Struct hdrhistogram::Histogram
[−]
[src]
pub struct Histogram { // some fields omitted }
Instance of a Histogram.
Methods
impl Histogram
fn init(lowest_trackable_value: u64, highest_trackable_value: u64, significant_figures: u32) -> Result<Histogram>
Create a new Histogram instance. lowest_trackable_value
..highest_trackable_value
defines
the range of values in the Histogram
. lowest_trackable_value
must be >= 1.
significant_figures
defines the precision, and must be in the range 1..5.
HistogramErr
is the catch-all failure case. It doesn't report much detail because the
underlying library doesn't.
Example
let mut h = Histogram::init(1, 100000, 2).unwrap(); h.record_value(10); // record a single count of '10'
fn reset(&mut self)
Zero all histogram state in place.
fn record_value(&mut self, value: u64) -> bool
Record a specific value. Returns true if successful.
h.record_value(5); assert_eq!(h.total_count(), 1);
fn record_values(&mut self, value: u64, count: u64) -> bool
Record multiple counts of a specific value. Returns true if successful.
h.record_values(5, 10); assert_eq!(h.total_count(), 10);
fn record_corrected_value(&mut self, value: u64, expected_interval: u64) -> bool
Record a value, correcting for coordinated omission. This should be used when accumulating latency measurements taked on a regular timebase (expected_interval). Any latency that's well above that interval implies some kind of outage in which sampled were lost. This corrects for those lost samples to preserve the integrity of the overall statistics.
fn record_corrected_values(&mut self, value: u64, count: u64, expected_interval: u64) -> bool
As with record_corrected_value()
but multiple counts of the value.
fn add(&mut self, other: &Histogram) -> u64
Sum two histograms, modifying self
in place. Returns the number of samples dropped;
samples will be dropped if they're out of the range lowest_trackable_value .. highest_trackable_value
.
fn add_while_correcting_for_coordinated_omission(&mut self, other: &Histogram, expected_interval: u64) -> u64
As with add
but corrects of potential lost samples while doing periodic latency
measurements, as in record_corrected_value
. Only one correction should be applied.
fn total_count(&self) -> u64
Total of all counters
fn min(&self) -> u64
Smallest recorded value.
h.record_value(1); h.record_value(5); assert_eq!(h.min(), 1);
fn max(&self) -> u64
Largest recorded value.
h.record_value(1); h.record_value(5); assert_eq!(h.max(), 5);
fn value_at_percentile(&self, percentile: f64) -> u64
Value at a particular percentile (0-100).
h.record_values(20, 10); h.record_value(40); assert_eq!(h.value_at_percentile(50.0), 20); assert_eq!(h.value_at_percentile(99.0), 40);
fn stddev(&self) -> f64
Standard deviation of values.
fn mean(&self) -> f64
Mean of values.
fn values_are_equivalent(&self, a: u64, b: u64) -> bool
Returns true if two values are the same according to the lowest, highest and significant figures parameters.
fn lowest_equivalent_value(&self, value: u64) -> u64
Lowest value equivalent to the given value.
fn count_at_value(&self, value: u64) -> u64
Count for a specific value.
fn linear_iter<'a>(&'a self, value_units_per_bucket: u64) -> LinearIter<'a>
Linear iterator over values. Results are returned in equally weighted buckets.
let mut h = Histogram::init(1, 100000, 3).unwrap(); for i in 1..100 { h.record_values(i, i); } for (i, c) in h.linear_iter(1).enumerate() { // 100 buckets println!("bucket {} = {}", i, c.count_added_in_this_iteration_step); }
fn log_iter<'a>(&'a self, value_units_per_bucket: u64, log_base: f64) -> LogIter<'a>
Logarithmic iterator over values. Results are returned in logarithmically weighted buckets.
fn recorded_iter<'a>(&'a self) -> RecordedIter<'a>
Iterator over recorded values.
fn percentile_iter<'a>(&'a self, ticks_per_half_distance: u32) -> PercentileIter<'a>
Iterator over percentiles.
fn encode(&self) -> Result<String>
Encode Histogram
state into a Base64 encoded string.
fn decode(base64: &String) -> Result<Histogram>
Decode Histogram
state from a Base64 string generated by encode
.