测试数据 10 万条:
Single==1 == 5 Multi==1 == 21 Single==2 == 2 Multi==2 == 5 Single==3 == 2 Multi==3 == 2 Single==4 == 2 Multi==4 == 3 Single==5 == 2 Multi==5 == 3 Single==6 == 2 Multi==6 == 4 Single==7 == 17 Multi==7 == 2 Single==8 == 2 Multi==8 == 2 Single==9 == 2 Multi==9 == 4 Single==10 == 2 Multi==10 == 2 Single==11 == 2 Multi==11 == 3 Single==12 == 2 Multi==12 == 5 Single==13 == 2 Multi==13 == 1 Single==14 == 2 Multi==14 == 2 Single==15 == 1 Multi==15 == 2 Single==16 == 2 Multi==16 == 2 Single==17 == 2 Multi==17 == 1 Single==18 == 2 Multi==18 == 2 Single==19 == 1 Multi==19 == 2 Single==20 == 1 Multi==20 == 2
测试数据 100 万条:
Single==1 == 16 Multi==1 == 20 Single==2 == 10 Multi==2 == 13 Single==3 == 11 Multi==3 == 14 Single==4 == 12 Multi==4 == 12 Single==5 == 11 Multi==5 == 12 Single==6 == 11 Multi==6 == 13 Single==7 == 11 Multi==7 == 12 Single==8 == 12 Multi==8 == 12 Single==9 == 11 Multi==9 == 12 Single==10 == 11 Multi==10 == 12 Single==11 == 11 Multi==11 == 12 Single==12 == 11 Multi==12 == 15 Single==13 == 11 Multi==13 == 12 Single==14 == 11 Multi==14 == 12 Single==15 == 11 Multi==15 == 14 Single==16 == 11 Multi==16 == 13 Single==17 == 11 Multi==17 == 16 Single==18 == 12 Multi==18 == 14 Single==19 == 12 Multi==19 == 14 Single==20 == 14 Multi==20 == 12
测试数据 1000 万条:
Single==1 == 94 Multi==1 == 138 Single==2 == 105 Multi==2 == 167 Single==3 == 124 Multi==3 == 155 Single==4 == 126 Multi==4 == 145 Single==5 == 116 Multi==5 == 145 Single==6 == 119 Multi==6 == 148 Single==7 == 117 Multi==7 == 147 Single==8 == 118 Multi==8 == 146 Single==9 == 117 Multi==9 == 146 Single==10 == 117 Multi==10 == 146 Single==11 == 117 Multi==11 == 148 Single==12 == 117 Multi==12 == 146 Single==13 == 117 Multi==13 == 146 Single==14 == 117 Multi==14 == 146 Single==15 == 117 Multi==15 == 159 Single==16 == 117 Multi==16 == 147 Single==17 == 116 Multi==17 == 146 Single==18 == 116 Multi==18 == 146 Single==19 == 120 Multi==19 == 145 Single==20 == 116 Multi==20 == 147
测试数据 2000 万条
Single==1 == 191 Multi==1 == 288 Single==2 == 224 Multi==2 == 303 Single==3 == 253 Multi==3 == 298 Single==4 == 240 Multi==4 == 301 Single==5 == 236 Multi==5 == 298 Single==6 == 234 Multi==6 == 299 Single==7 == 234 Multi==7 == 298 Single==8 == 235 Multi==8 == 300 Single==9 == 233 Multi==9 == 297 Single==10 == 236 Multi==10 == 297 Single==11 == 237 Multi==11 == 298 Single==12 == 236 Multi==12 == 301 Single==13 == 236 Multi==13 == 297 Single==14 == 237 Multi==14 == 302 Single==15 == 236 Multi==15 == 300 Single==16 == 240 Multi==16 == 310 Single==17 == 234 Multi==17 == 298 Single==18 == 235 Multi==18 == 296 Single==19 == 232 Multi==19 == 301 Single==20 == 234 Multi==20 == 298
测试代码如下,可自行测试:
import lombok.Data;
import org.junit.Test;
import org.springframework.util.StopWatch;
import java.util.List;
import java.util.stream.Collectors;
