Unexpected Error: UNION types agtype and bigint cannot be matched

[CelineWuest]

When running the following query:

RETURN 0 AS x UNION RETURN COUNT(0) AS x

I encountered the following error:

pq: UNION types agtype and bigint cannot be matched

I believe the query mentioned above is semantically and syntactically correct and thus no error should be thrown here.
Additionally, the query runs successfully in neo4j.

I encountered this issue when testing queries against the apache/age:PG13_latest docker image.

Steps to reproduce

Run the following query and observe it throws an error:

RETURN 0 AS x UNION RETURN COUNT(0) AS x

Expected behavior

The query should run successfully

Actual behavior

The query fails with the error message UNION types agtype and bigint cannot be matched.

[jrgemignani]

@DominicWuest Neo4j doesn't use nor have AGTYPE. Because of that, it won't complain.

The problem here is that you are trying to UNION a container type (or composite type) without specifying what it contains. AGTYPE can be many things, you need to specify what it is in this case.

 psql-11.5-5432-pgsql=# SELECT * from cypher('graph1', $$ RETURN 0::pg_bigint AS x UNION RETURN COUNT(0) AS x $$) as g1(result agtype);
 result
--------
 0
 1
(2 rows)

psql-11.5-5432-pgsql=#

[CelineWuest]

Hi @jrgemignani, thanks for looking into my issue!
I'm sorry, I was unaware of this discrepancy between AGE's implementation and the openCypher specs, I will close this issue as "Won't fix" since it seems like it's intended behavior.

[jrgemignani]

@DominicWuest Some messages, unfortunately, are out of our control as they are from PostgreSQL. It "might" be possible to address some of these, but not all. This is typical of a casting warnings. It will either not find the function or not be able to match something.

[CelineWuest]

I see, thanks for clearing that up @jrgemignani. Do you have any tips about how I could differentiate between fixable and unfixable cases?
By the way, how would you want me to report crashes? Should I also just create issues for those or do you want me to follow the apache reporting procedure?

[jrgemignani]

@DominicWuest For crashes, they need to be as detailed as possible and reproducible. Basically, show us how to reproduce the crash. That would be the PG version, AGE version, how you are connected to the database, graph database dataset, and a reproducible command that causes a crash against the previously mentioned dataset.

[CelineWuest]

Alright, noted @jrgemignani!
I'll put extra care into these reports and submit them soon.