Core Java

Can/Should I use parallel streams in a transaction context?

Introduction

To make a long story short, you should not use transactions within a parallel stream. This is because each thread in the parallel stream has its own name thus it does participate in the transaction.

The Streams API is designed to work correctly under certain guidelines. In practice, to benefit from parallelism, each operation is not allowed to change the state of shared objects (such operations are called side-effect-free). Provided you follow this guideline, the internal implementation of parallel streams cleverly splits the data, assigns different parts to independent threads, and merges the final result.

This is primary originated because of the way Transactions are implemented. In  sort, a ThreadLocal variable is used to mark each method participating in the transaction. ThreadLocal variables are not able to keep their vale within a parallel stream. To demonstrate that I have created the following test

01
02
03
04
05
06
07
08
09
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
import org.junit.Assert;
import org.junit.Test;
 
import java.util.concurrent.atomic.AtomicBoolean;
import java.util.stream.IntStream;
 
public class ThreadNameTest {
 
    @Test
    public void threadLocalTest(){
        ThreadContext.set("MAIN");
        AtomicBoolean hasNamechanged = new AtomicBoolean(false);
        IntStream.range(0,10000000).boxed().parallel().forEach(n->{
            if(!"MAIN".equals(ThreadContext.get())){
                hasNamechanged.set(true);
            }
        });
        Assert.assertTrue(hasNamechanged.get());
    }
 
    private static class ThreadContext {
        private static ThreadLocal<String> val = ThreadLocal.withInitial(() -> "empty");
 
        public ThreadContext() {
        }
 
        public static String get() {
            return val.get();
        }
 
        public static void set(String x) {
            ThreadContext.val.set(x);
        }
    }
 
}

The higher the IntStream.range value the more certain the test will succeed.

Now take a look at this github project https://github.com/diakogiannis/transactionplayground/

The TransactionPlayground project

I created a services that loads cats in 4 different ways

  1. sequentially curl -I -X GET http://localhost:8080/api/cats/all
  2. sequentially but throwing an exception in order to create a rollback mark curl -I -X GET http://localhost:8080/api/cats/all-exception
  3. in parallel curl -I -X GET http://localhost:8080/api/cats/all-parallel
  4. in parallel but throwing an exception in order to create a rollback mark curl -I -X GET http://localhost:8080/api/cats/all-parallel-exception

There are also 2 helper calls

  1. cleanup curl -I -X DELETE http://localhost:8080/api/cats/
  2. and one to actually view the cats curl -X GET http://localhost:8080/api/cats/

Start the project

please execute mvn clean package wildfly-swarm:run

Normal Ordered

Call curl -I -X GET http://localhost:8080/api/cats/all and then curl -X GET http://localhost:8080/api/cats/

Normal Without Order aka Parallel

Call for cleanup curl -I -X DELETE http://localhost:8080/api/cats/ Call curl -I -X GET http://localhost:8080/api/cats/all-parallel and then curl -X GET http://localhost:8080/api/cats/

The expected result is to see a list of cats. Without ordering. This is why parallel stream is first come-first served and reads randomly from the list.

Normal With exception

Call for cleanup curl -I -X DELETE http://localhost:8080/api/cats/ Call curl -I -X GET http://localhost:8080/api/cats/all-exception and then curl -X GET http://localhost:8080/api/cats/

The expected result is an empty list. This is because the transaction was marked as rollback, so the jdbc transaction was rolledback thus all entries were not persisted to the database following the ACID model.

Parallel With exception

Call for cleanup curl -I -X DELETE http://localhost:8080/api/cats/ Call curl -I -X GET http://localhost:8080/api/cats/all-parallel-exception and then curl -X GET http://localhost:8080/api/cats/

The expected result is NOT an empty list. This is because each thread in the parallel stream opens its own jdbc transaction and commits when done. So each time you do this, you get some cats displayed up until the point you get an Exception and the execution stops. Rollback is made only in one thread.

Published on Java Code Geeks with permission by Alexius Diakogiannis, partner at our JCG program. See the original article here: Can/Should I use parallel streams in a transaction context?

Opinions expressed by Java Code Geeks contributors are their own.

Alexius Diakogiannis

Author of JEE.gr, JEE Architect, Scrum Master, Enthusiastic Entrepreneur, Passionate Archer, Linux Lover and JAVA Geek!
Subscribe
Notify of
guest

This site uses Akismet to reduce spam. Learn how your comment data is processed.

0 Comments
Oldest
Newest Most Voted
Inline Feedbacks
View all comments
Back to top button