Multi-stream: per-minute error rate with nested groupingBy

Two classifiers stacked - the minute, then the outcome - give a per-window breakdown in one pass. The rate is computed from the inner map, so a minute with no errors still reports 0% rather than being missing.

Code

record Req(int atSeconds, int status) {}

var reqs = List.of(new Req(5, 200), new Req(50, 200), new Req(65, 500), new Req(70, 200),
                   new Req(130, 200), new Req(131, 500), new Req(190, 500));

Map<Integer, Map<String, Long>> perMinute = reqs.stream()
        .collect(Collectors.groupingBy(r -> r.atSeconds() / 60, TreeMap::new,
                 Collectors.groupingBy(r -> r.status() >= 500 ? "err" : "ok",
                                       TreeMap::new, Collectors.counting())));

perMinute.forEach((minute, counts) -> {
    long total = counts.values().stream().mapToLong(Long::longValue).sum();
    System.out.println("minute " + minute + ": " + counts + "  errorRate="
            + String.format(Locale.ROOT, "%.0f%%", 100.0 * counts.getOrDefault("err", 0L) / total));
});

Output

minute 0: {ok=2}  errorRate=0%
minute 1: {err=1, ok=1}  errorRate=50%
minute 2: {err=1, ok=1}  errorRate=50%
minute 3: {err=1}  errorRate=100%