RxDart vs FxDart › #50 · One source, two independent readers

One source, two independent readers

Toss-up

• tee

Requirement

One reading source must feed two independent computations — the total and the peak — while running exactly once. The source increments a counter each time it runs; print the total, the peak, and the counter to prove the single pass. The data is in the code; both versions must print the lines shown under Expected output.

total: 74
peak: 25
source runs: 1

Side by side

Why they differ

Both models hit the same wall here: their sources restart per consumer. Listening to a plain single-subscription stream twice is an error; iterating a sync* generator twice quietly runs it twice. And both libraries answer with the same idea — share one pass. RxDart makes the stream connectable: publish() defers the source, both reductions subscribe, and connect() starts the single subscription that feeds them. FxDart's tee keeps both reductions in step instead: each element advances the total and the peak before the next one is pulled, so the single pass never has to remember anything.

Both avoid a buffer, and for the same underlying reason — every reader sees each element while it is the current one. That is what connect() buys by making the readers attach first, and what tee buys by taking the readers as folds: a seed and a step, rather than two pipelines free to advance independently. The constraint is the price. publish() will feed any stream operators you care to subscribe; tee only feeds folds. When the two readers really are independent pipelines, FxDart's answer is fork — every fork of the same iterable object is a cursor over one shared, buffered pass — and there the buffer comes back, holding every value until the slowest cursor has consumed it. So this is a tie on capability: the general tool costs memory on both sides, and the specialised one is free on both. Pick the one matching the model the rest of your code already lives in.