Struct glean_core::metrics::TimingDistributionMetric

source ·

pub struct TimingDistributionMetric { /* private fields */ }

Expand description

A timing distribution metric.

Timing distributions are used to accumulate and store time measurement, for analyzing distributions of the timing data.

Implementations§

source §

impl TimingDistributionMetric

source

pub fn new(meta: CommonMetricData, time_unit: TimeUnit) -> Self

Creates a new timing distribution metric.

source

pub fn start(&self) -> TimerId

Starts tracking time for the provided metric.

This records an error if it’s already tracking time (i.e. set_start was already called with no corresponding set_stop_and_accumulate): in that case the original start time will be preserved.

Arguments

start_time - Timestamp in nanoseconds.

Returns

A unique TimerId for the new timer.

source

pub fn stop_and_accumulate(&self, id: TimerId)

Stops tracking time for the provided metric and associated timer id.

Adds a count to the corresponding bucket in the timing distribution. This will record an error if no set_start was called.

Arguments

id - The TimerId to associate with this timing. This allows for concurrent timing of events associated with different ids to the same timespan metric.
stop_time - Timestamp in nanoseconds.

source

pub fn cancel(&self, id: TimerId)

Aborts a previous start call.

No error is recorded if no start was called.

Arguments

id - The TimerId to associate with this timing. This allows for concurrent timing of events associated with different ids to the same timing distribution metric.

source

pub fn accumulate_samples(&self, samples: Vec<i64>)

Accumulates the provided signed samples in the metric.

This is required so that the platform-specific code can provide us with 64 bit signed integers if no u64 comparable type is available. This will take care of filtering and reporting errors for any provided negative sample.

Please note that this assumes that the provided samples are already in the “unit” declared by the instance of the metric type (e.g. if the instance this method was called on is using TimeUnit::Second, then samples are assumed to be in that unit).

Arguments

samples - The vector holding the samples to be recorded by the metric.

Notes

Discards any negative value in samples and report an ErrorType::InvalidValue for each of them. Reports an ErrorType::InvalidOverflow error for samples that are longer than MAX_SAMPLE_TIME.

source

pub fn accumulate_single_sample(&self, sample: i64)

Accumulates precisely one signed sample and appends it to the metric.

Precludes the need for a collection in the most common use case.

Sign is required so that the platform-specific code can provide us with a 64 bit signed integer if no u64 comparable type is available. This will take care of filtering and reporting errors for any provided negative sample.

Please note that this assumes that the provided sample is already in the “unit” declared by the instance of the metric type (e.g. if the instance this method was called on is using crate::TimeUnit::Second, then sample is assumed to be in that unit).

Arguments

sample - The singular sample to be recorded by the metric.

Notes

Discards any negative value and reports an ErrorType::InvalidValue. Reports an ErrorType::InvalidOverflow error if the sample is longer than MAX_SAMPLE_TIME.

source

pub fn accumulate_raw_samples_nanos(&self, samples: Vec<u64>)

Accumulates the provided samples in the metric.

Arguments

samples - A list of samples recorded by the metric. Samples must be in nanoseconds.

Notes

Reports an ErrorType::InvalidOverflow error for samples that are longer than MAX_SAMPLE_TIME.

source

pub fn accumulate_raw_duration(&self, duration: Duration)

Accumulates precisely one duration to the metric.

Like TimingDistribution::accumulate_single_sample, but for use when the duration is:

measured externally, or
is in a unit different from the timing_distribution’s internal TimeUnit.

Arguments

duration - The single duration to be recorded in the metric.

Notes

Reports an ErrorType::InvalidOverflow error if duration is longer than MAX_SAMPLE_TIME.

The API client is responsible for ensuring that duration is derived from a monotonic clock source that behaves consistently over computer sleep across the application’s platforms. Otherwise the resulting data may not share the same guarantees that other timing_distribution metrics’ data do.

source