RxDart vs FxDart › #21 · Three-reading moving average

Three-reading moving average

FxDart wins

• fx
• windowed
• average
• map

Requirement

Smooth eight hourly temperature readings with a three-reading moving average: for every window of 3 consecutive readings, print the window and its mean to one decimal place — six full windows, no partials. The data is in the code; both versions must print the lines shown under Expected output.

21.0 21.6 22.4 -> avg 21.7
21.6 22.4 23.1 -> avg 22.4
22.4 23.1 22.8 -> avg 22.8
23.1 22.8 22.2 -> avg 22.7
22.8 22.2 21.9 -> avg 22.3
22.2 21.9 21.4 -> avg 21.8

Side by side

Why they differ

RxDart spells a sliding window as a parameterisation of batching: bufferCount(3, 1) — buffers of three, a new buffer starting every one event. It works, but the encoding leaks twice. You have to know that the second argument is startBufferEvery and that 1 means "sliding"; and at the end of the stream the operator flushes its still-open buffers, so a ramp-down partial like [21.9, 21.4] comes out too and a where((w) => w.length == 3) has to stand guard for a case the requirement never mentioned.

FxDart has a word for the concept itself: windowed(3) yields exactly the full windows, and partial: true is the explicit opt-in for the ramp-down — the default matches what a moving average means. Add average as a library function (RxDart has no aggregate helpers, so the mean is a hand-rolled reduce-and-divide) and the pull side states the requirement while the push side encodes it. That gap is vocabulary, not model — but the vocabulary exists because windows over an iterable are a pull-native idea, and this one goes to FxDart.