Implement initializers

lib/compiler.js

@@ -285,7 +285,7 @@ PEG.Compiler = {
 
         rule_ref:
           function(node) {
-            if (typeof(ast[node.name]) === "undefined") {
+            if (typeof(ast.rules[node.name]) === "undefined") {
               throw new PEG.GrammarError(
                 "Referenced rule \"" + node.name + "\" does not exist."
               );
@@ -299,14 +299,14 @@ PEG.Compiler = {
 
       function check(node) { checkFunctions[node.type](node); }
 
-      for (var rule in ast) {
-        check(ast[rule]);
+      for (var rule in ast.rules) {
+        check(ast.rules[rule]);
       }
     },
 
     /* Checks that the start rule is defined. */
     function(ast, startRule) {
-      if (typeof(ast[startRule]) === "undefined") {
+      if (typeof(ast.rules[startRule]) === "undefined") {
         throw new PEG.GrammarError(
           "Missing \"" + startRule + "\" rule."
         );
@@ -356,7 +356,7 @@ PEG.Compiler = {
                 "Left recursion detected for rule \"" + node.name + "\"."
               );
             }
-            check(ast[node.name], appliedRules);
+            check(ast.rules[node.name], appliedRules);
           },
 
         literal:       nop,
@@ -368,8 +368,8 @@ PEG.Compiler = {
         checkFunctions[node.type](node, appliedRules);
       }
 
-      for (var rule in ast) {
-        check(ast[rule], []);
+      for (var rule in ast.rules) {
+        check(ast.rules[rule], []);
       }
     }
   ],
@@ -432,18 +432,18 @@ PEG.Compiler = {
           replaceFunctions[node.type](node, from, to);
         }
 
-        for (var rule in ast) {
-          replace(ast[rule], from, to);
+        for (var rule in ast.rules) {
+          replace(ast.rules[rule], from, to);
         }
       }
 
-      for (var rule in ast) {
-        if (isProxyRule(ast[rule])) {
-          replaceRuleRefs(ast, ast[rule].name, ast[rule].expression.name);
+      for (var rule in ast.rules) {
+        if (isProxyRule(ast.rules[rule])) {
+          replaceRuleRefs(ast, ast.rules[rule].name, ast.rules[rule].expression.name);
           if (rule === startRule) {
-            startRule = ast[rule].expression.name;
+            startRule = ast.rules[rule].expression.name;
           }
-          delete ast[rule];
+          delete ast.rules[rule];
         }
       }
 
@@ -452,6 +452,10 @@ PEG.Compiler = {
   ],
 
   _compileFunctions: {
+    initializer: function(node) {
+      return node.code;
+    },
+
     rule: function(node) {
       var resultVar = PEG.Compiler.generateUniqueIdentifier("result");
 
@@ -871,9 +875,13 @@ PEG.Compiler = {
       startRule = newAstNadStartRule[1];
     }
 
+    var initializerCode = ast.initializer !== null
+      ?  this.compileNode(ast.initializer)
+      : "";
+
     var parseFunctionDefinitions = [];
-    for (var rule in ast) {
-      parseFunctionDefinitions.push(this.compileNode(ast[rule]));
+    for (var rule in ast.rules) {
+      parseFunctionDefinitions.push(this.compileNode(ast.rules[rule]));
     }
 
     var source = this.formatCode(
@@ -998,6 +1006,8 @@ PEG.Compiler = {
       "        return { line: line, column: column };",
       "      }",
       "      ",
+      "      ${initializerCode}",
+      "      ",
       "      var result = parse_${startRule}({ reportMatchFailures: true });",
       "      ",
       "      /*",
@@ -1054,6 +1064,7 @@ PEG.Compiler = {
       "  return result;",
       "})()",
       {
+        initializerCode:          initializerCode,
         parseFunctionDefinitions: parseFunctionDefinitions.join("\n\n"),
         startRule:                startRule
       }

