Skip to content

Commit cd3f84c

Browse files
kkrik-eskkrik-es
authored
Switch TDigestState to use HybridDigest by default (#96904)
* Initial import for TDigest forking. * Fix MedianTest. More work needed for TDigestPercentile*Tests and the TDigestTest (and the rest of the tests) in the tdigest lib to pass. * Fix Dist. * Fix AVLTreeDigest.quantile to match Dist for uniform centroids. * Update docs/changelog/96086.yaml * Fix `MergingDigest.quantile` to match `Dist` on uniform distribution. * Add merging to TDigestState.hashCode and .equals. Remove wrong asserts from tests and MergingDigest. * Fix style violations for tdigest library. * Fix typo. * Fix more style violations. * Fix more style violations. * Fix remaining style violations in tdigest library. * Update results in docs based on the forked tdigest. * Fix YAML tests in aggs module. * Fix YAML tests in x-pack/plugin. * Skip failing V7 compat tests in modules/aggregations. * Fix TDigest library unittests. Remove redundant serializing interfaces from the library. * Remove YAML test versions for older releases. These tests don't address compatibility issues in mixed cluster tests as the latter contain a mix of older and newer nodes, so the output depends on which node is picked as a data node since the forked TDigest library is not backwards compatible (produces slightly different results). * Fix test failures in docs and mixed cluster. * Reduce buffer sizes in MergingDigest to avoid oom. * Exclude more failing V7 compatibility tests. * Update results for JdbcCsvSpecIT tests. * Update results for JdbcDocCsvSpecIT tests. * Revert unrelated change. * More test fixes. * Use version skips instead of blacklisting in mixed cluster tests. * Switch TDigestState back to AVLTreeDigest. * Update docs and tests with AVLTreeDigest output. * Update flaky test. * Remove dead code, esp around tracking of incoming data. * Update docs/changelog/96086.yaml * Delete docs/changelog/96086.yaml * Remove explicit compression calls. This was added to prevent concurrency tests from failing, but it leads to reduces precision. Submit this to see if the concurrency tests are still failing. * Revert "Remove explicit compression calls." This reverts commit 5352c96. * Remove explicit compression calls to MedianAbsoluteDeviation input. * Add unittests for AVL and merging digest accuracy. * Fix spotless violations. * Delete redundant tests and benchmarks. * Fix spotless violation. * Use the old implementation of AVLTreeDigest. The latest library version is 50% slower and less accurate, as verified by ComparisonTests. * Update docs with latest percentile results. * Update docs with latest percentile results. * Remove repeated compression calls. * Update more percentile results. * Use approximate percentile values in integration tests. This helps with mixed cluster tests, where some of the tests where blocked. * Fix expected percentile value in test. * Revert in-place node updates in AVL tree. Update quantile calculations between centroids and min/max values to match v.3.2. * Add SortingDigest and HybridDigest. The SortingDigest tracks all samples in an ArrayList that gets sorted for quantile calculations. This approach provides perfectly accurate results and is the most efficient implementation for up to millions of samples, at the cost of bloated memory footprint. The HybridDigest uses a SortingDigest for small sample populations, then switches to a MergingDigest. This approach combines to the best performance and results for small sample counts with very good performance and acceptable accuracy for effectively unbounded sample counts. * Remove deps to the 3.2 library. * Remove unused licenses for tdigest. * Revert changes for SortingDigest and HybridDigest. These will be submitted in a follow-up PR for enabling MergingDigest. * Remove unused Histogram classes and unit tests. Delete dead and commented out code, make the remaining tests run reasonably fast. Remove unused annotations, esp. SuppressWarnings. * Remove Comparison class, not used. * Revert "Revert changes for SortingDigest and HybridDigest." This reverts commit 2336b11. * Use HybridDigest as default tdigest implementation Add SortingDigest as a simple structure for percentile calculations that tracks all data points in a sorted array. This is a fast and perfectly accurate solution that leads to bloated memory allocation. Add HybridDigest that uses SortingDigest for small sample counts, then switches to MergingDigest. This approach delivers extreme performance and accuracy for small populations while scaling indefinitely and maintaining acceptable performance and accuracy with constant memory allocation (15kB by default). Provide knobs to switch back to AVLTreeDigest, either per query or through ClusterSettings. * Small fixes. * Add javadoc and tests. * Add javadoc and tests. * Remove special logic for singletons in the boundaries. While this helps with the case where the digest contains only singletons (perfect accuracy), it has a major issue problem (non-monotonic quantile function) when the first singleton is followed by a non-singleton centroid. It's preferable to revert to the old version from 3.2; inaccuracies in a singleton-only digest should be mitigated by using a sorted array for small sample counts. * Revert changes to expected values in tests. This is due to restoring quantile functions to match head. * Revert changes to expected values in tests. This is due to restoring quantile functions to match head. * Tentatively restore percentile rank expected results. * Use cdf version from 3.2 Update Dist.cdf to use interpolation, use the same cdf version in AVLTreeDigest and MergingDigest. * Revert "Tentatively restore percentile rank expected results." This reverts commit 7718dbb. * Revert remaining changes compared to main. * Revert excluded V7 compat tests. * Exclude V7 compat tests still failing. * Exclude V7 compat tests still failing. * Remove ClusterSettings tentatively. * Initial import for TDigest forking. * Fix MedianTest. More work needed for TDigestPercentile*Tests and the TDigestTest (and the rest of the