Spring非常适合将类实例粘合在一起。你知道你的Hibernate类总是需要一个数据源，Spring将它们连接在一起(也有一个数据源的实现)。

数据访问对象总是需要Hibernate访问，Spring将Hibernate类连接到dao中。

此外，Spring基本上为您提供了一系列库的可靠配置，并指导您应该使用哪些库。

Spring真的是一个很棒的工具。(我说的不是Spring MVC，只是基本框架)。

简而言之，我认为Spring是应用程序中的“粘合剂”。它用于集成不同的框架和您自己的代码。

Spring是Enterprise JavaBeans (EJB)技术的一个很好的替代方案。它还具有web框架和web服务框架组件。

我认为这有两部分:

"What exactly is Spring for" -> see the accepted answer by victor hugo. "[...] Spring is [a] good framework for web development" -> people saying this are talking about Spring MVC. Spring MVC is one of the many parts of Spring, and is a web framework making use of the general features of Spring, like dependency injection. It's a pretty generic framework in that it is very configurable: you can use different db layers (Hibernate, iBatis, plain JDBC), different view layers (JSP, Velocity, Freemarker...)

注意，不使用Spring MVC，你也可以在web应用程序中很好地使用Spring。我想说，大多数Java web应用程序都这样做，同时使用其他web框架，如Wicket, Struts, Seam，…

春天有三样东西。

Spring handles Dependency Injection and I recommend you read Martin Fowler's excellent introduction on dependency injection. The second thing Spring does is wrap excellent Java libraries in a very elegant way to use in your applications. For a good example see how Spring wraps Task Executors and Quartz Scheduler. Thirdly Spring provides a bunch of implementations of web stuff like REST, an MVC web framework and more. They figure since you are using Spring for the first two, maybe you can just use it for everything your web app needs.

The problem is that Spring DI is really well thought out, the wrappers around other things are really well thought out in that the other things thought everything out and Spring just nicely wraps it. The Spring implementations of MVC and REST and all the other stuff is not as well done (YMMV, IMHO) but there are exceptions (Spring Security is da bomb). So I tend to use Spring for DI, and its cool wrappers but prefer other stuff for Web (I like Tapestry a lot), REST (Jersey is really robust), etc.

其优点是依赖注入(DI)。这意味着将对象创建任务外包出去。让我用一个例子来解释。

public interface Lunch
{
   public void eat();
}

public class Buffet implements Lunch
{
   public void eat()
   {
      // Eat as much as you can 
   }
}

public class Plated implements Lunch
{
   public void eat()
   {
      // Eat a limited portion
   }
}

现在在我的代码中，我有一个类LunchDecide，如下所示:

public class LunchDecide {
    private Lunch todaysLunch;
    public LunchDecide(){
        this.todaysLunch = new Buffet(); // choose Buffet -> eat as much as you want
        //this.todaysLunch = new Plated(); // choose Plated -> eat a limited portion 
    }
}

在上面的类中，根据心情选择Buffet()或Plated()。但是这个系统是紧密耦合的。每次需要不同类型的对象时，我们都需要更改代码。在本例中，注释掉一行!假设有50个不同的人使用50个不同的类。那会是一团乱麻。在这种情况下，我们需要解耦系统。让我们重写LunchDecide类。

public class LunchDecide {
    private Lunch todaysLunch;
    public LunchDecide(Lunch todaysLunch){
        this.todaysLunch = todaysLunch
        }
    }

Notice that instead of creating an object using new keyword we passed the reference to an object of Lunch Type as a parameter to our constructor. Here, object creation is outsourced. This code can be wired either using Xml config file (legacy) or Java Annotations (modern). Either way, the decision on which Type of object would be created would be done there during runtime. An object would be injected by Xml into our code - Our Code is dependent on Xml for that job. Hence, Dependency Injection (DI). DI not only helps in making our system loosely coupled, it simplifies writing of Unit tests since it allows dependencies to be mocked. Last but not the least, DI streamlines Aspect Oriented Programming (AOP) which leads to further decoupling and increase of modularity. Also note that above DI is Constructor Injection. DI can be done by Setter Injection as well - same plain old setter method from encapsulation.