Missing relationship data causes parent data to not be returned

[molexx]

If a query nests to a relationship and there are no rows for that relationship then the parent's data is not returned.

e.g.:

query {
    Humans {
        select {
            id
            name
            homePlanet
        }
    }
}

returns 5 humans. Some have a homePlanet of null and that's fine.

But if you also want to see those humans' favoriteDroid:

query {
    Humans {
        select {
            id
            name
            homePlanet
            favoriteDroid {
              name
            }
        }
    }
}

it only returns 2 humans. Those who don't have a favoriteDroid aren't listed. I think they should be listed and just have a favouriteDroid of null?

[igdianov]

@molexx Good catch! I have created the PR to fix this. Do you think it will make sense to apply JPA schema semantics in order to build queries based on optional attribute specified using JPA's association annotations:

    /** 
     * (Optional) Whether the association is optional. If set 
     * to false then a non-null relationship must always exist.
     */
    boolean optional() default true;

[molexx]

Thanks!

That makes sense... But does JPA enforce that anyway when it runs the query and tries to build its object graph?

[igdianov]

I think Hibernate persistence provider does use optional parameter on association annotations as a hint to create query with fetch graph at runtime when loading the entities. For example, relations of the actual entity specified with optional=false result in an INNER JOIN to their respective relation when loaded, whereas relations specified with optional=true result in LEFT JOINs.

I had struggled with this when building HSQL query criteria for @ManyToMany association attributes using INNER JOIN, because Hibernate would add another LEFT OUTER JOIN to the query automatically and I could not understand why until now. Apparently, @ManyToMany is always optional, so Hibernate always uses left outer join to fetch data.

It would make sense to mimic default Hibernate behavior and derive default join type from optional attribute on the provided associations to build the query. At the same time, I think it is best not to be very opinionated about it and allow to override the the association join type in the actual query, i.e.

    Humans {
        select {
            id
            name
            homePlanet
            favoriteDroid(optional: true) {
              name
            }
        }
    }

The default value of the optional argument in GraphQL schema will be derived from JPA annotation where allowed,

    @ManyToOne(fetch = FetchType.LAZY, optional = true)
    @JoinColumn(name = "favorite_droid_id")
    Droid favoriteDroid;

The default value for optional attribute is true, and so using OUTER JOIN by default makes better sense then INNER JOIN.

What do you think?

[molexx]

Yes that sounds neat and flexible. Thanks!