Issue 1787 extended stats (#2247)

* first version of extended stats along with its tests

* using IntermediateExtendStats instead of IntermediateStats with all tests passing

* Created struct for request and response

* first test with extended_stats

* kahan summation and tests with approximate equality

* version ready for merge

* removed approx dependency

* refactor for using ExtendedStats only when needed

* interim version

* refined version with code formatted

* refactored a struct

* cosmetic refactor

* fix after merge

* fix format

* added extended_stat bench

* merge and new benchmark for extended stats

* split stat segment collectors

* wrapped intermediate extended stat with a box to limit memory usage

* Revert "wrapped intermediate extended stat with a box to limit memory usage"

This reverts commit 5b4aa9f393.

* some code reformat, commented kahan summation

* refactor after review

* refactor after code review

* fix after incorrectly restoring kahan summation

* modifications for code review + bug fix in merge_fruit

* refactor assert_nearly_equals macro

* update after code review

---------

Co-authored-by: Giovanni Cuccu <gcuccu@imolainformatica.it>
This commit is contained in:
giovannicuccu
2024-06-04 08:25:17 +02:00
committed by GitHub
parent c0686515a9
commit 1095c9b073
11 changed files with 1271 additions and 26 deletions

View File

@@ -47,6 +47,7 @@ fn bench_agg(mut group: InputGroup<Index>) {
register!(group, average_f64);
register!(group, average_f64_u64);
register!(group, stats_f64);
register!(group, extendedstats_f64);
register!(group, percentiles_f64);
register!(group, terms_few);
register!(group, terms_many);
@@ -105,7 +106,12 @@ fn stats_f64(index: &Index) {
});
exec_term_with_agg(index, agg_req)
}
fn extendedstats_f64(index: &Index) {
let agg_req = json!({
"extendedstats_f64": { "extended_stats": { "field": "score_f64", } }
});
exec_term_with_agg(index, agg_req)
}
fn percentiles_f64(index: &Index) {
let agg_req = json!({
"mypercentiles": {

View File

@@ -34,7 +34,7 @@ use super::bucket::{
DateHistogramAggregationReq, HistogramAggregation, RangeAggregation, TermsAggregation,
};
use super::metric::{
AverageAggregation, CountAggregation, MaxAggregation, MinAggregation,
AverageAggregation, CountAggregation, ExtendedStatsAggregation, MaxAggregation, MinAggregation,
PercentilesAggregationReq, StatsAggregation, SumAggregation, TopHitsAggregation,
};
@@ -146,6 +146,11 @@ pub enum AggregationVariants {
/// extracted values.
#[serde(rename = "stats")]
Stats(StatsAggregation),
/// Computes a collection of estended statistics (`min`, `max`, `sum`, `count`, `avg`,
/// `sum_of_squares`, `variance`, `variance_sampling`, `std_deviation`,
/// `std_deviation_sampling`) over the extracted values.
#[serde(rename = "extended_stats")]
ExtendedStats(ExtendedStatsAggregation),
/// Computes the sum of the extracted values.
#[serde(rename = "sum")]
Sum(SumAggregation),
@@ -170,6 +175,7 @@ impl AggregationVariants {
AggregationVariants::Max(max) => vec![max.field_name()],
AggregationVariants::Min(min) => vec![min.field_name()],
AggregationVariants::Stats(stats) => vec![stats.field_name()],
AggregationVariants::ExtendedStats(extended_stats) => vec![extended_stats.field_name()],
AggregationVariants::Sum(sum) => vec![sum.field_name()],
AggregationVariants::Percentiles(per) => vec![per.field_name()],
AggregationVariants::TopHits(top_hits) => top_hits.field_names(),

View File

@@ -11,8 +11,8 @@ use super::bucket::{
DateHistogramAggregationReq, HistogramAggregation, RangeAggregation, TermsAggregation,
};
use super::metric::{
AverageAggregation, CountAggregation, MaxAggregation, MinAggregation, StatsAggregation,
SumAggregation,
AverageAggregation, CountAggregation, ExtendedStatsAggregation, MaxAggregation, MinAggregation,
StatsAggregation, SumAggregation,
};
use super::segment_agg_result::AggregationLimits;
use super::VecWithNames;
@@ -276,6 +276,10 @@ impl AggregationWithAccessor {
field: ref field_name,
..
})
| ExtendedStats(ExtendedStatsAggregation {
field: ref field_name,
..
})
| Sum(SumAggregation {
field: ref field_name,
..

