直接更改谷歌地图标记图标？(Change google maps marker icon directly?)

 这里有三个问题要考虑：如果用户希望某些输入无效，该怎么办？ 如果无效输入是无意的，该怎么办？ 以及该语言中常见的模式是什么？ 
 
 当预期无效输入时 
 
 用户可能希望某些输入无效。 如果函数返回NaN ，则这些情况很容易处理。 
 
let values = [[2, 4, 6], [3, 5, 1], [7, 5, 9]];

let clamped = values.map(args => clamp(...args)).filter(n => !isNaN(n));

console.log(clamped); // [4, 7]
 
 如果该函数抛出异常，则用户必须使用try...catch语句 ，这将更加冗长且性能更低。 或者她可以在应用clamp()之前过滤掉无效输入，但是她必须确切地知道哪种输入是无效的，这可能比在这个例子中更复杂。 
 
 这显然是支持NaN选项的一个观点。 
 
 无效输入无意时 
 
 如果函数抛出异常，则可以更轻松地调试部分用户的正版错误。 例如，用户可能很容易犯这样的错误： 
 
clamp(-4, 0, -10);
 
 如果以上是返回NaN ，调试可能是噩梦，特别是在一个包含很多其他事情的大型代码库中。 如果函数抛出异常，用户会看到： 
 
RangeError: min may not be greater than max
  at clamp ...
 
 更简单。 这显然是一场胜利。 
 
 共同的语言模式 
 
 所以我们有一个平局。 我们最好的办法是问自己，WWJD（JavaScript会做什么）？ 例如，当我们将负数传递给Math.sqrt()时会发生什么？ 
 
console.log(Math.sqrt(-4)); // NaN
 
 它返回NaN 。 这是打破平局。 如果输入无效，JavaScript程序员会期望这些函数返回NaN 。 它们将用于处理源自此行为的错误，如果不是，那么他们最好习惯它。 
 
 所以回归NaN 。 

There are three questions to consider here: What if the user expects some inputs to be invalid? What if the invalid inputs are unintentional? and What patterns are commonly found in the language?
 
When invalid inputs are expected
 
The user might expect certain inputs to be invalid. If the function returns NaN, these cases are easy to deal with.
 
let values = [[2, 4, 6], [3, 5, 1], [7, 5, 9]];

let clamped = values.map(args => clamp(...args)).filter(n => !isNaN(n));

console.log(clamped); // [4, 7]
 
If the function were to throw an exception, the user would have to use a try...catch statement, which would be more verbose and less performant. Or she could filter out the invalid inputs before applying clamp(), but then she would have to know exactly which kinds of inputs are invalid, which could be more complicated than it is in this example.
 
This is clearly a point in favor of the NaN option.
 
When invalid inputs are unintended
 
Genuine errors on part of the user will be more easily debugged if the function throws an exception. For example, the user could easily make a mistake like this:
 
clamp(-4, 0, -10);
 
If the above were to return NaN, debugging could be nightmarish, especially in a large code base with lots of other things going on. If the function were to throw an exception, the user would see this:
 
RangeError: min may not be greater than max
  at clamp ...
 
Much simpler. This is clearly a win for throw.
 
Common language patterns
 
So we have a tie. Our best bet is to ask ourselves, WWJD (What would JavaScript do)? What happens, for example, when we pass a negative number to Math.sqrt()?
 
console.log(Math.sqrt(-4)); // NaN
 
It returns NaN. This is the tie-breaker. JavaScript programmers will expect functions like these to return NaN if the inputs are invalid. And they will be used to dealing with errors that originate from this behavior, or if not, then they had better get used to it.
 
So return NaN.