Switch to Closure Compiler for generating the minified version.

.gitmodules

@@ -4,6 +4,3 @@
 [submodule "vendor/marked"]
 	path = vendor/marked
 	url = git://github.com/chjj/marked.git
-[submodule "vendor/uglifyjs"]
-	path = vendor/uglifyjs
-	url = git://github.com/mishoo/UglifyJS.git

build.js

@@ -1,113 +1,228 @@
 #!/usr/bin/env node
 
 /* JSON 3 Builder | http://bestiejs.github.com/json3 */
-var path = require("path"), fs = require("fs"), marked = require(path.join(__dirname, "vendor", "marked")), ugly = require(path.join(__dirname, "vendor", "uglifyjs", "uglify-js"));
+var path = require("path"), fs = require("fs"), spawn = require("child_process").spawn, marked = require(path.join(__dirname, "vendor", "marked")),
+
+// The path to the Closure Compiler `.jar` file.
+closurePath = path.join(__dirname, "vendor", "closure-compiler.jar"),
+
+// The Closure Compiler options: enable advanced optimizations and suppress all
+// warnings apart from syntax and optimization errors.
+closureOptions = ["--compilation_level=advanced_optimizations", "--warning_level=quiet"];
 
 // Enable GitHub-Flavored Markdown.
 marked.setOptions({ "gfm": true });
 
 // Generate the GitHub project page.
-fs.readFile(path.join(__dirname, "README.md"), "utf8", function readSource(exception, source) {
+fs.readFile(path.join(__dirname, "README.md"), "utf8", function readInfo(exception, source) {
   if (exception) {
     console.log(exception);
   } else {
     // Read the project page template.
-    fs.readFile(path.join(__dirname, "page", "page.html"), "utf8", function readPage(exception, page) {
-      var headers, lines, lastSection, lastLevel, navigation;
-      if (exception) {
-        console.log(exception);
-      } else {
-        // Generate the page navigation. Ported from `mdtoc.rb` by Sam
-        // Stephenson.
-        headers = [];
-        lines = source.split(/\r?\n/);
-        // First pass: Scan the Markdown source looking for titles of the format:
-        // `### Title ###`. Record the line number, header level (number of
-        // octothorpes), and text of each matching title.
-        lines.forEach(function (line, index) {
-          var match = /^(\#{1,6})\s+(.+?)\s+\1$/.exec(line);
-          if (match) {
-            headers.push([index, match[1].length, match[2]]);
-          }
-        });
-        // Second pass: Iterate over all matched titles and compute their
-        // corresponding section numbers. Then replace the titles with annotated
-        // anchors.
-        headers.forEach(function(value) {
-          var index = value[0], level = value[1], text = value[2], section, length;
-          if (lastSection) {
-            // Duplicate the last section metadata array.
-            section = lastSection.slice(0);
-            if (lastLevel < level) {
-              section.push(1);
-            } else {
-              length = lastLevel - level;
-              while (length--) {
-                section.pop();
-              }
-              section[section.length - 1] += 1;
-            }
+    fs.readFile(path.join(__dirname, "page", "page.html"), "utf8", readTemplate);
+  }
+
+  // Interpolates the page template and writes the result to disk.
+  function readTemplate(exception, page) {
+    var headers, lines, lastSection, lastLevel, navigation;
+    if (exception) {
+      console.log(exception);
+    } else {
+      // Generate the page navigation. Ported from `mdtoc.rb` by Sam
+      // Stephenson.
+      headers = [];
+      lines = source.split(/\r?\n/);
+      // First pass: Scan the Markdown source looking for titles of the format:
+      // `### Title ###`. Record the line number, header level (number of
+      // octothorpes), and text of each matching title.
+      lines.forEach(function (line, index) {
+        var match = /^(\#{1,6})\s+(.+?)\s+\1$/.exec(line);
+        if (match) {
+          headers.push([index, match[1].length, match[2]]);
+        }
+      });
+      // Second pass: Iterate over all matched titles and compute their
+      // corresponding section numbers. Then replace the titles with annotated
+      // anchors.
+      headers.forEach(function (value) {
+        var index = value[0], level = value[1], text = value[2], section, length;
+        if (lastSection) {
+          // Clone the last section metadata array.
+          section = lastSection.slice(0);
+          if (lastLevel < level) {
+            section.push(1);
           } else {
-            section = [1];
-          }
-          lines[index] = Array(level + 1).join("#") + "<a name=\"section_" + section.join(".") + "\"></a>" + text;
-          value.push(section);
-          lastSection = section;
-          lastLevel = level;
-        });
-        // Third pass: Iterate over matched titles once more to produce the table of
-        // contents.
-        navigation = headers.map(function (value) {
-          var index = value[0], level = value[1], text = value[2], section = value[3], name = section.join(".");
-          return "<li><a href=\"#section_" + name + "\">" + text + "</a></li>";
-        });
-        navigation.push("");
-        // Write the page source to disk.
-        fs.writeFile(path.join(__dirname, "index.html"), page.replace(/<%=\s*(.+?)\s*%>/g, function (match, data) {
-          switch (data) {
-            case "navigation":
-              // Insert the table of contents directly into the template.
-              return navigation.join("\n");
-            case "source":
-              // Convert the read me to HTML and insert it into the page body.
-              return marked(lines.join("\n"));
+            length = lastLevel - level;
+            while (length--) {
+              section.pop();
+            }
+            section[section.length - 1] += 1;
           }
-          return "";
-        }), function (exception) {
-          console.log(exception || "GitHub project page generated successfully.");
-        });
-      }
-    });
+        } else {
+          section = [1];
+        }
+        lines[index] = Array(level + 1).join("#") + "<a name=\"section_" + section.join(".") + "\"></a>" + text;
+        value.push(section);
+        lastSection = section;
+        lastLevel = level;
+      });
+      // Third pass: Iterate over matched titles once more to produce the table of
+      // contents.
+      navigation = headers.map(function (value) {
+        var index = value[0], level = value[1], text = value[2], section = value[3], name = section.join(".");
+        return "<li><a href=\"#section_" + name + "\">" + text + "</a></li>";
+      });
+      navigation.push("");
+      // Write the page source to disk.
+      fs.writeFile(path.join(__dirname, "index.html"), page.replace(/<%=\s*(.+?)\s*%>/g, function interpolate(match, data) {
+        switch (data) {
+          case "navigation":
+            // Insert the table of contents directly into the template.
+            return navigation.join("\n");
+          case "source":
+            // Convert the read me to HTML and insert it into the page body.
+            return marked(lines.join("\n"));
+        }
+        return "";
+      }), function writePage(exception) {
+        console.log(exception || "GitHub project page generated successfully.");
+      });
+    }
   }
 });
 
