+fileName);
154 if (!file.exists()) continue;
156 String newPrefix = (String)walk(t.child(1));
157 if (newPrefix.length() > 0) newPrefix += ".";
158 FileInputStream fis = new FileInputStream(file);
159 Tree tr = new CharParser(MetaGrammar.newInstance()).parse(fis).expand1();
160 return (GrammarNode)new GrammarBuilder(includes, newPrefix).walk(tr);
161 } catch (Exception e) {
162 throw new RuntimeException("while parsing " + file, e);
165 throw new RuntimeException("unable to find #include file \""+fileName+"\"");
167 throw new RuntimeException("unknown head: " + head + "\n"+t);
170 /** A grammar (a set of nonterminals) */
171 public class GrammarNode extends HashMap<String,NonTerminalNode> {
172 public GrammarNode(NonTerminalNode[] nonterminals) {
173 for(NonTerminalNode nt : nonterminals) {
174 if (nt==null) continue;
175 if (this.get(nt.name)!=null)
176 throw new RuntimeException("duplicate definition of nonterminal \""+nt.name+"\"");
177 this.put(nt.name, nt);
180 public GrammarNode(Object[] nt) { add(nt); }
181 private void add(Object o) {
183 else if (o instanceof Object[]) for(Object o2 : (Object[])o) add(o2);
184 else if (o instanceof NonTerminalNode) {
185 NonTerminalNode nt = (NonTerminalNode)o;
186 if (this.get(nt.name)!=null)
187 throw new RuntimeException("duplicate definition of nonterminal \""+nt.name+"\"");
188 this.put(nt.name, nt);
190 else if (o instanceof GrammarNode)
191 for(NonTerminalNode n : ((GrammarNode)o).values())
194 public String toString() {
196 for(NonTerminalNode nt : values()) ret += nt + ", ";
199 public Union build(String rootNonterminal) {
200 Context cx = new Context(this);
202 for(GrammarBuilder.NonTerminalNode nt : values())
203 if (nt.name.equals(rootNonterminal))
204 return (Union)cx.get(nt.name);
209 public class UnionNode extends ElementNode {
210 public Seq[][] sequences;
211 public String sep = null;
213 public UnionNode(Seq[][] sequences) { this(sequences, false, null); }
214 public UnionNode(Seq[][] sequences, boolean rep, String sep) {
215 this.sequences = sequences;
219 public Atom toAtom(Context cx) {
221 for(Seq[] ss : sequences)
223 ret = ret==null ? s.toAtom(cx) : (Atom)ret.union(s.toAtom(cx));
226 public Element build(Context cx, NonTerminalNode cnt, boolean dropall) {
227 return buildIntoPreallocatedUnion(cx, cnt, dropall, new Union(null, false)); }
228 public Element buildIntoPreallocatedUnion(Context cx, NonTerminalNode cnt, boolean dropall, Union u) {
231 urep = new Union(null, false);
232 urep.add(Sequence.create(new Element[0], cnt.name));
234 ? Sequence.create(new Element[] { u }, 0)
235 : Sequence.create(new Element[] { cx.get(sep), u }, 1));
237 HashSet<Sequence> bad2 = new HashSet<Sequence>();
238 for(int i=0; i<sequences.length; i++) {
239 Seq[] group = sequences[i];
240 Union u2 = new Union(null, false);
241 if (sequences.length==1) u2 = u;
242 for(int j=0; j<group.length; j++)
244 group[j].build(cx, u2, cnt, dropall);
246 Union u3 = new Union(null, false);
247 group[j].build(cx, u3, cnt, dropall);
248 Sequence s = Sequence.create(cnt.name,
249 new Element[] { u3, urep },
250 new boolean[] { false, false },
254 if (sequences.length==1) break;
255 Sequence seq = Sequence.create(u2);
256 for(Sequence s : bad2) seq = seq.andnot(s);
258 bad2.add(Sequence.create(u2));
264 public class NonTerminalNode extends UnionNode {
265 public boolean alwaysDrop;
266 public String name = null;
267 public boolean drop() { return alwaysDrop; }
