This makes dozens if not hundreds of embedded widgets I built with the previous behavior sort of useless. This might be a more a “true representation of the data” but you broke what I had deployed and replaced it with something worse. I don’t see an obvious way get this back to the previous behavior. It seems like timeseries line graphs are broken by design for sparse data.
I would much prefer that you had added a new parameter to the
average function to get this new behavior than breaking the existing behavior with no way to get it back.
In the absence of a backwards compatible solution, I feel like the average function should accept a default value so that one can actually get line graphs when asking for line graphs rather than a mix of points and line segments.
I recognize the ambiguity of no events, but the point of a line graph is to identify trends. The previous behavior was useful for this and it was reasonably easy to work with the NULL is 0 behavior.
This solution here is like the kid who takes his ball and goes home when the group can’t decide on the rules for the game or whether someone is safe or out. If I wanted more “truth” I should at least have been able to opt in to it.