derivative node returns value from the future

[phemmer]

The derivative() node computes the derivative between a point, and the one that follows it, but adds the calculated derivative value to the first point, not the second.
This is counterintuitive as this is not how a derivative is defined, nor is it how InfluxDB's derivative behaves.
Additionally the behavior results in the derivative() node buffering a point. If it behaved as expected, buffering would not be necessary.

foo v=1i 1000000000
foo v=2i 2000000000
foo v=4i 3000000000
foo v=8i 4000000000
foo v=16i 5000000000

stream|from().measurement('foo')|derivative('v').as('derivative')|log()

Expected:

"Fields":{"v":1},"Time":"1970-01-01T00:00:01Z"
"Fields":{"v":2,"derivative":1},"Time":"1970-01-01T00:00:02Z"
"Fields":{"v":4,"derivative":2},"Time":"1970-01-01T00:00:03Z"
"Fields":{"v":8,"derivative":4},"Time":"1970-01-01T00:00:04Z"
"Fields":{"v":16,"derivative":8},"Time":"1970-01-01T00:00:05Z"

Actual:

"Fields":{"v":1,"derivative":1},"Time":"1970-01-01T00:00:01Z"
"Fields":{"v":2,"derivative":2},"Time":"1970-01-01T00:00:02Z"
"Fields":{"v":4,"derivative":4},"Time":"1970-01-01T00:00:03Z"
"Fields":{"v":8,"derivative":8},"Time":"1970-01-01T00:00:04Z"

[phemmer]

Another issue I just encountered with the derivative() node:

foo v=1i 1000000000
foo v=2i 2000000000
foo x=3i 2500000000
foo v=4i 3000000000
foo v=8i 4000000000
foo v=16i 5000000000

stream|from().measurement('foo')|derivative('v').as('derivative')|log()

Actual:

"Fields":{"derivative":1,"v":1},"Time":"1970-01-01T00:00:01Z"
"Fields":{"derivative":4,"v":4},"Time":"1970-01-01T00:00:03Z"
"Fields":{"derivative":8,"v":8},"Time":"1970-01-01T00:00:04Z"

Due to the original issue this was opened for, I expected a couple different things, but this output was not one of them. I can't make sense of what's going on here. Obviously we have a point where the v field is missing, so we can't use it to calculate a derivative. But what happened to the v=2i point? The derivative between the "v":1 and "v":4 point certainly isn't 1.

[nathanielc]

@phemmer Looking through your examples I agree the actual behavior is quite confusing. We will take a look at the best way to fix this while minimizing any breaking changes to existing behavior.

[phemmer]

I actually started work on this. Basically I just adjusted it so that the derivative is added to the latter point, not the former one, and the first point in the stream/batch is dropped (currently in batch the last point is dropped).
But I can abandon the work if you want to handle it yourself.

[nathanielc]

Go ahead a put up a PR on what you have, we can start the discussion from there. Your solution sounds like what I was thinking.