Struct MetricHistogram 

pub struct MetricHistogram { /* private fields */ }

Fixed-bucket latency histogram.

Tracks the distribution of values (typically durations in microseconds) across predefined buckets. No allocator needed — the bucket array is inline. Uses 10 buckets: 100us, 500us, 1ms, 5ms, 10ms, 50ms, 100ms, 500ms, 1s, +Inf.

Buckets are cumulative: each bucket count includes all observations that also fall into lower buckets. This matches the Prometheus histogram convention.

Examples

use grafos_observe::MetricHistogram;

let h = MetricHistogram::new();
h.observe(50);     // 50us — lands in the <=100us bucket
h.observe(2_000);  // 2ms  — lands in the <=5ms bucket

assert_eq!(h.count(), 2);
assert_eq!(h.sum(), 2050);
assert_eq!(h.bucket_count(0), 1); // <=100us: only the 50us observation
assert_eq!(h.bucket_count(3), 2); // <=5ms: both observations

Implementations

impl MetricHistogram

pub const NUM_BUCKETS: usize = 10usize

Number of histogram buckets.

pub const BUCKET_BOUNDS: [u64; 10]

Bucket boundaries in microseconds.

pub const fn new() -> Self

Create a new histogram with the default bucket boundaries.

pub fn observe(&self, value_us: u64)

Record an observation (value in microseconds).

Increments the count for every bucket whose bound is >= the value (cumulative histogram, matching Prometheus convention).

pub fn bucket_count(&self, index: usize) -> u64

Read the cumulative count for a specific bucket index.

pub fn bucket_bound(&self, index: usize) -> u64

Read the bucket upper bound for a specific bucket index (in microseconds).

pub fn count(&self) -> u64

Total number of observations.

pub fn sum(&self) -> u64

Sum of all observed values (in microseconds).

Trait Implementations

impl Default for MetricHistogram

fn default() -> Self

Returns the “default value” for a type.