Fix handling of null values with AnyOf

[ob-stripe]

r? @remi-stripe @fred-stripe
cc @stripe/api-libraries

Fix handling of null values with AnyOf in the encoder. null values are now ignored, like they would be for a non-AnyOf property.

[remi-stripe]

So this looks good to me. Is it worth adding a test for Source which is critical in this library so that we don't break it again? the created one is a bit of an edge-case and something we wouldn't care to remove the test for.

Also, is there a world in which you wanted to unset something and now it's broken (like how AdditionalOwners could be unset before)?

Sorry for the vague questions, I'm having trouble grasping whether it's possible for something else to be an edge-case.

[ob-stripe]

Is it worth adding a test for Source which is critical in this library so that we don't break it again? the created one is a bit of an edge-case and something we wouldn't care to remove the test for.

Yep, good idea. I added a new ChargeCreateOptionsTest test class to test the various cases for ChargeCreateOptions.Source.

Also, is there a world in which you wanted to unset something and now it's broken (like how AdditionalOwners could be unset before)?

I think we're good:

• list fields can be unset by setting the value to an empty list
• metadata keys can be unset by setting the key to either null or an empty string (in this case, we do encode null as an empty string since we know for sure that the null value was explicitly inserted in the dictionary by the user)

Both cases are already covered in FormEncoderTest.

Can you think of any other cases for unsetting values?

[ob-stripe]

ptal @remi-stripe

[remi-stripe]

Can you think of any other cases for unsetting values?

There's this whole "to unset a hash, send an empty string" like to empty Charge.shipping you do shipping="". Does that still work? Or would it be introduced as a separate concept like we did the Empty in stripe-java?

[ob-stripe]

There's this whole "to unset a hash, send an empty string" like to empty Charge.shipping you do shipping="". Does that still work? Or would it be introduced as a separate concept like we did the Empty in stripe-java?

AFAICT that was never supported in stripe-dotnet. It would be easy to implement something similar to stripe-java's EmptyParam.EMPTY with AnyOf though, but let's do that separately since it would be a new feature.

[ob-stripe]

Released as 27.1.3.