Post migration to JDK 8: Functions, lambdas and others

[guidomedina]

Post migration to JDK 8:

• Functions, lambdas and use of computeIfAbsent to behave like a true cache.
• Migrate date formatters and calendars to Java 8 JSR-310.

There is a new class org.quickfixj.SimpleCache, is it at the right place?

[guidomedina]

This is an interesting function, I think we don't need a calendar for that, if it could be migrated to JDK 8 to avoid such instantiation.

Also, JDK 8 JSR-310 date & time formatters are now thread safe, I'll see if I can add more of that stuff on Friday, busy days ahead on my other project.

Maybe migrated to JDK 8 will make it faster than calculating the hash for a String look in a Map

[chrjohn]

Hi @guidomedina , thank you, I will look at it later this week or weekend.

[chrjohn]

Just curious: Shouldn't we then have a code path that directly uses computeIfAbsent if number of CPUs is greater than 32 (for the very rare case ;)).

[guidomedina]

Maybe is not worth the effort, look at 1st comment on the question:
http://stackoverflow.com/questions/33901070/java-8-concurrent-hash-map-get-performance-alternative

This is actually the recommendation by the Doug Lea the author of ConcurrentHashMap for Java 8 (aka CHMv8):
http://jsr166-concurrency.10961.n7.nabble.com/Re-ConcurrentHashMap-computeIfAbsent-td11687.html

[guidomedina]

Rebased and added some plugin updates, I was looking at the date time stuff and it will take some effort to do so I think what has been done is good enough for now, later we can comeback to it and migrate more stuff to Java 8.

[chrjohn]

Sorry, did not have time yet to take a deeper look, but anyway thanks for the PR. 👍

Just a quick question: does the conversion from for-each-loop to the functional style have any benefit? I do not find it easier to understand. I mean, maybe there are cases where it might be handy, but in this commit it is just the same logic but with a different header.

[guidomedina]

I didn't change any for-each, I explicitly left these out, the only functional changes are anonymous classes and functional interfaces which are interfaces with one method that are inferred as functional by Java 8 hence allow you to omit boiler plate code.

Almost two years ago I found difficult to understand Java 8, then it becomes very easy.

[guidomedina]

I have a custom QFJ 1.7.x compiled version, running what is on master in production for weeks now, no issues, I also introduced a NullStore to avoid storing market data messages which are not needed to be resend, I should probably PR it, WDYT?

import org.slf4j.LoggerFactory;
import quickfix.*;

import java.util.Collection;
import java.util.Date;

public class NullStore implements MessageStore {

  public static final class NullStoreFactory implements MessageStoreFactory {

    @Override
    public MessageStore create(SessionID sessionID) {
      return new NullStore(sessionID);
    }
  }

  private final SessionID sessionID;

  private Date creationTime = new Date();
  private int nextSenderMsgSeqNum = 1;
  private int nextTargetMsgSeqNum = 1;

  public NullStore(SessionID sessionID) {
    this.sessionID = sessionID;
  }

  public void get(int startSequence, int endSequence, Collection<String> messages) {
  }

  public Date getCreationTime() {
    return creationTime;
  }

  public int getNextSenderMsgSeqNum() {
    return nextSenderMsgSeqNum;
  }

  public int getNextTargetMsgSeqNum() {
    return nextTargetMsgSeqNum;
  }

  public void incrNextSenderMsgSeqNum() {
    nextSenderMsgSeqNum++;
  }

  public void incrNextTargetMsgSeqNum() {
    nextTargetMsgSeqNum++;
  }

  public void reset() {
    creationTime = new Date();
    nextSenderMsgSeqNum = 1;
    nextTargetMsgSeqNum = 1;
  }

  public boolean set(int sequence, String message) {
    return true;
  }

  public void setNextSenderMsgSeqNum(int next) {
    nextSenderMsgSeqNum = next;
  }

  public void setNextTargetMsgSeqNum(int next) {
    nextTargetMsgSeqNum = next;
  }

  public void refresh() {
    final String text = "null store does not support refresh!";
    if (sessionID != null) {
      final Session session = Session.lookupSession(sessionID);
      session.getLog().onErrorEvent("ERROR: " + text);
    } else {
      LoggerFactory.getLogger(quickfix.MemoryStore.class).error(text);
    }
  }
}

[chrjohn]

Hi @guidomedina,

yes, please open a separate PR. You could also use the PersistMessages=N option to not persist any messages.
But I like your implementation anyway since it is even doing less processing than a MemoryStore used in conjunction with PersistMessages=N.

Cheers,
Chris.

[chrjohn]

Hi @guidomedina , could you please reopen this? I didn't yet merge the JDK8 changes.
Thanks

[chrjohn]

OK, just saw #88 :)

[guidomedina]

I split them so that one doesn't interfere with the other.