Java字符串拆分为数组，是否应该存储在变量中？

Question

With our project we are going for MAXIMUM possible "speed". Would it be faster to split the string multiple times or to store it in a variable and use it?

Example

    String example = "1,2,3,4";
    System.out.println(example.split(",")[0]);
    System.out.println(example.split(",")[1]);
    System.out.println(example.split(",")[3]);
    System.out.println(example.split(",")[4]);

or

    String[] example = "1,2,3,4".split(",");
    System.out.println(example[0]);
    System.out.println(example[1]);
    System.out.println(example[2]);
    System.out.println(example[3]);

Which one would execute in less time?

Answer 1

In general, working with arrays is pretty fast and parsing Strings is not, so parsing the String only once and then working with the array will be significantly faster.

Anyhow, this is easily measurable, just create a for with a few thousand elements and try both options to see the difference.

Answer 2

Following is faster

String[] example = "1,2,3,4".split(",");

Proof:

Time taken to run first method 10000 times: 640 ms
Time taken to run second method 10000 times: 531 ms

Time taken to run first method 100000 times: 4144 ms
Time taken to run second method 100000 times: 3763 ms

Answer 3

Putting Reece's answer to the test, I used this code

    long startTime, endTime;

    int amountOfStrings=10000;

    System.err.println("Generating "+amountOfStrings+" String(s)");
    startTime=System.currentTimeMillis();
    String example="";
    for(int i=1; i<amountOfStrings-1; i++) {
        example+=i+",";
    }
    example+=amountOfStrings;
    endTime=System.currentTimeMillis();
    System.err.println("Done in "+(endTime-startTime)+"ms");

    System.err.println("Taking the multisplit-approach");
    startTime=System.currentTimeMillis();
    for(int i=0; i<amountOfStrings-1; i++) {
        System.out.println(example.split(",")[i]);
    }
    endTime=System.currentTimeMillis();
    System.err.println("Done in "+(endTime-startTime)+"ms");

    System.err.println("Taking the onesplit-approach");
    startTime=System.currentTimeMillis();
    String[] data=example.split(",");
    for(int i=0; i<amountOfStrings-1; i++) {
        System.out.println(data[i]);
    }
    endTime=System.currentTimeMillis();
    System.err.println("Done in "+(endTime-startTime)+"ms");

Executed this command:

$ java -jar ~/Desktop/Untitled.jar > ~/Desktop/Untitled.log

And got this output:

Generating 10000 String(s)
Done in 292ms
Taking the multisplit-approach
Done in 3906ms
Taking the onesplit-approach
Done in 275ms

This clearly indicates that the onesplit-approach ( String[] data=example.split(",") ) is much faster than the multisplit-approach ( example.split(",")[i] ), using 10,000 Strings.

Sidenote: Using 100 Strings is the multisplit-approach 24ms and the onesplit-approach 6ms, 10 String multisplit 1ms and onesplit 0ms. I don't expect the multisplit-approach to be faster when using more numbers.

Happy coding :) -Charlie