-// Compress JSON 3 using UglifyJS.
-fs.readFile(path.join(__dirname, "lib", "json3.js"), "utf8", function (exception, source) {
+// Compress JSON 3 using the Closure Compiler.
+fs.readFile(path.join(__dirname, "lib", "json3.js"), "utf8", function readSource(exception, source) {
   var results;
   if (exception) {
     console.log(exception);
   } else {
-    results = "";
-    // Preserve the copyright header.
-    ugly.parser.tokenizer(source)().comments_before.forEach(function (comment) {
-      // Remove the leading `!` character from YUI-style comments.
-      results += comment.type == "comment1" ? "//" + comment.value + "\n" : ("/*" + comment.value.slice(comment.value.charAt(0) == "!" ? 1 : 0) + "*/");
-    });
-    results += "\n;" + ugly.uglify.gen_code(
-      // Enable unsafe transformations.
-      ugly.uglify.ast_squeeze_more(
-        ugly.uglify.ast_squeeze(
-          // Munge variable and function names, excluding the special `define`
-          // function exposed by asynchronous module loaders.
-          ugly.uglify.ast_mangle(ugly.parser.parse(source), {
-            "except": ["define"]
-          }
-      ))), {
-      "ascii_only": true
-    }) + ";";
-    // Split lines at 500 characters for consistency with Closure Compiler.
-    fs.writeFile(path.join(__dirname, "lib", "json3.min.js"), ugly.uglify.split_lines(results, 500), function (exception) {
+    // Shell out to the Closure Compiler executable. Requires Java 6 or higher.
+    invoke("java", ["-jar", closurePath].concat(closureOptions), source, compressSource);
+  }
+
+  // Post-processes the compressed source and writes the result to disk.
+  function compressSource(exception, compressed) {
+    if (exception) {
+      console.log(exception);
+    } else {
+      // Extract the JSON 3 header and wrap the compressed source in an
+      // immediately-invoked function expression (enabling advanced
+      // optimizations causes the Compiler to add global variables).
+      compressed = extractComments(source)[0] + "\n;(function(){" + compressed + "}());";
+      // Write the compressed version to disk.
+      fs.writeFile(path.join(__dirname, "lib", "json3.min.js"), compressed, writeSource);
+    }
+
+    // Checks the `gzip`-ped size of the compressed version by shelling out to the
+    // Unix `gzip` executable.
+    function writeSource(exception) {
       console.log(exception || "Compressed version generated successfully.");
-    });
+      // Automatically check the `gzip`-ped size of the compressed version.
+      // This requires shelling out to the Unix `gzip` executable.
+      invoke("gzip", ["-9f", "-c"], compressed, "binary", function checkSize(exception, results) {
+        if (!exception) {
+          console.log("Compressed version size: %d KB.", results.length);
+        }
+      });
+    }
+  }
+});
+
+// Internal: Invokes a process with the given `name`, `parameters,` and standard
+// `input`. Yields the result to a `callback` function. The optional `encoding`
+// argument specifies the output stream encoding.
+function invoke(name, parameters, input, encoding, callback) {
+  // The standard output stream, error stream, and process instance.
+  var error = "", output = "", process = spawn(name, parameters);
+  if (typeof encoding == "string") {
+    // Explicitly set the encoding of the output stream if one is specified.
+    process.stdout.setEncoding(encoding);
+  } else {
+    callback = encoding;
+    encoding = null;
+  }
+  process.stdout.on("data", function onData(data) {
+    // Append the data to the output stream.
+    output += data;
+  });
+  process.stderr.on("data", function onError(data) {
+    // Append the error message to the error stream.
+    error += data;
+  });
+  process.on("exit", function onExit(status) {
+    var exception;
+    // `status` specifies the process exit code.
+    if (status) {
+      exception = new Error(error);
+      exception.status = status;
+    }
+    callback(exception, output);
+  });
+  // Proxy the standard input to the process.
+  process.stdin.end(input);
+}
+
+// Internal: Extracts line and block comments from a JavaScript `source`
+// string. Returns an array containing the comments.
+function extractComments(source) {
+  var index = 0, length = source.length, results = [], symbol, position, original;
+  while (index < length) {
+    symbol = source[index];
+    switch (symbol) {
+      // Parse line and block comments.
+      case "/":
+        original = symbol;
+        symbol = source[++index];
+        switch (symbol) {
+          // Extract line comments.
+          case "/":
+            position = source.indexOf("\n", index);
+            if (position < 0) {
+              // Check for CR line endings.
+              position = source.indexOf("\r", index);
+            }
+            results.push(original + source.slice(index, index = position < 0 ? length : position));
+            break;
+          // Extract block comments.
+          case "*":
+            position = source.indexOf("*/", index);
+            if (position < 0) {
+              throw SyntaxError("Unterminated block comment.");
+            }
+            // Advance past the end of the comment.
+            results.push(original + source.slice(index, index = position += 2));
+            break;
+          default:
+            index++;
+        }
+        break;
+      // Parse strings separately to ensure that any JavaScript comments within
+      // them are preserved.
+      case '"':
+      case "'":
+        for (position = index, original = symbol; index < length;) {
+          symbol = source[++index];
+          if (symbol == "\\") {
+            // Skip past escaped characters.
+            index++;
+          } else if ("\n\r\u2028\u2029".indexOf(symbol) > -1) {
+            // According to the ES 5.1 spec, strings may not contain unescaped
+            // line terminators.
+            throw SyntaxError("Illegal line continuation.");
+          } else if (symbol == original) {
+            break;
+          }
+        }
+        if (source[index] == original) {
+          index++;
+          break;
+        }
+        throw SyntaxError("Unterminated string.");
+      default:
+        // Advance to the next character.
+        index++;
+    }
   }
-});
+  return results;
+}