OrientDB Community 2.0.5 - Seems bug: two times hook event triggering

[Андрей Гришин]

Hi all!

I'm using OrientDB 2.0.5 release now and encountered one problem with my hook.

I have recursively linked object structure, which my hook need to process, and i've wrote simple demo showing problem below:

    abstract class AbstractClass {

    @OVersion

    private Object version;

    private AbstractClass parent;

   

@SuppressWarnings("unused")

public AbstractClass getParent() {

return parent;

}

public void setParent(AbstractClass parent) {

this.parent = parent;

}

    }

    @Entity

    class MyClass extends AbstractClass {

   

private String name;

public MyClass() {

super();

name = "no-name";

}

@SuppressWarnings("unused")

public String getName() {

return name;

}

public void setName(String name) {

this.name = name;

}

}

    

    @Entity

    class ContainerClass extends AbstractClass {

    private AbstractClass item;

public AbstractClass getItem() {

return item;

}

public void setItem(AbstractClass item) {

this.item = item;

}

    }

    

    public void linkRecursiveObjectsHookUpdateTest() {

    OObjectDatabaseTx db = new OObjectDatabaseTx("plocal:./LocalTestDB");

    ODocumentHookAbstract hook = new ODocumentHookAbstract() {

    @Override

    public RESULT onRecordBeforeCreate(ODocument iDocument) {

    System.out.println("This CREATE string appears one time (correct).");

    return RESULT.RECORD_NOT_CHANGED;

    }

    @Override

    public RESULT onRecordBeforeUpdate(ODocument iDocument) {

    System.out.println("This UPDATE string appears two times! (incorrect?) " + Arrays.asList(iDocument.getDirtyFields()));

    // Here i have some 'pacth' code, which fix this problem for me, but i'm not sure, if it is correct way to fix

    // OObjectDatabaseTx db;

    // if (iDocument.getDatabase().getDatabaseOwner() instanceof OObjectDatabaseTx) {

    // db = (OObjectDatabaseTx) iDocument.getDatabase().getDatabaseOwner();

    // } else {

    // db = new OObjectDatabaseTx((ODatabaseDocumentTx) iDocument.getDatabase().getDatabaseOwner());

    // }

    // db.unsetDirty(db.getUserObjectByRecord(iDocument, null));

    return RESULT.RECORD_NOT_CHANGED;

    }

@Override

public DISTRIBUTED_EXECUTION_MODE getDistributedExecutionMode() {

return DISTRIBUTED_EXECUTION_MODE.BOTH;

}

};

hook.setIncludeClasses("MyClass");

db.registerHook(hook, HOOK_POSITION.FIRST); // can be any position

if (!db.exists()) {

db.create();

} else {

db.open("admin", "admin");

}

db.setAutomaticSchemaGeneration(true);

db.getEntityManager().registerEntityClass(MyClass.class);

db.getEntityManager().registerEntityClass(ContainerClass.class);

db.reload(); // Sometimes entity registration transparently fails and first save command fires exception (first try block), seems that reload helps (another bug?)

MyClass obj = new MyClass();

obj.setName("name");

ContainerClass box = new ContainerClass();

obj.setParent(box);

box.setItem(obj);

db.begin();

try {

box = db.save(box);

db.commit();

} catch (Exception e) {

db.rollback();

db.close();

throw e;

}

obj = (MyClass) box.getItem();

db.begin();

try {

obj.setName("new-name");

obj = db.save(obj); // I have first even triggering here

db.commit(); // and the second here

} catch (Exception e) {

db.rollback();

throw e;

} finally {

db.close();

}

}

This code gives me next output:

mar 18, 2015 12:08:42 PM com.orientechnologies.common.log.OLogManager log

INFO: OrientDB auto-config DISKCACHE=1 018MB (heap=876MB os=3 942MB disk=166 585MB)

This CREATE string appears one time (correct).

This UPDATE string appears two times! (incorrect?) [name]

This UPDATE string appears two times! (incorrect?) []

I must notice, that problem occurs only when i register my hook from client-side and it not exists on server-side registered hook.

For me is important to understand, is this problem is bug or not? And if not, then why and how it can be evaded for me?

Thanks,

Andrey