Look at the usage scenario. If it is purely computational, use the for loop directly. If there are other IO operations, consider the parallel stream, which mainly depends on the cost of computing and thread synchronization

In fact, lambda still runs in a thread, so how to implement it?

Think of threads as a task pool. To create a task each time is to insert a task into the task pool. The program continues to move down until all the current code has been executed, and then obtains the next task execution from the thread.

Lambda plays this little trick, which makes people feel asynchronous. In fact, it is a fragmented operation of tasks.

The performance difference between a loop and a lambda depends on the task scheduling speed, but I think that in most cases, the for loop is more efficient.

The main advantage of lambda is that it is not easy to block.

According to the test results of Alex Zhitnitsky, sticking to the traditional Java programming style - iterator and for each loop - has better performance than the implementation of Java 8.

Using iterators or for each loops is the most effective way to traverse ArrayList, and its performance is twice as good as the traditional for loop with index values.

Among the methods of Java 8, the performance of parallel streams is the best. Be careful, however, that in some cases it may cause the program to run slower. Lambda expressions have a speed between stream and parallel stream.

No, after stream is enabled for parallel, the amount of large data will increase

No, the simpler the syntax, the slower the execution. This also explains why python is not as efficient as java

There are parallel streams in lambda, which can be multithreaded. However, for will still be faster if only one million pieces of data are traversed. When the data volume is small, multithreading may not be faster than single thread. However, when the data volume is large, tens of millions or even hundreds of millions, the parallel stream may be faster

Lambda and stream are two different things. The problem should be the comparison between stream and for loops. Normal serial stream is 1.4-1.5 times slower than for; The parallel stream loop will be faster and only suitable for large amount of data calculation. Otherwise, the thread overhead and synchronization operation will suffer performance loss. The running speed is about 0.6-0.8 times of the for loop

In many cases, lambda will be fast, and most compilers will turn lambda into inline function. If the for loop is not in the inline function, the code will not be cached. If the for loop is in the override function, it will be slower. Therefore, lambda is a good alternative to polymorphism (virtual pointer).

It depends on the situation.

One kind of flow in the lambda expression is called parallelSteam, which is called parallel flow. Compared with stream flow, the processing of convection is parallel, so it is certainly more efficient than the simple for loop. However, parallel streams have a disadvantage that they are prone to concurrent security problems. So this thing should be used with caution.

Of course, stream stream also has its own advantages. If you have multiple processing operations, such as filtering the elements of specific conditions in a collection, transforming them, and then traversing the output, multiple operations are involved at this time. If you use the traditional for loop, it is not as efficient as stream. Because it encapsulates the algorithms and methods for operating the set, it is processed from the bottom. If it is just a traversal operation, the efficiency of the two is not very different.

More importantly, using stream stream can make the code more concise. Although it is not very friendly to old programmers, you will like this writing method if you spend some time researching.