268 public NonTerminalNode(String name, Seq[][] sequences, boolean rep, String sep, boolean alwaysDrop) {
269 super(sequences, rep, sep==null?null:(prefix + sep));
270 this.name = prefix + name;
271 this.alwaysDrop = alwaysDrop;
273 public Element build(Context cx, NonTerminalNode cnt, boolean dropall) { return cx.get(name); }
277 public boolean alwaysDrop = false;
278 public boolean drop() { return alwaysDrop; }
279 HashSet<Seq> and = new HashSet<Seq>();
280 HashSet<Seq> not = new HashSet<Seq>();
281 ElementNode[] elements;
284 public Seq(ElementNode e) { this(new ElementNode[] { e }); }
285 public Seq(ElementNode[] elements) {
286 this.elements = elements;
287 for(int i=0; i<elements.length; i++)
288 if (elements[i]==null)
289 throw new RuntimeException();
291 public Atom toAtom(Context cx) {
292 if (elements.length != 1)
293 throw new Error("you attempted to use ->, **, ++, or a similar character-class"+
294 " operator on a [potentially] multicharacter production");
295 return elements[0].toAtom(cx);
297 public Seq tag(String tag) { this.tag = tag; return this; }
298 public Seq follow(ElementNode follow) { this.follow = follow; return this; }
299 public Seq and(Seq s) { and.add(s); return this; }
300 public Seq andnot(Seq s) { not.add(s); return this; }
302 Seq ret = new Seq(elements);
309 public Seq separate(ElementNode sep) {
310 ElementNode[] elements = new ElementNode[this.elements.length * 2 - 1];
311 for(int i=0; i<this.elements.length; i++) {
312 elements[i*2] = this.elements[i];
313 if (i<this.elements.length-1)
314 elements[i*2+1] = new DropNode(sep);
316 this.elements = elements;
319 public Sequence build(Context cx, Union u, NonTerminalNode cnt, boolean dropall) {
320 Sequence ret = build0(cx, cnt, dropall);
321 for(Seq s : and) ret = ret.and(s.build(cx, null, cnt, true));
322 for(Seq s : not) ret = ret.andnot(s.build(cx, null, cnt, true));
323 if (u!=null) u.add(ret);
326 public Sequence build0(Context cx, NonTerminalNode cnt, boolean dropall) {
327 boolean[] drops = new boolean[elements.length];
328 Element[] els = new Element[elements.length];
330 for(int i=0; i<elements.length; i++) {
331 if (dropall) drops[i] = true;
332 else drops[i] = elements[i].drop();
333 if (elements[i] instanceof ReferenceNode)
334 if (((ReferenceNode)elements[i]).resolve(cx).drop())
336 if (elements[i].getOwnerTag() != null)
337 tag = elements[i].getOwnerTag();
341 boolean multiNonDrop = false;
342 for(int i=0; i<drops.length; i++)
344 if (idx==-1) idx = i;
345 else multiNonDrop = true;
346 for(int i=0; i<elements.length; i++) {
347 if (!multiNonDrop && i==idx && tag!=null && elements[i] instanceof RepeatNode) {
348 els[i] = ((RepeatNode)elements[i]).build(cx, cnt, dropall, tag);
351 els[i] = elements[i].build(cx, cnt, dropall);
353 if (tag==null && multiNonDrop)
354 throw new RuntimeException("multiple non-dropped elements in sequence: " + Sequence.create(els, ""));
358 ? Sequence.create(els, illegalTag)
359 : Sequence.create(tag, els, drops, false);
360 else if (tag==null) ret = Sequence.create(els, idx);
361 else ret = Sequence.create(tag, els, drops, false);
364 ret = Sequence.create(tag, els, drops, false);
365 if (this.follow != null)
366 ret = ret.followedBy(this.follow.toAtom(cx));
371 public class ReferenceNode extends ElementNode {
372 public String nonTerminal;
373 public ReferenceNode() { }
