Multi-stream: tumbling windows are just groupingBy on time/size

A fixed window needs no windowing library: integer-divide the timestamp by the window width and that quotient IS the window id. A TreeMap downstream keeps the windows in clock order.

Code

record Event(int at, String kind) {}

var events = List.of(new Event(2, "ok"), new Event(7, "ok"), new Event(11, "err"),
                     new Event(13, "ok"), new Event(26, "err"), new Event(28, "ok"));
int window = 10;

Map<Integer, Long> perWindow = events.stream()
        .collect(Collectors.groupingBy(e -> e.at() / window, TreeMap::new, Collectors.counting()));

perWindow.forEach((w, n) ->
        System.out.println("[" + (w * window) + "-" + (w * window + window - 1) + "s] " + n));

Output

[0-9s] 2
[10-19s] 2
[20-29s] 2