lib/metagrammar.pegjs

@@ -1,8 +1,20 @@
 grammar
-  = __ rules:rule+ {
-      var result = {};
-      PEG.ArrayUtils.each(rules, function(rule) { result[rule.name] = rule; });
-      return result;
+  = __ initializer:initializer? rules:rule+ {
+      var rulesConverted = {};
+      PEG.ArrayUtils.each(rules, function(rule) { rulesConverted[rule.name] = rule; });
+
+      return {
+        initializer: initializer !== "" ? initializer : null,
+        rules:       rulesConverted
+      }
+    }
+
+initializer
+  = code:action semicolon? {
+      return {
+        type: "initializer",
+        code: code
+      };
     }
 
 rule

test/compiler-test.js

@@ -379,6 +379,18 @@ test("actions", function() {
   doesNotParse(notAMatchParser, "b");
 });
 
+test("initializer", function() {
+  var variableDefinitionParser = PEG.buildParser(
+    '{ a = 42; }; start = "a" { return a; }'
+  );
+  parses(variableDefinitionParser, "a", 42);
+
+  var functionDefinitionparser = PEG.buildParser(
+    '{ function f() { return 42; } }; start = "a" { return f(); }'
+  );
+  parses(variableDefinitionParser, "a", 42);
+});
+
 test("rule references", function() {
   var parser = PEG.buildParser([
     'start   = static / dynamic',

test/metagrammar-test.js

@@ -20,6 +20,13 @@ global.grammarParserDoesNotParseWithMessage = function(input, message) {
 
 module("Grammar Parser");
 
+function initializer(code) {
+  return {
+    type: "initializer",
+    code: code
+  };
+}
+
 function rule(name, displayName, expression) {
   return {
     type:        "rule",
@@ -119,7 +126,10 @@ var sequenceLiterals = sequence([literalAbcd, literalEfgh, literalIjkl]);
 var choiceLiterals = choice([literalAbcd, literalEfgh, literalIjkl]);
 
 function oneRuleGrammar(expression) {
-  return { start: rule("start", null, expression) };
+  return {
+    initializer: null,
+    rules:       { start: rule("start", null, expression) }
+  };
 }
 
 var simpleGrammar = oneRuleGrammar(literal("abcd"));
@@ -144,19 +154,42 @@ function actionGrammar(action) {
   return oneRuleGrammar(action_(literal("a"), action));
 }
 
+var initializerGrammar = {
+  initializer: initializer(" code "),
+  rules: {
+    a: rule("a", null, literalAbcd),
+  }
+};
+
 /* Canonical grammar is "a: \"abcd\"; b: \"efgh\"; c: \"ijkl\";". */
 test("parses grammar", function() {
-  grammarParserParses('a = "abcd"', { a: rule("a", null, literalAbcd) });
+  grammarParserParses(
+    'a = "abcd"',
+    {
+      initializer: null,
+      rules:       { a: rule("a", null, literalAbcd) }
+    }
+  );
+  grammarParserParses('{ code }; a = "abcd"', initializerGrammar);
   grammarParserParses(
     'a = "abcd"; b = "efgh"; c = "ijkl"',
     {
-      a: rule("a", null, literalAbcd),
-      b: rule("b", null, literalEfgh),
-      c: rule("c", null, literalIjkl)
+      initializer: null,
+      rules: {
+        a: rule("a", null, literalAbcd),
+        b: rule("b", null, literalEfgh),
+        c: rule("c", null, literalIjkl)
+      }
     }
   );
 });
 
+/* Canonical initializer is "{ code }". */
+test("parses initializer", function() {
+  grammarParserParses('{ code }a = "abcd"', initializerGrammar);
+  grammarParserParses('{ code };a = "abcd"', initializerGrammar);
+});
+
 /* Canonical rule is "a: \"abcd\"". */
 test("parses rule", function() {
   grammarParserParses(
@@ -166,7 +199,8 @@ test("parses rule", function() {
   grammarParserParses(
     'start "start rule" = "abcd" / "efgh" / "ijkl"',
     {
-      start: rule("start", "start rule", choiceLiterals)
+      initializer: null,
+      rules:       { start: rule("start", "start rule", choiceLiterals) }
     }
   );
   grammarParserParses(