374 public NonTerminalNode resolve(Context cx) { return cx.grammar.get(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 Element build(Context cx, NonTerminalNode cnt, boolean dropall) {
378 if (!this.nonTerminal.startsWith(prefix)) nonTerminal = prefix + nonTerminal;
379 Element ret = cx.get(nonTerminal);
380 if (ret == null) throw new RuntimeException("unknown nonterminal \""+nonTerminal+"\"");
385 public class LiteralNode extends ElementNode {
386 private String string;
387 private final String thePrefix = prefix;
388 private boolean caret;
389 public LiteralNode(String string) { this(string, false); }
390 public LiteralNode(String string, boolean caret) {
391 this.string = string;
394 public String getOwnerTag() { return caret ? thePrefix+string : super.getOwnerTag(); }
395 public String toString() { return "\""+string+"\""; }
396 public boolean drop() { return true; }
397 public Atom toAtom(Context cx) {
398 if (string.length()!=1) return super.toAtom(cx);
399 edu.berkeley.sbp.util.Range.Set set = new edu.berkeley.sbp.util.Range.Set();
400 set.add(string.charAt(0), string.charAt(0));
401 return CharAtom.set(set);
403 public Element build(Context cx, NonTerminalNode cnt, boolean dropall) { return CharAtom.string(string); }
406 public class AtomNode extends ElementNode {
407 AtomNodeRange[] ranges;
408 public AtomNode(AtomNodeRange[] ranges) { this.ranges = ranges; }
409 public AtomNode(AtomNodeRange range) { this.ranges = new AtomNodeRange[] { range }; }
410 public Element build(Context cx, NonTerminalNode cnt, boolean dropall) { return toAtom(cx); }
411 public Atom toAtom(Context cx) {
412 edu.berkeley.sbp.util.Range.Set set = new edu.berkeley.sbp.util.Range.Set();
413 for(AtomNodeRange r : ranges) set.add(r.first, r.last);
414 return CharAtom.set(set);
417 public class AtomNodeRange {
420 public AtomNodeRange(char only) { first = only; last = only; }
421 public AtomNodeRange(char first, char last) { this.first = first; this.last = last; }
424 public class RepeatNode extends ElementNode {
425 public ElementNode e, sep;
426 public boolean zero, many, max;
427 public RepeatNode(ElementNode e, ElementNode sep, boolean zero, boolean many, boolean max) {
428 this.e = e; this.sep = sep; this.zero = zero; this.many = many; this.max = max;}
429 public Atom toAtom(Context cx) { return sep==null ? e.toAtom(cx) : super.toAtom(cx); }
430 public Element build(Context cx, NonTerminalNode cnt, boolean dropall) {
431 if (!dropall && !drop() && !e.drop())
432 throw new Error("you need a tag on this repetition: " + build(cx, cnt, dropall, ""));
433 return build(cx, cnt, dropall, illegalTag);
435 public Element build(Context cx, NonTerminalNode cnt, boolean dropall, Object repeatTag) {
437 ? Repeat.repeat(e.build(cx, null, dropall), zero, many, sep==null ? null : sep.build(cx, null, dropall), repeatTag)
439 ? Repeat.repeatMaximal(e.toAtom(cx), zero, many, repeatTag)
440 : Repeat.repeatMaximal(e.build(cx, null, dropall), zero, many, sep.toAtom(cx), repeatTag);
444 public abstract class ElementNode {
445 public String getOwnerTag() { return null; }
446 public boolean drop() { return false; }
447 public Atom toAtom(Context cx) { throw new Error("can't convert a " + this.getClass().getName() + " to an atom: " + this); }
448 public abstract Element build(Context cx, NonTerminalNode cnt, boolean dropall);
451 public abstract class ElementNodeWrapper extends ElementNode {
452 protected ElementNode _e;
453 public ElementNodeWrapper(ElementNode e) { this._e = e; }
