// Idea: For every position, store the productions recognized to start there, then infer up to higher classes !752 // a recognition is identified by (startPos, className, endPos) // key 1 = start position, key 2 = class name, value = end position static Map> recog; static MultiMap> productionMap = new MultiMap; static boolean debug = false; p { S rulesText = loadSnippet("#1002281"); S inputText = loadSnippet("#1002286") + "\n" + loadSnippet("#1002280"); S mainProd = "line"; for (S rule : toLinesFullTrim(rulesText)) pcall { //printF("Processing rule: *", rule); L lr = splitAtJavaToken(rule, "="); if (l(lr) != 2) { print("Weird rule: " + rule); continue; } S l = lr.get(0), r = lr.get(1); L tokr = javaTok(r); assertEquals(structure(tokr), 3, l(tokr)); S className = assertIdentifier(get(tokr, 1)); L tok = javaTok(l); tok = mergeBracketThingies(tok); //printStructure(tok); productionMap.put(className, tok); } print(n(productionMap.size(), "production") + "."); print(); for (S line : toLinesFullTrim(inputText)) { print(); print(line); L tok = javaTok(line); //printStructure(tok); Pos pos = new Pos(tok); L x = parseTop(pos, mainProd); if (x.contains(l(tok))) print(" parsed"); else if (!empty(x)) print(" beginning matches"); else print(" not parsed"); print(" " + structure(recog)); if (!empty(x)) { L out = explainMatch(new Pos(tok), x.get(0), mainProd); print("Analysis: " + structure(out)); } } } static L parseTop(Pos pos, S mainProd) { // init structures recog = new TreeMap; for (int i = pos.i; i < l(pos.tok); i += 2) recog.put(i, new MultiMap); boolean anyChange; do { anyChange = false; for (int i = pos.i; i < l(pos.tok); i += 2) { Pos pos2 = new Pos(pos.tok, i); for (S className : productionMap.keySet()) { MultiMap rr = recog.get(i); L recs = rr.getActual(className); L> prods = productionMap.get(className); for (L prod : prods) { int n = l(recs); matchProd(pos2, new Pos(prod), className, recs); anyChange = anyChange || l(recs) > n; } rr.clean(className); } } } while (anyChange); ret recog.get(pos.i).get(mainProd); } static class Pos { L tok; int i = 1; *() {} *(L *tok) {} *(L *tok, int *i) {} boolean end() { ret i >= l(tok)-1; } S get() { ret tok.get(i); } public Pos clone() { ret new Pos(tok, i); } public boolean equals(O o) { if (!(o instanceof Pos)) ret false; Pos pos = cast o; ret tok == pos.tok && i == pos.i; } S rest() { ret join(subList(tok, i)); } Pos plus(int x) { ret new Pos(tok, i + x); } } static void copy(Pos a, Pos b) { b.tok = a.tok; b.i = a.i; } static void debug(S bla, Pos pos) { if (debug) print(bla + " on " + quote(pos.rest())); } static void matchProd(Pos pos, Pos prod, S forClass, L out) { if (prod.end()) setAdd(out, pos.i); else { S p = prod.get(); if (isBracketedID(p) && neq(p, "")) { L r = recog.get(pos.i).get(unbracket(p)); // keep parsing for every option for (int i : cloneList(r)) matchProd(new Pos(pos.tok, i), prod.plus(2), forClass, out); } else { // it's a literal if (pos.end()) ret; // need a token to match S t = pos.get(); if (!matchToken(p, t)) ret; matchProd(pos.plus(2), prod.plus(2), forClass, out); } } } static boolean matchToken(S p, S t) { if (eq(p, "")) { if (!isQuoted(t)) ret false; } else if (!(eq(p, "*") || eqic(p, t))) ret false; // token mismatch ret true; } // assumes that there is a match (pos, class, endPos) // and gives explanations of how it was probably matched static L explainMatch(Pos pos, int endPos, S forClass) { L> prods = productionMap.get(forClass); new L out; for (L prod : prods) explainMatch2(pos, new Pos(prod), endPos, forClass, litlist(forClass, join(prod)), out); ret out; } // same, but with fixed production static void explainMatch2(Pos pos, Pos prod, int endPos, S forClass, L match, L out) { if (prod.end()) { if (pos.i == endPos) out.add(cloneList(match)); } else { S p = prod.get(); if (isBracketedID(p) && neq(p, "")) { S className = unbracket(p); L r = recog.get(pos.i).get(className); // keep parsing for every option for (int i : cloneList(r)) { match.add(litlist(pos.i, i)); explainMatch2(new Pos(pos.tok, i), prod.plus(2), endPos, forClass, match, out); removeLast(match); } } else { // it's a literal if (pos.end()) ret; // need a token to match S t = pos.get(); if (!matchToken(p, t)) ret; //match.add(litlist(pos.i, p, pos.i+2)); explainMatch2(pos.plus(2), prod.plus(2), endPos, forClass, match, out); removeLast(match); } } }