Revision control
Copy as Markdown
Other Tools
// This Source Code Form is subject to the terms of the Mozilla Public
// License, v. 2.0. If a copy of the MPL was not distributed with this
mod common;
use crate::common::*;
use std::time::Duration;
use serde_json::json;
use glean_core::metrics::*;
use glean_core::storage::StorageManager;
use glean_core::{test_get_num_recorded_errors, ErrorType};
use glean_core::{CommonMetricData, Lifetime};
// Tests ported from glean-ac
#[test]
fn serializer_should_correctly_serialize_timing_distribution() {
let (mut tempdir, _) = tempdir();
let duration = 60;
let time_unit = TimeUnit::Nanosecond;
{
let (glean, dir) = new_glean(Some(tempdir));
tempdir = dir;
let metric = TimingDistributionMetric::new(
CommonMetricData {
name: "distribution".into(),
category: "telemetry".into(),
send_in_pings: vec!["store1".into()],
disabled: false,
lifetime: Lifetime::Ping,
..Default::default()
},
time_unit,
);
let id = 4u64.into();
metric.set_start(id, 0);
metric.set_stop_and_accumulate(&glean, id, duration);
let snapshot = metric
.get_value(&glean, "store1")
.expect("Value should be stored");
assert_eq!(snapshot.count, 1);
assert_eq!(snapshot.sum, duration as i64);
}
// Make a new Glean instance here, which should force reloading of the data from disk
// so we can ensure it persisted, because it has User lifetime
{
let (glean, _t) = new_glean(Some(tempdir));
let snapshot = StorageManager
.snapshot_as_json(glean.storage(), "store1", true)
.unwrap();
// We check the exact format to catch changes to the serialization.
let expected = json!({
"sum": duration,
"values": {
"58": 1,
}
});
assert_eq!(
expected,
snapshot["timing_distribution"]["telemetry.distribution"]
);
}
}
#[test]
fn set_value_properly_sets_the_value_in_all_stores() {
let (glean, _t) = new_glean(None);
let store_names: Vec<String> = vec!["store1".into(), "store2".into()];
let duration = 1;
let metric = TimingDistributionMetric::new(
CommonMetricData {
name: "distribution".into(),
category: "telemetry".into(),
send_in_pings: store_names.clone(),
disabled: false,
lifetime: Lifetime::Ping,
..Default::default()
},
TimeUnit::Nanosecond,
);
let id = 4u64.into();
metric.set_start(id, 0);
metric.set_stop_and_accumulate(&glean, id, duration);
// We check the exact format to catch changes to the serialization.
let expected = json!({
"sum": 1,
"values": {
"1": 1,
}
});
for store_name in store_names {
let snapshot = StorageManager
.snapshot_as_json(glean.storage(), &store_name, true)
.unwrap();
assert_eq!(
expected,
snapshot["timing_distribution"]["telemetry.distribution"]
);
}
}
#[test]
fn timing_distributions_must_not_accumulate_negative_values() {
let (glean, _t) = new_glean(None);
let duration = 60;
let time_unit = TimeUnit::Nanosecond;
let metric = TimingDistributionMetric::new(
CommonMetricData {
name: "distribution".into(),
category: "telemetry".into(),
send_in_pings: vec!["store1".into()],
disabled: false,
lifetime: Lifetime::Ping,
..Default::default()
},
time_unit,
);
// Flip around the timestamps, this should result in a negative value which should be
// discarded.
let id = 4u64.into();
metric.set_start(id, duration);
metric.set_stop_and_accumulate(&glean, id, 0);
assert!(metric.get_value(&glean, "store1").is_none());
// Make sure that the errors have been recorded
assert_eq!(
Ok(1),
test_get_num_recorded_errors(&glean, metric.meta(), ErrorType::InvalidValue)
);
}
#[test]
fn the_accumulate_samples_api_correctly_stores_timing_values() {
let (glean, _t) = new_glean(None);
let metric = TimingDistributionMetric::new(
CommonMetricData {
name: "distribution".into(),
category: "telemetry".into(),
send_in_pings: vec!["store1".into()],
disabled: false,
lifetime: Lifetime::Ping,
..Default::default()
},
TimeUnit::Second,
);
// Accumulate the samples. We intentionally do not report
// negative values to not trigger error reporting.
metric.accumulate_samples_sync(&glean, &[1, 2, 3]);
let snapshot = metric
.get_value(&glean, "store1")
.expect("Value should be stored");
let seconds_to_nanos = 1000 * 1000 * 1000;
// Check that we got the right sum.
assert_eq!(snapshot.sum, 6 * seconds_to_nanos);
// Check that we got the right number of samples.
assert_eq!(snapshot.count, 3);
// We should get a sample in 3 buckets.
