因为到处复制大对象会浪费时间和内存。

指针是获得对另一个变量的间接引用的一种方式。它们不是保存变量的值，而是告诉您变量的地址。这在处理数组时特别有用，因为使用指向数组中第一个元素(它的地址)的指针，可以通过递增指针(指向下一个地址位置)快速找到下一个元素。

我读过的关于指针和指针算术的最好解释是在K & R的《C编程语言》中。开始学习c++的一本好书是c++ Primer。

这里有一个略有不同，但有深刻见解的观点，为什么C的许多特性是有意义的:http://steve.yegge.googlepages.com/tour-de-babel#C

基本上，标准的CPU体系结构是Von Neumann体系结构，在这样的机器上，能够引用内存中数据项的位置并对其进行运算是非常有用的。如果您了解汇编语言的任何变体，您将很快看到这在低级别上是多么重要。

c++让指针有点令人困惑，因为它有时会为你管理指针，并以“引用”的形式隐藏它们的效果。如果你直接使用C语言，对指针的需求就更加明显了:没有其他方法可以实现引用调用，它是存储字符串的最佳方式，是迭代数组的最佳方式，等等。

关于你的第二个问题，通常你在编程时不需要使用指针，但是有一个例外，那就是当你创建一个公共API时。

The problem with C++ constructs that people generally use to replace pointers are very dependent on the toolset that you use which is fine when you have all the control you need over the source code, however if you compile a static library with visual studio 2008 for instance and try to use it in a visual studio 2010 you will get a ton of linker errors because the new project is linked with a newer version of STL which is not backwards compatible. Things get even nastier if you compile a DLL and give an import library that people use in a different toolset because in that case your program will crash sooner or later for no apparent reason.

因此，为了将大型数据集从一个库移动到另一个库，如果你不想强迫其他人使用与你使用的相同工具，你可以考虑给一个指向应该复制数据的函数的数组指针。这样做的好处是它甚至不必是一个c风格的数组，你可以使用std::vector并通过给出第一个元素&向量[0]的地址来给出指针，并使用std::vector在内部管理数组。

Another good reason to use pointers in C++ again relates to libraries, consider having a dll that cannot be loaded when your program runs, so if you use an import library then the dependency isn't satisfied and the program crashes. This is the case for instance when you give a public api in a dll alongside your application and you want to access it from other applications. In this case in order to use the API you need to load the dll from its' location (usually it's in a registry key) and then you need to use a function pointer to be able to call functions inside the DLL. Sometimes the people that make the API are nice enough to give you a .h file that contain helper functions to automate this process and give you all the function pointers that you need, but if not you can use LoadLibrary and GetProcAddress on windows and dlopen and dlsym on unix to get them (considering that you know the entire signature of the function).

使用指针而不是变量的一种方法是消除所需的重复内存。例如，如果您有某个大型复杂对象，您可以使用指针为您所做的每次引用指向该变量。对于变量，您需要为每个副本复制内存。

Pointers allow you to refer to the same space in memory from multiple locations. This means that you can update memory in one location and the change can be seen from another location in your program. You will also save space by being able to share components in your data structures. You should use pointers any place where you need to obtain and pass around the address to a specific spot in memory. You can also use pointers to navigate arrays: An array is a block of contiguous memory that has been allocated with a specific type. The name of the array contains the value of the starting spot of the array. When you add 1, that takes you to the second spot. This allows you to write loops that increment a pointer that slides down the array without having an explicit counter for use in accessing the array.

下面是一个C语言的例子:

char hello[] = "hello";

char *p = hello;

while (*p)
{
    *p += 1; // increase the character by one

    p += 1; // move to the next spot
}

printf(hello);

打印

ifmmp

因为它取每个字符的值并加1。