cancel
Showing results for 
Search instead for 
Did you mean: 

Head's Up! These forums are read-only. All users and content have migrated. Please join us at community.neo4j.com.

3.5.0 ent cartesian error?

senica
Node Link

Something like this happened before on a version update and it turned out to be a bug in the cartesian return, so forgive me if I'm jumping the gun here....

On v3.4.10-enterprise everything worked as expected. On 3.5.0 things are out of place in the return.

Here is the query:

MATCH (u:user {uuid: $uid})-[:allowed_to]->(c:company {uuid: $cid})
    CREATE (u)-[:owns]->(p:payment $data)<-[:allows]-(c)
    WITH *
    MATCH (a:account {uuid: $account})<-[:allows]-(c)

    WITH u, c, p, a
    UNWIND $distribution as distribution
    MATCH (pv:payment_voucher {uuid: distribution.payment_voucher})-[:bills]->(r:receivable)
    MATCH (r)-[:references]->(pr:product)-[:deposits]->(pa:account)
    CREATE (u)-[:owns]->(d:distribution)<-[:allows]-(c)
    SET d = distribution
    CREATE (p)-[:splits]->(d)
    CREATE (d)-[:pays]->(pv)
    MERGE (p)-[:deposits]->(a)
    CREATE (r)-[:references]->(p)
    CREATE (r)-[:references]->(d)

    WITH p, r, a, pr, pa, collect(d) AS distributions, sum(d.amount) as distribution_amount, pv ORDER BY pv.due_date ASC
    SET pa.balance = pa.balance + distribution_amount
    SET a.balance = a.balance + distribution_amount
    SET pv.paid = pv.paid + distribution_amount
    SET pv.unpaid = pv.amount - pv.paid
    SET pv.status = CASE
      WHEN pv.unpaid = 0 THEN 'paid'
      WHEN pv.due_date <= $now THEN 'pastdue'
      WHEN pv.unpaid < pv.amount THEN 'partial'
      ELSE 'unpaid' END

    WITH p, r, a, pr{.*, account: pa} AS product, sum(distribution_amount) AS voucher_amount, collect(pv{.*, distributions: distributions}) AS payment_vouchers

    SET r.paid = r.paid + voucher_amount
    SET r.unpaid = r.total - r.paid
    SET r.status = CASE
      WHEN r.unpaid = 0 THEN 'paid'
      // WHEN ANY (x IN payment_vouchers WHERE x.status = 'pastdue') THEN 'pastdue'
      // we don't have all the payment_vouchers to check if any are past due.
      // so, if it was past due before, then it's past due now.
      WHEN r.status = 'pastdue' THEN 'pastdue'
      WHEN r.unpaid < r.total THEN 'partial'
      ELSE 'unpaid' END

    WITH p, collect(r{.*, product: product, account: a, payment_vouchers: payment_vouchers}) AS receivables

    RETURN p{.*,
      receivables: receivables
    }

I'm using the nodejs neo4j-driver 1.7.2

The most notable thing that I've found which led me to a cartesian result error is that result[0].uuid points to an account, not a payment. Whereas in 3.4.10, result[0].uuid pointed to the uuid of a payment as expected.

I looked through the release notes to see if anything notable changed, but we aren't doing anything fancy in here that would make me think it was suppose to be a breaking change.

Any thoughts?

11 REPLIES 11

Don't use numeric id's to index result colums, always use names.

Can you share the profile / explain for both versions?

senica
Node Link

Sorry, modified the wrong thing 😞 One moment

senica
Node Link

PROFILE 3.5.0
Here is the profile. From the browser, it is return a receivable at the top level of the object instead of the payment. This seems to differ from the nodejs neo4j-driver (but I need to confirm that )

senica
Node Link

EXPLAIN 3.5.0
Here is the explain:

senica
Node Link

So I was mistaken about the browser and the driver returning different results. They are the same results. However, the issue is still the same in that it returns a receivable instead of a payment.

I think it would be best if you could share the statement and the misbehavior in a GitHub issue with a dataset that allows the Neo4j Cypher team to reproduce the issue.

If you find any way to minimizing it to the relevant bits that would be great.

Also do you have the profile from the other version? And please note which is which.

senica
Node Link

I setup the older version 3.4.10. This one works as expected.

Here is the PROFILE 3.4.10

senica
Node Link

EXPLAIN 3.4.10

senica
Node Link

thanks a lot. it will be handled there.