The TargetingId should be included in the FeatureEvaluation event even if it's null/empty.

[zhenlan]

From what I can see in the telemetry data, TargetingId appears to be skipped if it's null/empty, but we do send a FeatureEvaluation event for it. We should include it even if it's null/empty.

(We've discussed this. Open an issue to track it.)

[rossgrambo]

Currently null and empty string are treated as different targets. Should we continue that? I believe python is treating them as the same. If we're going to emit it, the languages will need to do the same thing. cc; @mrm9084

[jimmyca15]

Currently null and empty string are treated as different targets

Can you explain what you mean here? Do you mean if someone sets TargetingContext.UserId to null vs "", there may be a different variant returned?

[rossgrambo]

They get the same variant right now, but if we removed the null or empty check on the outgoing telemetry- we can emit events where the userId is null and other events where the userId is "".

Python would never emit an event for null, as it defaults to "". So a null userId in C# would have a differently shaped impression event than that same "null" user in python.

tldr; should we commit to "" as the null/empty targetingId?

[jimmyca15]

As far as I'm concerned, there is no such thing as a null user id. Our system should effectively treat it as empty string.

[rossgrambo]

Turns out both null and empty are ignored by Application Insights SDK and the field won't be added. So unfortunately we can't include this on the impression event.

We decided to have the ActivityEvent optionally include these as well to follow suit.