Added test that parentId is not undefined when using fetchAll with relations

[kirrg001]

no issue

• the fix for that was added in Fix for issue #629 #1716

[ricardograca]

Small problem. I just ran the proposed test against master and it passed correctly. Can you verify?

[kirrg001]

It wasn't for me 🤔But i was on 0.10.3 when writing these tests. Maybe it is already fixed. Let me double check.

[ricardograca]

Probably fixed in #1716.

[kirrg001]

1. Integration Tests Dialect: mysql Relations Bookshelf Relations Eager Loading - Collections eager loads "belongsToMany" models correctly and parent is not undefined:
   TypeError: Cannot read property 'should' of undefined
   at . (test/integration/relations.js:229:68)
   From previous event:
   at Context. (test/integration/relations.js:228:14)

So yeah on 0.10.3 this test fails when i remove my code adaptions.
But it does not on latest master. Then it was probably already fixed. I just wonder where 🤔I need to check the history.

We could still merge test improvements to assert that relatedData has the correct values.

[ricardograca]

Sure, merging just the test sounds good to me.

[kirrg001]

Probably fixed in #1716.

Yeah maybe.

return new EagerRelation(this.models, response, new this.model()).fetch(options);

this.models are already instantiated models (the collection). I don't understand why we have to instantiate new this.model() empty as third parameter to then call target.belongsToMany.

Furthermore:

https://github.com/bookshelf/bookshelf/blob/master/src/base/eager.js#L52

This calls e.g. tags: function() { return this.belongsToMany(...); }
And the result is always relation.relatedData.parentId=undefined. So i think the fix which was merged happens too late?

[ricardograca]

So i think the fix which was merged happens too late?

Could be. I never delved too deep into the relations code. Are you going to investigate? It's quite possible that your fix is better than the already merged one, in which case it should be replaced. The current test suite has a test in place for #629.

[kirrg001]

Are you going to investigate?

Yes will do. But not before next week 👍Will report back here.
I don't mind which fix we take as long as it's the correct one :)

[kirrg001]

I have not investigated much, but my gut feeling tells me to keep the fix in master as is.
Because my fix instantiates for every model in the collection a new eager relation (performance decrease). I am not saying my fix is incorrect, i think the problem is that the relation code needs to be reconsidered as a whole if we want to improve the code.

I'll just update my PR to check that the relational data is not undefined.