列表<列表<字符串>&GT; vs. String [] []

时间:2017-01-11 13:57:16

标签: java performance math matrix processing-efficiency

我想用Java创建一个2D String'matrix'对象。我的两个目标是提高效率和简化代码。

我听说使用ArrayList比String [] []效率更高。首先,我想知道这是否属实,如果是,那么效率会更高?

另一件事是我必须能够在'矩阵'中添加列和行。我能够开发一些有效的算法来向String [] []添加行和列。我很想知道以这种方式开发算法来操作2D List是否值得 - 是否会有显着的性能提升?

感谢您的帮助!

2 个答案:

答案 0 :(得分:2)

首先简单回答:直接数组操作会更快,因为您可以避免通用类的开销。通过利用Unsafe实例(非公共API)进行直接阵列访问,您甚至可以获得更高的性能。

对于意见部分:如果性能如此重要,我宁愿看看如何同时执行或分配不同系统。在这种情况下,我宁愿避免直接使用数组,因为这很难维护。但当然这一切都取决于您的具体用例。

因此,与所有与性能相关的问题一样,针对您的特定用例进行一些测量,并自行决定哪种方法最适合您。

就像我在评论中所说的那样,我已经摆弄了自己的Matrix实现,它有两个专门的Map实现(CompactArrayMapDirectArrayMap),它们都是关键映射的int。键自然顺序,非常像数组但具有地图功能。

答案 1 :(得分:1)

List是一个界面。有了这个界面,您可以使用不同的实现,例如ArrayListLinkedListCopyOnWriteArrayList

所以你的问题也许应该改写为

ArrayList<ArrayList<String>> vs String[][]

ArrayList是使用数组实现的列表。它有一些方法可以让你处理特定位置的元素:

void add(int index, E element);
E    get(int index);

我曾经认为ArrayList几乎和数组一样快,但我想 实际使用情况将决定实际的性能差异。下面我补充说 一些轶事实验的一些结果显示数组是 比ArrayList更快,特别是在2-D矩阵的人口中。 这两者的访问时间似乎彼此相同。

ArrayList为您提供了一些优势,例如您无需了解 大小提前。然而,如果我确定我需要的是一个数组而不是一个通用列表(例如,正如你所说,对于矩阵计算),那么我可以使用一个数组来获得更简洁的语法。

String s1 = array[i][j];
array[i][j] = s1;

对战

String s2 = list.get(j).get(i);
list.get(j).add(i, s2);

有关区分接口与其实现的更多信息,可以参考Oracle / Sun的本教程:

https://docs.oracle.com/javase/tutorial/collections/implementations/list.html

轶事实验

import java.util.ArrayList;
import java.util.List;
import java.util.Random;

public class ArrayListVsString {

    private static final int NUM_TESTS = 5;

    public static void main(String[] args) {
        List<List<String>>  list;
        String[][]          array; 
        int height  = 500;
        int width   = 1000;
        String __ = "    "; // indent
        for (int n=0; n<NUM_TESTS; n++) {
            System.out.format("Testing 2D matrix of %dx%d: Test %d%n"
                    , height, width, n);
            /*
             * Time to populate the matrices
             */
            long startTime = System.nanoTime();
            // array
            String subTestArray = "2-D array";
            array = new String[width][height];
            for (int i=0; i<width; i++) {
                for (int j=0; j<height; j++) {
                    array[i][j] = getRandomString();
                }
            }
            startTime 
                = logElapsedTime(startTime
                                , __ + "creating matrix as " 
                                     + subTestArray);
            // array-list
            String subTestList = "2-D array-list";
            list = new ArrayList<>(height);
            for (int j=0; j<height; j++) {
                List<String> row = new ArrayList<>(width);
                list.add(j, row);
                for (int i=0; i<width; i++) {
                    String element = getRandomString();
                    list.get(j).add(i, element);
                }
            }
            startTime 
                = logElapsedTime(startTime
                            , __ + "creating matrix as " 
                                 + subTestList);

            int hash = 0;
            /*
             * Time to do a full walk through all the elements
             */
            // array
            for (int i=0; i<width; i++) {
                for (int j=0; j<height; j++) {
                    hash += (array[i][j]).hashCode();
                }
            }
            startTime 
                = logElapsedTime(startTime
                                , __ + "full walk of matrix as" 
                                     + subTestArray);

            // array-list
            for (int j=0; j<height; j++) {
                for (int i=0; i<width; i++) {
                    hash += list.get(j).get(i).hashCode();
                }
            }
            startTime 
                = logElapsedTime(startTime
                                , __ + "full walk of matrix as "
                                     + subTestList);
            list = null;

        }

    }

    private static Random random = new Random();

    private static String getRandomString() {
        return String.valueOf(random.nextInt());
    }

    private static long logElapsedTime(long startTimeNano
            , String action) {
        long elapsedTimeNano = System.nanoTime() - startTimeNano;
        System.out.format("%s took %,d ms%n"
                    , action, elapsedTimeNano/1000000);

        return System.nanoTime();
    }
}

<强>结果

Testing 2D matrix of 500x1000: Test 0
    creating matrix as 2-D array took 117 ms
    creating matrix as 2-D array-list took 232 ms
    full walk of matrix as2-D array took 25 ms
    full walk of matrix as 2-D array-list took 31 ms
Testing 2D matrix of 500x1000: Test 1
    creating matrix as 2-D array took 65 ms
    creating matrix as 2-D array-list took 186 ms
    full walk of matrix as2-D array took 20 ms
    full walk of matrix as 2-D array-list took 30 ms
Testing 2D matrix of 500x1000: Test 2
    creating matrix as 2-D array took 61 ms
    creating matrix as 2-D array-list took 60 ms
    full walk of matrix as2-D array took 14 ms
    full walk of matrix as 2-D array-list took 15 ms
Testing 2D matrix of 500x1000: Test 3
    creating matrix as 2-D array took 66 ms
    creating matrix as 2-D array-list took 358 ms
    full walk of matrix as2-D array took 16 ms
    full walk of matrix as 2-D array-list took 15 ms
Testing 2D matrix of 500x1000: Test 4
    creating matrix as 2-D array took 45 ms
    creating matrix as 2-D array-list took 55 ms
    full walk of matrix as2-D array took 14 ms
    full walk of matrix as 2-D array-list took 15 ms