Fix for issues #622 & #623

[ningsuhen]

Fix for issues #622 & #623

polymorphic hasOne needs separate handling and hasMany->referencesMany has a unique case which is fixed for MongoDB as suggested by @fabien.

[raymondfeng]

LGTM. @fabien Please review.

[fabien]

I wonder if we should check that obj is actually defined at this stage (defensive coding ...)? There are probably other parts of this code where this might be appropriate.

[ningsuhen]

I think that it makes sense only for the referencesMany -> hasMany case because it is possible to have no value for a map key. In other cases, it should always have a map entry.

[fabien]

@raymondfeng other than the comment, LGTM

[raymondfeng]

@ningsuhen Can you address the comment from @fabien so that we can merge the PR?

[ningsuhen]

@raymondfeng , I had replied to the comment. I think it is applicable only in case of referencesMany <-> hasMany because target[relation.keyTo] can be an Array of ObjectIds instead of an ObjectId. Even if there are other cases where it can be an array, we need to handle it as an array apart from just null checking. Is there any other case where referencesMany can be used in conjunction with another relation?

[fabien]

@raymondfeng LGTM

[ningsuhen]

@fabien, @raymondfeng , I have added the checks anyway :).