Models with a Base Class Fail if the Model (a subclass) is what's really in the DbSet about graphdiff HOT 7 CLOSED

I committed and pushed a fix for the bug that gets the test I was struggling with to work, and the several I was checking on, but causes the 4 OwnedCollection tests to fail.

The core of my change is I check the DbContext for DbSets to determine which type to actually use to safely get an ObjectSet on the ObjectContext in GetKeysFor(). But again I suspect there's a smarter way to answer the which-class question... .

from graphdiff.

This is now fixed in my fork:

https://github.com/b9chris/GraphDiff

All tests passing including the test I introduced, and the 4 I mentioned.

I'd like to merge it into main when you get a chance.

from graphdiff.

hey b9chris,

I've had a look at the code and it looks great, and caching of metadata as well 👍
I did have a look around to see if there was a simpler way to find the EntityObject for a derived poco.. but no luck. I'll have a look again tomorrow but it seems reflection of the types as you've done is the best way.

Let me know if there is anything you would like to add, otherwise I'll update everything tomorrow.

from graphdiff.

Excellent! Thanks for the merge. I don't have any further modifications to make since it's both passing all tests and currently used in production code on our end.

from graphdiff.

After a lot of reading and playing with the Entity Framework I think I may have found a simpler way to get the information we need.

var metadata = ((IObjectContextAdapter)db).ObjectContext.MetadataWorkspace;
var type = metadata.GetItems<EntityType>(DataSpace.OSpace).Where(p => p.FullName == entityType.FullName).Single();

foreach (string name in type.KeyMembers.Select(k => k.Name))
    yield return entityType.GetProperty(name);

It seems the ObjectContext is smart enough to 'know about' entities which are derivatives of a base type registered with it, and keeps the EntityType for it in its metadata. I've tested the above with your tests and some extras I have created and they pass.

If you can think of a test that could break the above please let me know. I'll push to master soon.

from graphdiff.

07cacb0

from graphdiff.

Excellent! Yeah as long as it passes all tests it should do just as well. Thanks for looking into that, I knew what I threw together was a real hack, but hey first pass to get it working.

from graphdiff.