How to do an upsert block with multiple linked variable from separate query blocks

Users Dgraph
2

Bulk Upsert is not highly recommended for everyone. In that case, you need to know how to control collateral damage when using such a procedure. My recommendation is to limit this query to one operation at a time.

For example, add a temporary predicate stating that there was a migration. e.g. “<migrated>” and add the check @filter (NOT has (<migrated>)) to your query. That way you avoid messing things up.

To limit one operation at a time, you must use the “first” paging parameter.

e.g

addresses(func: has(Address.tempCountry), first: 1) @filter(NOT has(<migrated>))

...
set {
      uid(address_uid) <Address.country> uid(country_uid) .
      uid(address_uid) <migrated>  "" .
      uid(country_uid) <Country.usedBy> uid(address_uid) .
    }

I would also separate the delete operation for later. I would add a “<to_be_deleted> "Address.tempCountry" . ” predicate and later I would delete them all via upsert block. Just to avoid deleting important data unintentionally.

I didn’t cover every part of the problem, as I didn’t fully understand it. If you provide a sample exemplifying the problem in practice, I may be able to go deeper. But in general, Bulk Upsert does not work in all cases.

Let me know if you have more questions.

Cheers.

2 Likes