View File

@@ -8,7 +8,9 @@ use rustc_hash::FxHashMap;
use serde::{Deserialize, Serialize};
use super::bucket::GetDocCount;
use super::metric::{PercentilesMetricResult, SingleMetricResult, Stats, TopHitsMetricResult};
use super::metric::{
ExtendedStats, PercentilesMetricResult, SingleMetricResult, Stats, TopHitsMetricResult,
};
use super::{AggregationError, Key};
use crate::TantivyError;
@@ -88,6 +90,8 @@ pub enum MetricResult {
Min(SingleMetricResult),
/// Stats metric result.
Stats(Stats),
/// ExtendedStats metric result.
ExtendedStats(Box<ExtendedStats>),
/// Sum metric result.
Sum(SingleMetricResult),
/// Percentiles metric result.
@@ -104,6 +108,7 @@ impl MetricResult {
MetricResult::Max(max) => Ok(max.value),
MetricResult::Min(min) => Ok(min.value),
MetricResult::Stats(stats) => stats.get_value(agg_property),
MetricResult::ExtendedStats(extended_stats) => extended_stats.get_value(agg_property),
MetricResult::Sum(sum) => Ok(sum.value),
MetricResult::Percentiles(_) => Err(TantivyError::AggregationError(
AggregationError::InvalidRequest("percentiles can't be used to order".to_string()),

View File

@@ -19,8 +19,8 @@ use super::bucket::{
GetDocCount, Order, OrderTarget, RangeAggregation, TermsAggregation,
};
use super::metric::{
IntermediateAverage, IntermediateCount, IntermediateMax, IntermediateMin, IntermediateStats,
IntermediateSum, PercentilesCollector, TopHitsTopNComputer,
IntermediateAverage, IntermediateCount, IntermediateExtendedStats, IntermediateMax,
IntermediateMin, IntermediateStats, IntermediateSum, PercentilesCollector, TopHitsTopNComputer,
};
use super::segment_agg_result::AggregationLimits;
use super::{format_date, AggregationError, Key, SerializedKey};
@@ -215,6 +215,9 @@ pub(crate) fn empty_from_req(req: &Aggregation) -> IntermediateAggregationResult
Stats(_) => IntermediateAggregationResult::Metric(IntermediateMetricResult::Stats(
IntermediateStats::default(),
)),
ExtendedStats(_) => IntermediateAggregationResult::Metric(
IntermediateMetricResult::ExtendedStats(IntermediateExtendedStats::default()),
),
Sum(_) => IntermediateAggregationResult::Metric(IntermediateMetricResult::Sum(
IntermediateSum::default(),
)),
@@ -282,6 +285,8 @@ pub enum IntermediateMetricResult {
Min(IntermediateMin),
/// Intermediate stats result.
Stats(IntermediateStats),
/// Intermediate stats result.
ExtendedStats(IntermediateExtendedStats),
/// Intermediate sum result.
Sum(IntermediateSum),
/// Intermediate top_hits result
@@ -306,6 +311,9 @@ impl IntermediateMetricResult {
IntermediateMetricResult::Stats(intermediate_stats) => {
MetricResult::Stats(intermediate_stats.finalize())
}
IntermediateMetricResult::ExtendedStats(intermediate_stats) => {
MetricResult::ExtendedStats(intermediate_stats.finalize())
}
IntermediateMetricResult::Sum(intermediate_sum) => {
MetricResult::Sum(intermediate_sum.finalize().into())
}
@@ -346,6 +354,12 @@ impl IntermediateMetricResult {
) => {
stats_left.merge_fruits(stats_right);
}
(
IntermediateMetricResult::ExtendedStats(extended_stats_left),
IntermediateMetricResult::ExtendedStats(extended_stats_right),
) => {
extended_stats_left.merge_fruits(extended_stats_right);
}
(IntermediateMetricResult::Sum(sum_left), IntermediateMetricResult::Sum(sum_right)) => {
sum_left.merge_fruits(sum_right);
}

File diff suppressed because it is too large Load Diff

View File

@@ -18,6 +18,7 @@
mod average;
mod count;
mod extended_stats;
mod max;
mod min;
mod percentiles;
@@ -29,6 +30,7 @@ use std::collections::HashMap;
pub use average::*;
pub use count::*;
pub use extended_stats::*;
pub use max::*;
pub use min::*;
pub use percentiles::*;

