最近我读了很多关于函数式编程的东西，大部分我都能理解，但有一件事我就是搞不懂，那就是无状态编码。在我看来，通过删除可变状态来简化编程就像通过删除仪表盘来“简化”一辆汽车:最终产品可能更简单，但希望它能与最终用户交互。
几乎我能想到的每个用户应用程序都将状态作为核心概念。如果你写了一个文档(或一个SO post)，状态会随着每一个新的输入而改变。或者如果你玩电子游戏，会有大量的状态变量，从所有角色的位置开始，这些角色往往会不断移动。如果不跟踪不断变化的值，您怎么可能做任何有用的事情呢?
每次我发现一些讨论这个问题的东西，它都是用真正的技术函数语言写的，假设我没有浓厚的FP背景。有谁知道如何向那些对命令式编码有很好的、扎实的理解，但在函数方面完全是n00b的人解释这一点吗?
编辑:到目前为止，一堆回复似乎试图让我相信不可变值的优点。我懂你的意思。这很有道理。我不明白的是，在没有可变变量的情况下，如何跟踪必须不断变化的值。
当前回答
Functional programming avoids state and emphasizes functionality. There's never any such thing as no state, though the state might actually be something that's immutable or baked into the architecture of what you're working with. Consider the difference between a static web server that just loads up files off the filesystem versus a program that implements a Rubik's cube. The former is going to be implemented in terms of functions designed to turn a request into a file path request into a response from the contents of that file. Virtually no state is needed beyond a tiny bit of configuration (the filesystem 'state' is really outside the scope of the program. The program works the same way regardless of what state the files are in). In the latter though, you need to model the cube and your program implementation of how operations on that cube change its state.
其他回答
Functional programming avoids state and emphasizes functionality. There's never any such thing as no state, though the state might actually be something that's immutable or baked into the architecture of what you're working with. Consider the difference between a static web server that just loads up files off the filesystem versus a program that implements a Rubik's cube. The former is going to be implemented in terms of functions designed to turn a request into a file path request into a response from the contents of that file. Virtually no state is needed beyond a tiny bit of configuration (the filesystem 'state' is really outside the scope of the program. The program works the same way regardless of what state the files are in). In the latter though, you need to model the cube and your program implementation of how operations on that cube change its state.
这很简单。在函数式编程中，你可以使用任意多的变量……但前提是它们是局部变量(包含在函数中)。因此，只需将代码包装在函数中，在这些函数之间来回传递值(作为传递的参数和返回值)……这就是它的全部!
这里有一个例子:
function ReadDataFromKeyboard() {
$input_values = $_POST[];
return $input_values;
}
function ProcessInformation($input_values) {
if ($input_values['a'] > 10)
return ($input_values['a'] + $input_values['b'] + 3);
else if ($input_values['a'] > 5)
return ($input_values['b'] * 3);
else
return ($input_values['b'] - $input_values['a'] - 7);
}
function DisplayToPage($data) {
print "Based your input, the answer is: ";
print $data;
print "\n";
}
/* begin: */
DisplayToPage (
ProcessInformation (
GetDataFromKeyboard()
)
);
JavaScript provides very clear examples of the different ways of approaching mutable or immutable state\values within its core because the ECMAScript specifications were not able to settle on a universal standard so one must continue to memorize or doublecheck which functions create a new object that they return or modify the original object passed to it. If your entire language is immutable then you know you are always getting a new (copied & possibly modified) result and never have to worry about accidentally modifying the variable before passing it into a function.
你知道哪个会返回一个新的对象，哪个会改变下面例子中的原始对象吗?
Array.prototype.push()
String.prototype.slice()
Array.prototype.splice()
String.prototype.trim()
除了别人给出的很好的答案，想想Java中的Integer和String类。这些类的实例是不可变的，但这并不意味着仅仅因为它们的实例不可更改，这些类就毫无用处。不可变性给了你一定的安全性。您知道，如果使用String或Integer实例作为Map的键，则该键不能更改。将其与Java中的Date类进行比较:
Date date = new Date();
mymap.put(date, date.toString());
// Some time later:
date.setTime(new Date().getTime());
你已经无声地改变了地图中的一个键!使用不可变对象(如在函数式编程中)要干净得多。更容易推断会发生什么副作用——没有!这意味着程序员更容易，优化器也更容易。
这就是没有COMMON块的FORTRAN的工作方式:您将编写具有传入值和局部变量的方法。就是这样。
面向对象编程将我们的状态和行为结合在一起，但当我在1994年第一次从c++中接触到它时，它还是一个新思想。
天啊，当我还是机械工程师的时候，我是一个函数式程序员，而我却不知道!