// These numbers are a bit magic, but they correspond to
// `hist.sample_to_bucket_minimum(i * seconds_to_nanos)` for `i = 1..=3`.
assert_eq!(1, snapshot.values[&984625593]);
assert_eq!(1, snapshot.values[&1969251187]);
assert_eq!(1, snapshot.values[&2784941737]);
// No errors should be reported.
assert!(test_get_num_recorded_errors(&glean, metric.meta(), ErrorType::InvalidValue).is_err());
}
#[test]
fn the_accumulate_samples_api_correctly_handles_negative_values() {
let (glean, _t) = new_glean(None);
let metric = TimingDistributionMetric::new(
CommonMetricData {
name: "distribution".into(),
category: "telemetry".into(),
send_in_pings: vec!["store1".into()],
disabled: false,
lifetime: Lifetime::Ping,
..Default::default()
},
TimeUnit::Nanosecond,
);
// Accumulate the samples.
metric.accumulate_samples_sync(&glean, &[-1, 1, 2, 3]);
let snapshot = metric
.get_value(&glean, "store1")
.expect("Value should be stored");
// Check that we got the right sum.
assert_eq!(snapshot.sum, 6);
// Check that we got the right number of samples.
assert_eq!(snapshot.count, 3);
// We should get a sample in each of the first 3 buckets.
assert_eq!(1, snapshot.values[&1]);
assert_eq!(1, snapshot.values[&2]);
assert_eq!(1, snapshot.values[&3]);
// 1 error should be reported.
assert_eq!(
Ok(1),
test_get_num_recorded_errors(&glean, metric.meta(), ErrorType::InvalidValue)
);
}
#[test]
fn the_accumulate_samples_api_correctly_handles_overflowing_values() {
let (glean, _t) = new_glean(None);
let metric = TimingDistributionMetric::new(
CommonMetricData {
name: "distribution".into(),
category: "telemetry".into(),
send_in_pings: vec!["store1".into()],
disabled: false,
lifetime: Lifetime::Ping,
..Default::default()
},
TimeUnit::Nanosecond,
);
// The MAX_SAMPLE_TIME is the same from `metrics/timing_distribution.rs`.
const MAX_SAMPLE_TIME: u64 = 1000 * 1000 * 1000 * 60 * 10;
let overflowing_val = MAX_SAMPLE_TIME as i64 + 1;
// Accumulate the samples.
metric.accumulate_samples_sync(&glean, &[overflowing_val, 1, 2, 3]);
let snapshot = metric
.get_value(&glean, "store1")
.expect("Value should be stored");
// Overflowing values are truncated to MAX_SAMPLE_TIME and recorded.
assert_eq!(snapshot.sum as u64, MAX_SAMPLE_TIME + 6);
// Check that we got the right number of samples.
assert_eq!(snapshot.count, 4);
// We should get a sample in each of the first 3 buckets.
assert_eq!(1, snapshot.values[&1]);
assert_eq!(1, snapshot.values[&2]);
assert_eq!(1, snapshot.values[&3]);
// 1 error should be reported.
assert_eq!(
Ok(1),
test_get_num_recorded_errors(&glean, metric.meta(), ErrorType::InvalidOverflow)
);
}
#[test]
fn large_nanoseconds_values() {
let (glean, _t) = new_glean(None);
let metric = TimingDistributionMetric::new(
CommonMetricData {
name: "distribution".into(),
category: "telemetry".into(),
send_in_pings: vec!["store1".into()],
disabled: false,
lifetime: Lifetime::Ping,
..Default::default()
},
TimeUnit::Nanosecond,
);
let time = Duration::from_secs(10).as_nanos() as u64;
assert!(time > u64::from(u32::MAX));
let id = 4u64.into();
metric.set_start(id, 0);
metric.set_stop_and_accumulate(&glean, id, time);
let val = metric
.get_value(&glean, "store1")
.expect("Value should be stored");
// Check that we got the right sum and number of samples.
assert_eq!(val.sum, time as i64);
}
#[test]
fn stopping_non_existing_id_records_an_error() {
let (glean, _t) = new_glean(None);
let metric = TimingDistributionMetric::new(
CommonMetricData {
name: "non_existing_id".into(),
category: "test".into(),
send_in_pings: vec!["store1".into()],
disabled: false,
lifetime: Lifetime::Ping,
..Default::default()
},
TimeUnit::Nanosecond,
);
let id = 3785u64.into();
metric.set_stop_and_accumulate(&glean, id, 60);
// 1 error should be reported.