View File

@@ -1,3 +1,5 @@
use std::fmt::Debug;
use serde::{Deserialize, Serialize};
use super::*;
@@ -85,13 +87,15 @@ impl Stats {
#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
pub struct IntermediateStats {
/// The number of extracted values.
count: u64,
pub(crate) count: u64,
/// The sum of the extracted values.
sum: f64,
pub(crate) sum: f64,
/// delta for sum needed for [Kahan algorithm for summation](https://en.wikipedia.org/wiki/Kahan_summation_algorithm)
pub(crate) delta: f64,
/// The min value.
min: f64,
pub(crate) min: f64,
/// The max value.
max: f64,
pub(crate) max: f64,
}
impl Default for IntermediateStats {
@@ -99,6 +103,7 @@ impl Default for IntermediateStats {
Self {
count: 0,
sum: 0.0,
delta: 0.0,
min: f64::MAX,
max: f64::MIN,
}
@@ -109,7 +114,13 @@ impl IntermediateStats {
/// Merges the other stats intermediate result into self.
pub fn merge_fruits(&mut self, other: IntermediateStats) {
self.count += other.count;
self.sum += other.sum;
// kahan algorithm for sum
let y = other.sum - (self.delta + other.delta);
let t = self.sum + y;
self.delta = (t - self.sum) - y;
self.sum = t;
self.min = self.min.min(other.min);
self.max = self.max.max(other.max);
}
@@ -141,9 +152,15 @@ impl IntermediateStats {
}
#[inline]
fn collect(&mut self, value: f64) {
pub(in crate::aggregation::metric) fn collect(&mut self, value: f64) {
self.count += 1;
self.sum += value;
// kahan algorithm for sum
let y = value - self.delta;
let t = self.sum + y;
self.delta = (t - self.sum) - y;
self.sum = t;
self.min = self.min.min(value);
self.max = self.max.max(value);
}
@@ -288,7 +305,6 @@ impl SegmentAggregationCollector for SegmentStatsCollector {
#[cfg(test)]
mod tests {
use serde_json::Value;
use crate::aggregation::agg_req::{Aggregation, Aggregations};

View File

@@ -11,12 +11,12 @@ use super::agg_req_with_accessor::{AggregationWithAccessor, AggregationsWithAcce
use super::bucket::{SegmentHistogramCollector, SegmentRangeCollector, SegmentTermCollector};
use super::intermediate_agg_result::IntermediateAggregationResults;
use super::metric::{
AverageAggregation, CountAggregation, MaxAggregation, MinAggregation,
AverageAggregation, CountAggregation, ExtendedStatsAggregation, MaxAggregation, MinAggregation,
SegmentPercentilesCollector, SegmentStatsCollector, SegmentStatsType, StatsAggregation,
SumAggregation,
};
use crate::aggregation::bucket::TermMissingAgg;
use crate::aggregation::metric::TopHitsSegmentCollector;
use crate::aggregation::metric::{SegmentExtendedStatsCollector, TopHitsSegmentCollector};
pub(crate) trait SegmentAggregationCollector: CollectorClone + Debug {
fn add_intermediate_aggregation_result(
@@ -148,6 +148,9 @@ pub(crate) fn build_single_agg_segment_collector(
accessor_idx,
*missing,
))),
ExtendedStats(ExtendedStatsAggregation { missing, sigma, .. }) => Ok(Box::new(
SegmentExtendedStatsCollector::from_req(req.field_type, *sigma, accessor_idx, *missing),
)),
Sum(SumAggregation { missing, .. }) => Ok(Box::new(SegmentStatsCollector::from_req(
req.field_type,
SegmentStatsType::Sum,

View File

@@ -871,7 +871,10 @@ mod tests {
use crate::schema::{Field, Schema, FAST, STORED, TEXT};
use crate::time::format_description::well_known::Rfc3339;
use crate::time::OffsetDateTime;
use crate::{DateTime, DocAddress, DocId, Index, IndexWriter, Order, Score, SegmentReader};
use crate::{
assert_nearly_equals, DateTime, DocAddress, DocId, Index, IndexWriter, Order, Score,
SegmentReader,
};
fn make_index() -> crate::Result<Index> {
let mut schema_builder = Schema::builder();

View File

@@ -397,16 +397,20 @@ pub mod tests {
#[macro_export]
macro_rules! assert_nearly_equals {
($left:expr, $right:expr) => {{
match (&$left, &$right) {
(left_val, right_val) => {
assert_nearly_equals!($left, $right, 0.0005);
}};
($left:expr, $right:expr, $epsilon:expr) => {{
match (&$left, &$right, &$epsilon) {
(left_val, right_val, epsilon_val) => {
let diff = (left_val - right_val).abs();
let add = left_val.abs() + right_val.abs();
if diff > 0.0005 * add {
if diff > *epsilon_val {
panic!(
r#"assertion failed: `(left ~= right)`
left: `{:?}`,
right: `{:?}`"#,
&*left_val, &*right_val
r#"assertion failed: `abs(left-right)>epsilon`
left: `{:?}`,
right: `{:?}`,
epsilon: `{:?}`"#,
&*left_val, &*right_val, &*epsilon_val
)
}
}