import java.util.stream.Stream;
public class StreamTest {
@Data
public static class TestDTO {
private Integer id;
private String na;
}
@Test
public void test(){
StopWatch watch = new StopWatch();
int total = 20000000;
TestDTO[] arr = new TestDTO[total];
int i = 0;
while (i++ < total) {
TestDTO source = new TestDTO();
source.setId(i);
source.setNa("a" + i);
arr[i-1] = source;
}
List list = Stream.of(arr).collect(Collectors.toList());
int k = 0;
while (k++ < 20) {
watch.start("Single==" + k);
List collect2 = list.stream().filter(
t -> t.getId() % 3 == 0
&& t.getId() % 2 == 0
&& t.getId() % 5 == 0
&& t.getNa().endsWith("70")).collect(Collectors.toList());
watch.stop();
watch.start("Multi==" + k);
List collect = list.stream()
.filter(t -> t.getId() % 3 == 0)
.filter(t -> t.getId() % 2 == 0)
.filter(t -> t.getId() % 5 == 0)
.filter(t -> t.getNa().endsWith("70"))
.collect(Collectors.toList());
watch.stop();
}
Stream.of(watch.getTaskInfo()).forEach(
ti -> System.out.println(ti.getTaskName() + " == " + ti.getTimeMillis()));
}
}
原因如下:假定数据为1000万条,单个filter只需要扫描1000万次,多个filter扫描1000万+第一个filter的结果数+第二个filter的结果数+...,依此类推。
基于些理论,代码改动如下:
@Test
public void test(){
StopWatch watch = new StopWatch();
int total = 10000000;
TestDTO[] arr = new TestDTO[total];
int i = 0;
while (i++ < total) {
TestDTO source = new TestDTO();
source.setId(i);
source.setNa("a" + i);
arr[i-1] = source;
}
List list = Stream.of(arr).collect(Collectors.toList());
int k = 0;
while (k++ < 20) {
watch.start("Single==" + k);
List collect2 = list.stream().filter(
t -> t.getId() % 1 == 0
&& t.getId() % 1 == 0
&& t.getId() % 2 == 0
&& t.getId() % 5 == 0
&& t.getNa().endsWith("70")).collect(Collectors.toList());
watch.stop();
watch.start("Multi==" + k);
List collect = list.stream()
.filter(t -> t.getId() % 1 == 0)
.filter(t -> t.getId() % 1 == 0)
.filter(t -> t.getId() % 2 == 0)
.filter(t -> t.getId() % 5 == 0)
.filter(t -> t.getNa().endsWith("70"))
.collect(Collectors.toList());
watch.stop();
}
Stream.of(watch.getTaskInfo()).forEach(
ti -> System.out.println(ti.getTaskName() + " == " + ti.getTimeMillis()));
}
测试结果:
Single==1 == 94 Multi==1 == 222 Single==2 == 110 Multi==2 == 234 Single==3 == 128 Multi==3 == 226 Single==4 == 116 Multi==4 == 236 Single==5 == 120 Multi==5 == 229 Single==6 == 120 Multi==6 == 237 Single==7 == 124 Multi==7 == 235 Single==8 == 118 Multi==8 == 236 Single==9 == 117 Multi==9 == 234 Single==10 == 116 Multi==10 == 228 Single==11 == 123 Multi==11 == 235 Single==12 == 117 Multi==12 == 227 Single==13 == 117 Multi==13 == 224 Single==14 == 115 Multi==14 == 232 Single==15 == 116 Multi==15 == 229 Single==16 == 116 Multi==16 == 232 Single==17 == 122 Multi==17 == 245 Single==18 == 121 Multi==18 == 235 Single==19 == 117 Multi==19 == 226 Single==20 == 116 Multi==20 == 231总结:单个filter比多个filter性能更好。