454 public String getOwnerTag() { return _e.getOwnerTag(); }
455 public boolean drop() { return _e.drop(); }
456 public Atom toAtom(Context cx) { return _e.toAtom(cx); }
457 public Element build(Context cx, NonTerminalNode cnt, boolean dropall) { return _e.build(cx, cnt, dropall); }
460 public class TildeNode extends ElementNodeWrapper {
461 public TildeNode(ElementNode e) { super(e); }
462 public Atom toAtom(Context cx) { return (Atom)((Topology<Character>)_e.toAtom(cx).complement()); }
463 public Element build(Context cx, NonTerminalNode cnt, boolean dropall) { return toAtom(cx); }
466 public class DropNode extends ElementNodeWrapper {
467 public DropNode(ElementNode e) { super(e); }
468 public boolean drop() { return true; }
471 public Seq and2(Seq s, Seq a) { a.alwaysDrop = true; return s.and(a); }
472 public Seq andnot2(Seq s, Seq a) { a.alwaysDrop = true; return s.andnot(a); }
473 public Seq arrow(Seq s, ElementNode e) { return s.follow(e); }
474 public Seq tag(String tagname, Seq s) { return s.tag(tagname); }
475 public Seq seq(ElementNode[] elements) { return new Seq(elements); }
476 public Seq seq2(ElementNode[] elements) { return new Seq(elements); }
477 public ElementNode plusmax(final ElementNode e) { return new RepeatNode(e, null, false, true, true); }
478 public ElementNode plus(final ElementNode e) { return new RepeatNode(e, null, false, true, false); }
479 public ElementNode plusmaxfollow(final ElementNode e, final ElementNode sep) { return new RepeatNode(e, sep, false, true, true); }
480 public ElementNode plusfollow(final ElementNode e, final ElementNode sep) { return new RepeatNode(e, sep, false, true, false); }
481 public ElementNode starmax(final ElementNode e) { return new RepeatNode(e, null, true, true, true); }
482 public ElementNode star(final ElementNode e) { return new RepeatNode(e, null, true, true, false); }
483 public ElementNode starmaxfollow(final ElementNode e, final ElementNode sep) { return new RepeatNode(e, sep, true, true, true); }
484 public ElementNode starfollow(final ElementNode e, final ElementNode sep) { return new RepeatNode(e, sep, true, true, false); }
485 public ElementNode question(final ElementNode e) { return new RepeatNode(e, null, true, true, false); }
487 //////////////////////////////////////////////////////////////////////////////
489 public class Context {
490 public HashMap<String,Union> map = new HashMap<String,Union>();
491 public GrammarNode grammar;
492 public Context(Tree t) { }
493 public Context(GrammarNode g) { this.grammar = g; }
494 public Union build() {
496 for(NonTerminalNode nt : grammar.values()) {
497 Union u = get(nt.name);
498 if ("s".equals(nt.name))
503 public Union peek(String name) { return map.get(name); }
504 public void put(String name, Union u) { map.put(name, u); }
505 public Union get(String name) {
506 Union ret = map.get(name);
507 if (ret != null) return ret;
508 NonTerminalNode nt = grammar.get(name);
510 throw new Error("warning could not find " + name);
512 ret = new Union(name, false);
514 nt.buildIntoPreallocatedUnion(this, nt, nt.drop(), ret);
521 public class XTree extends ElementNode {
523 public XTree(Seq seq) { this.body = seq; }
524 public Element build(Context cx, NonTerminalNode cnt, boolean dropall) {
525 Union u = new Union(null, false);
526 Sequence s = body.build(cx, u, null, dropall);
527 Union u2 = new Union(null, false);
528 u2.add(Sequence.create(new Element[] {