assert_eq!(
Ok(1),
test_get_num_recorded_errors(&glean, metric.meta(), ErrorType::InvalidState)
);
}
#[test]
fn the_accumulate_raw_samples_api_correctly_stores_timing_values() {
let (glean, _t) = new_glean(None);
let metric = TimingDistributionMetric::new(
CommonMetricData {
name: "distribution".into(),
category: "telemetry".into(),
send_in_pings: vec!["store1".into()],
disabled: false,
lifetime: Lifetime::Ping,
..Default::default()
},
TimeUnit::Second,
);
let seconds_to_nanos = 1000 * 1000 * 1000;
metric.accumulate_raw_samples_nanos_sync(
&glean,
[seconds_to_nanos, 2 * seconds_to_nanos, 3 * seconds_to_nanos].as_ref(),
);
let snapshot = metric
.get_value(&glean, "store1")
.expect("Value should be stored");
// Check that we got the right sum.
assert_eq!(snapshot.sum, 6 * seconds_to_nanos as i64);
// Check that we got the right number of samples.
assert_eq!(snapshot.count, 3);
// We should get a sample in 3 buckets.
// These numbers are a bit magic, but they correspond to
// `hist.sample_to_bucket_minimum(i * seconds_to_nanos)` for `i = 1..=3`.
assert_eq!(1, snapshot.values[&984625593]);
assert_eq!(1, snapshot.values[&1969251187]);
assert_eq!(1, snapshot.values[&2784941737]);
// No errors should be reported.
assert!(test_get_num_recorded_errors(&glean, metric.meta(), ErrorType::InvalidState).is_err());
}
#[test]
fn raw_samples_api_error_cases() {
let (glean, _t) = new_glean(None);
let metric = TimingDistributionMetric::new(
CommonMetricData {
name: "distribution".into(),
category: "telemetry".into(),
send_in_pings: vec!["store1".into()],
disabled: false,
lifetime: Lifetime::Ping,
..Default::default()
},
TimeUnit::Nanosecond,
);
// 10minutes in nanoseconds
let max_sample_time = 1000 * 1000 * 1000 * 60 * 10;
metric.accumulate_raw_samples_nanos_sync(
&glean,
&[
0, /* rounded up to 1 */
1, /* valid */
max_sample_time + 1, /* larger then the maximum, will record an error and the maximum */
],
);
let snapshot = metric
.get_value(&glean, "store1")
.expect("Value should be stored");
// Check that we got the right sum.
assert_eq!(snapshot.sum, 2 + max_sample_time as i64);
// Check that we got the right number of samples.
assert_eq!(snapshot.count, 3);
// We should get a sample in 3 buckets.
// These numbers are a bit magic, but they correspond to
// `hist.sample_to_bucket_minimum(i * seconds_to_nanos)` for `i = {1, max_sample_time}`.
assert_eq!(2, snapshot.values[&1]);
assert_eq!(1, snapshot.values[&599512966122]);
// 1 error should be reported.
assert_eq!(
Ok(1),
test_get_num_recorded_errors(&glean, metric.meta(), ErrorType::InvalidOverflow)
);
}
#[test]
fn timing_distribution_is_tracked_across_upload_toggle() {
let (mut glean, _t) = new_glean(None);
let metric = TimingDistributionMetric::new(
CommonMetricData {
name: "distribution".into(),
category: "telemetry".into(),
send_in_pings: vec!["store1".into()],
disabled: false,
lifetime: Lifetime::Ping,
..Default::default()
},
TimeUnit::Nanosecond,
);
let id = 4u64.into();
let duration = 100;
// Timer is started.
metric.set_start(id, 0);
// User disables telemetry upload.
glean.set_upload_enabled(false);
// App code eventually stops the timer.
// We should clear internal state as upload is disabled.
metric.set_stop_and_accumulate(&glean, id, duration);
assert_eq!(None, metric.get_value(&glean, "store1"));
// App code eventually starts the timer again.
// Upload is disabled, so this should not have any effect.
metric.set_start(id, 100);
// User enables telemetry upload again.
glean.set_upload_enabled(true);
// App code eventually stops the timer.
// The full timespan is recorded.
metric.set_stop_and_accumulate(&glean, id, 100 + duration);
let data = metric.get_value(&glean, "store1").unwrap();
assert_eq!(1, data.count);
assert_eq!(100, data.sum);
// Make sure that the error has been recorded
assert!(test_get_num_recorded_errors(&glean, metric.meta(), ErrorType::InvalidState).is_err());
}