[Bug] Java Api .command(query) result in a temporary RID

[matanshukry]

OrientDB Version: 3.0.21

Java Version: 1.8.0_211

OS: Windows 10

Expected behavior

The result of a create query will include a positive (real) RID

Actual behavior

The result of a create query includes a negative (temporary) RID

Steps to reproduce

// ODatabase session = ... // obtain a session to the database
OResultSet result = session.command("CREATE VERTEX User SET name=\"bla1\"");
// RID: result.next().getRecord().get().getIdentity()

[wolf4ood]

Hi @matanshukry

did you begin a transaction before the command?

session.begin();

Thanks

[matanshukry]

@wolf4ood Yup. Also doing .commit() after, but that's after I'm getting the recordid anyway.

[wolf4ood]

Hi @matanshukry

this works for me

@Test
  public void testTx() {

    OrientDB orientDB = new OrientDB("embedded:", OrientDBConfig.defaultConfig());

    orientDB.createIfNotExists("test", ODatabaseType.MEMORY);

    ODatabaseSession database = orientDB.open("test", "admin", "admin");

    database.begin();

    OVertex vertex;

    try (OResultSet resultSet = database.command("create vertex V set name = 'Test'")) {
      vertex = resultSet.next().getVertex().get();
    }

    database.commit();

    Assert.assertTrue(vertex.getIdentity().isPersistent());

  }

you should get the identity after the commit, since new records in tx have temporary rids

Let me know if this helps

Thanks

[matanshukry]

@wolf4ood I am getting the identity before the .commit(), which seems to be issue. Although I'm using breakpoints, and although I'm getting a temporary identity at that point, if I'll query the database using another session/console, I can see the new row with a persistent identity. Hence it exist, so why can't I access it?

Also, my current flow, which I believe is a common flow, is to convert the result of a db operation immediately when I have the result. Hence:
(1) session.begin()
(2) userA = findUserBy..(...)
(3) userB = createUser..(..)
(4) session.commit()

where each method such as findUserBy/createUser would go something like this:
(1) OResult result = session.command(query, args)
(2) User user = convertFromDb(result);

[matanshukry]

@wolf4ood Actually, it seems it doesn't work after the .commit() either. What I tried after your comment:

(1) session.begin()
(2) OResult result = session.command(query, args);
(3) this.user = convertResultToUser(result)
(4) this.savedResult = result
(5) session.commit()
(6) session.close()
(7) result.getRecord().map(ORecord::getIdentity).filter(ORID::isPersistent).orElse(null);

Line (7) is null. I can debug and see cluster position is still negative (-2).

[wolf4ood]

Hi @matanshukry

that's strange. Do you have a test case to reproduce this?

[wolf4ood]

Oh i was testing using embedded and works fine. By switching to remote it remains negative

Let me check

[matanshukry]

cool thx. I tried checking by saving OResultSet instead of just OResult, and even calling .next() only after the .commit(), all the same. Note it only happens when i'm calling .begin() and .commit()

[wolf4ood]

@matanshukry

yes because with begin and commit there is a server side transaction in progress. My understanding that at commit time there is no rebinding of final rids, which happens in embedded.

[matanshukry]

@wolf4ood Yeah I realized that too now. Just to clarify though, when you're saying "my understanding" - are you referring to how things work, or how things should work?

Also, I do notice ORecordAbstract.setIdentity() is called twice when using .begin and .commit().
Although the second time is called before the .commit(), in here: https://github.com/orientechnologies/orientdb/blob/develop/client/src/main/java/com/orientechnologies/orient/core/db/document/ODatabaseDocumentRemote.java#L375

According to OTransactionOptimisticClient.replaceContent(), it also looks like it's expecting a new non-temporary rid at that moment.

[wolf4ood]

Hi @matanshukry

my understanding from a debug session is that the rebind is not working properly.
After commit the OVertex should contain the final rid

[wolf4ood]

Hi @matanshukry

i was able to fix it. I'm running the test and if passes i will push it

Thanks