Allow empty string to represent apex records

[kntsoriano]

Fixes #451.

Description

When creating a ZoneRecordInput and ZoneRecordUpdateInput with name='', the to_json() command omits the name attribute because of the omitempty command. The omitempty command removes empty strings by default.

This PR changes the behavior to allow a name with an empty string to be rendered by the to_json command. This name with an empty string represents an apex record.

[kntsoriano]

@dallasread Who else do you recommend to review this PR?

[weppos]

@kntsoriano have you tested that this behavior doesn't conflict with the update? We had this issue in the past, where forcing the empty name to always render, caused issues with updates where we update would reset all names to empty.

This is because on update you can provide a partial changeset (e.g. only TTL), and name should be omitted. But after this change you always pass name which in this case would be empty, causing the update to also reset the name.

Please make sure this change is not introducing bugs on updates.

[kntsoriano]

@weppos Thanks. I have not considered that and I will look into it tomorrow.

[weppos]

@weppos Thanks. I have not considered that and I will look into it tomorrow.

Please let's make sure we have sufficient test coverage for all these cases. Tests should catch that.

[kntsoriano]

Thanks. I've updated the code to allow passing in an empty string "" to represent the name of an apex record. This is also what we recommend in the API documentation.

We should be safe from the issue where a partial update clears the name as the default value of the name parameter is None and not "". If None is passed, the name attribute is not included in the payload. If "" is passed, the name attribute with the value "" is included in the payload.

I'm requesting another round of review.

[kntsoriano]

cc @lokst @dallasread

[lokst]

Thanks for the fix and adding the test!

[dallasread]

Looks good to me – thanks for adding the spec.