下面这段话帮助我理解了它们之间的区别:

数据封装是一种捆绑数据的机制 使用它们的函数和数据抽象是一种机制 只暴露接口，隐藏实现细节 来自用户。

你可以在这里阅读更多。

让我们以堆栈为例。它可以使用数组或链表来实现。但它支持的操作是推送和弹出。

Now abstraction is exposing only the interfaces push and pop. The underlying representation is hidden (is it an array or a linked list?) and a well-defined interface is provided. Now how do you ensure that no accidental access is made to the abstracted data? That is where encapsulation comes in. For example, classes in C++ use the access specifiers which ensure that accidental access and modification is prevented. And also, by making the above-mentioned interfaces as public, it ensures that the only way to manipulate the stack is through the well-defined interface. In the process, it has coupled the data and the code that can manipulate it (let's not get the friend functions involved here). That is, the code and data are bonded together or tied or encapsulated.

封装要求模块化。它要求您创建具有数据和处理数据的方法的对象。在这种情况下，您可以将其视为一个模块。

抽象为您提供了类的一般视图。

简而言之:

抽象是一种帮助我们识别哪些特定信息是必要的，哪些信息应该隐藏的技术。

因此，封装是一种以隐藏对象的细节和实现细节的方式封装信息的技术。

这里的大多数答案都关注于OOP，但封装开始得更早:

Every function is an encapsulation; in pseudocode: point x = { 1, 4 } point y = { 23, 42 } numeric d = distance(x, y) Here, distance encapsulates the calculation of the (Euclidean) distance between two points in a plane: it hides implementation details. This is encapsulation, pure and simple. Abstraction is the process of generalisation: taking a concrete implementation and making it applicable to different, albeit somewhat related, types of data. The classical example of abstraction is C’s qsort function to sort data: The thing about qsort is that it doesn't care about the data it sorts — in fact, it doesn’t know what data it sorts. Rather, its input type is a typeless pointer (void*) which is just C’s way of saying “I don't care about the type of data” (this is also called type erasure). The important point is that the implementation of qsort always stays the same, regardless of data type. The only thing that has to change is the compare function, which differs from data type to data type. qsort therefore expects the user to provide said compare function as a function argument.

封装和抽象是密切相关的，因此您可以认为它们确实是不可分割的。就实际而言，这可能是对的;也就是说，这里有一个不太抽象的封装:

class point {
    numeric x
    numeric y
}

我们封装了点的坐标，但是我们没有实质性地将它们抽象出来，只是在逻辑上对它们进行分组。

这里有一个抽象的例子，它不是封装:

T pi<T> = 3.1415926535

这是一个具有给定值(π)的泛型变量pi，声明并不关心变量的确切类型。诚然，我很难在真实的代码中找到这样的东西:抽象实际上总是使用封装。然而，上面的内容在c++(14)中确实存在，通过变量模板(=变量的通用模板);使用稍微复杂一点的语法，例如:

template <typename T> constexpr T pi = T{3.1415926535};

封装:对对象的实际用户隐藏不需要的/不期望的/适当的实现细节。 如。

List<string> list = new List<string>();
list.Sort(); /* Here, which sorting algorithm is used and hows its 
implemented is not useful to the user who wants to perform sort, that's 
why its hidden from the user of list. */

抽象:是一种提供泛化的方法，因此是处理大量不同对象的通用方法。如。

class Aeroplane : IFlyable, IFuelable, IMachine
{ // Aeroplane's Design says:
  // Aeroplane is a flying object
  // Aeroplane can be fueled
  // Aeroplane is a Machine
}
// But the code related to Pilot, or Driver of Aeroplane is not bothered 
// about Machine or Fuel. Hence,
// pilot code:
IFlyable flyingObj = new Aeroplane();
flyingObj.Fly();
// fighter Pilot related code
IFlyable flyingObj2 = new FighterAeroplane();
flyingObj2.Fly();
// UFO related code 
IFlyable ufoObj = new UFO();
ufoObj.Fly();
// **All the 3 Above codes are genaralized using IFlyable,
// Interface Abstraction**
// Fly related code knows how to fly, irrespective of the type of 
// flying object they are.

// Similarly, Fuel related code:
// Fueling an Aeroplane
IFuelable fuelableObj = new Aeroplane();
fuelableObj.FillFuel();
// Fueling a Car
IFuelable fuelableObj2 = new Car(); // class Car : IFuelable { }
fuelableObj2.FillFuel();

// ** Fueling code does not need know what kind of vehicle it is, so far 
// as it can Fill Fuel**