您可以根据异常的类型捕获异常。引自MDC:

try {
   myroutine(); // may throw three exceptions
} catch (e if e instanceof TypeError) {
   // statements to handle TypeError exceptions
} catch (e if e instanceof RangeError) {
   // statements to handle RangeError exceptions
} catch (e if e instanceof EvalError) {
   // statements to handle EvalError exceptions
} catch (e) {
   // statements to handle any unspecified exceptions
   logMyErrors(e); // pass exception object to error handler
}

注意:条件捕获子句是Netscape(因此也是Mozilla/Firefox)扩展，它不是ECMAScript规范的一部分，因此不能依赖于特定的浏览器。

确保在遍历对象属性时使用hasOwnProperty方法:

for (p in anObject) {
    if (anObject.hasOwnProperty(p)) {
        //Do stuff with p here
    }
}

这样做是为了只访问object的直接属性，而不使用原型链下面的属性。

这个是超级隐藏的，只是偶尔有用;-)

可以使用原型链创建委托给另一个对象的对象，而无需更改原始对象。

var o1 = { foo: 1, bar: 'abc' };
function f() {}
f.prototype = o1;
o2 = new f();
assert( o2.foo === 1 );
assert( o2.bar === 'abc' );
o2.foo = 2;
o2.baz = true;
assert( o2.foo === 2 );
// o1 is unchanged by assignment to o2
assert( o1.foo === 1 );
assert( o2.baz );

这只包括o1上的“简单”值。如果你修改了一个数组或另一个对象，那么原型就不再“保护”原始对象。当你在Class定义/原型中有{}或[]时要小心。

在Crockford的《Javascript: The Good Parts》中也提到过:

parseInt()是危险的。如果你给它一个字符串而不告诉它正确的进制，它可能会返回意想不到的数字。例如，parseInt('010')返回8，而不是10。传递一个base给parseInt可以使它正确工作:

parseInt('010') // returns 8! (in FF3)
parseInt('010', 10); // returns 10 because we've informed it which base to work with.

具有公共接口的私有变量

它使用了一个简洁的小技巧，即自调用函数定义。 返回的对象中的所有内容在公共接口中都是可用的，而其他所有内容都是私有的。

var test = function () {
    //private members
    var x = 1;
    var y = function () {
        return x * 2;
    };
    //public interface
    return {
        setx : function (newx) {
            x = newx;
        },
        gety : function () {
            return y();
        }
    }
}();

assert(undefined == test.x);
assert(undefined == test.y);
assert(2 == test.gety());
test.setx(5);
assert(10 == test.gety());

下面是一些有趣的事情:

Comparing NaN with anything (even NaN) is always false, that includes ==, < and >. NaN Stands for Not a Number but if you ask for the type it actually returns a number. Array.sort can take a comparator function and is called by a quicksort-like driver (depends on implementation). Regular expression "constants" can maintain state, like the last thing they matched. Some versions of JavaScript allow you to access $0, $1, $2 members on a regex. null is unlike anything else. It is neither an object, a boolean, a number, a string, nor undefined. It's a bit like an "alternate" undefined. (Note: typeof null == "object") In the outermost context, this yields the otherwise unnameable [Global] object. Declaring a variable with var, instead of just relying on automatic declaration of the variable gives the runtime a real chance of optimizing access to that variable The with construct will destroy such optimzations Variable names can contain Unicode characters. JavaScript regular expressions are not actually regular. They are based on Perl's regexs, and it is possible to construct expressions with lookaheads that take a very, very long time to evaluate. Blocks can be labeled and used as the targets of break. Loops can be labeled and used as the target of continue. Arrays are not sparse. Setting the 1000th element of an otherwise empty array should fill it with undefined. (depends on implementation) if (new Boolean(false)) {...} will execute the {...} block Javascript's regular expression engine's are implementation specific: e.g. it is possible to write "non-portable" regular expressions.

[更新了一点，以回应良好的评论;请参阅评论]