tests) in the tdigest lib to pass. * Fix Dist. * Fix AVLTreeDigest.quantile to match Dist for uniform centroids. * Update docs/changelog/96086.yaml * Fix `MergingDigest.quantile` to match `Dist` on uniform distribution. * Add merging to TDigestState.hashCode and .equals. Remove wrong asserts from tests and MergingDigest. * Fix style violations for tdigest library. * Fix typo. * Fix more style violations. * Fix more style violations. * Fix remaining style violations in tdigest library. * Update results in docs based on the forked tdigest. * Fix YAML tests in aggs module. * Fix YAML tests in x-pack/plugin. * Skip failing V7 compat tests in modules/aggregations. * Fix TDigest library unittests. Remove redundant serializing interfaces from the library. * Remove YAML test versions for older releases. These tests don't address compatibility issues in mixed cluster tests as the latter contain a mix of older and newer nodes, so the output depends on which node is picked as a data node since the forked TDigest library is not backwards compatible (produces slightly different results). * Fix test failures in docs and mixed cluster. * Reduce buffer sizes in MergingDigest to avoid oom. * Exclude more failing V7 compatibility tests. * Update results for JdbcCsvSpecIT tests. * Update results for JdbcDocCsvSpecIT tests. * Revert unrelated change. * More test fixes. * Use version skips instead of blacklisting in mixed cluster tests. * Switch TDigestState back to AVLTreeDigest. * Update docs and tests with AVLTreeDigest output. * Update flaky test. * Remove dead code, esp around tracking of incoming data. * Remove explicit compression calls. This was added to prevent concurrency tests from failing, but it leads to reduces precision. Submit this to see if the concurrency tests are still failing. * Update docs/changelog/96086.yaml * Delete docs/changelog/96086.yaml * Revert "Remove explicit compression calls." This reverts commit 5352c96. * Remove explicit compression calls to MedianAbsoluteDeviation input. * Add unittests for AVL and merging digest accuracy. * Fix spotless violations. * Delete redundant tests and benchmarks. * Fix spotless violation. * Use the old implementation of AVLTreeDigest. The latest library version is 50% slower and less accurate, as verified by ComparisonTests. * Update docs with latest percentile results. * Update docs with latest percentile results. * Remove repeated compression calls. * Update more percentile results. * Use approximate percentile values in integration tests. This helps with mixed cluster tests, where some of the tests where blocked. * Fix expected percentile value in test. * Revert in-place node updates in AVL tree. Update quantile calculations between centroids and min/max values to match v.3.2. * Add SortingDigest and HybridDigest. The SortingDigest tracks all samples in an ArrayList that gets sorted for quantile calculations. This approach provides perfectly accurate results and is the most efficient implementation for up to millions of samples, at the cost of bloated memory footprint. The HybridDigest uses a SortingDigest for small sample populations, then switches to a MergingDigest. This approach combines to the best performance and results for small sample counts with very good performance and acceptable accuracy for effectively unbounded sample counts. * Remove deps to the 3.2 library. * Remove unused licenses for tdigest. * Revert changes for SortingDigest and HybridDigest. These will be submitted in a follow-up PR for enabling MergingDigest. * Remove unused Histogram classes and unit tests. Delete dead and commented out code, make the remaining tests run reasonably fast. Remove unused annotations, esp. SuppressWarnings. * Remove Comparison class, not used. * Revert "Revert changes for SortingDigest and HybridDigest." This reverts commit 2336b11. * Use HybridDigest as default tdigest implementation Add SortingDigest as a simple structure for percentile calculations that tracks all data points in a sorted array. This is a fast and perfectly accurate solution that leads to bloated memory allocation. Add HybridDigest that uses SortingDigest for small sample counts, then switches to MergingDigest. This approach delivers extreme performance and accuracy for small populations while scaling indefinitely and maintaining acceptable performance and accuracy with constant memory allocation (15kB by default). Provide knobs to switch back to AVLTreeDigest, either per query or through ClusterSettings. * Add javadoc and tests. * Remove ClusterSettings tentatively. * Restore bySize function in TDigest and subclasses. * Update Dist.cdf to match the rest. Update tests. * Revert outdated test changes. * Revert outdated changes. * Small fixes. * Update docs/changelog/96794.yaml * TDigestState uses MergingDigest by default. * Make HybridDigest the default implementation. * Update boxplot documentation. * Use HybridDigest for real. * Restore AVLTreeDigest as the default in TDigestState. TDigest.createHybridDigest nw returns the right type. The switch in TDigestState will happen in a separate PR as it requires many test updates. * Use execution_hint in tdigest spec. * Restore expected test values. * Fix Dist.cdf for empty digest. * Bump up TransportVersion. * More test updates. * Bump up TransportVersion for real. * Restore V7 compat blacklisting. * HybridDigest uses its final implementation during deserialization. * Restore the right TransportVersion in TDigestState.read * More test fixes. * More test updates. * Use TDigestExecutionHint instead of strings. * Add link to TDigest javadoc. * Spotless fix. * Small fixes. * Bump up TransportVersion. * Bump up the TransportVersion, again. * Update docs/changelog/96904.yaml * Delete 96794.yaml Delete existing changelog to get a new one. * Restore previous changelog. * Rename 96794.yaml to 96794.yaml * Update breaking change notes in changelog. * Remove mapping value from changelog. * Set a valid breaking area. * Use HybridDigest as default TDigest impl. * Update docs/changelog/96904.yaml * Use TDigestExecutionHint in MedianAbsoluteDeviationAggregator. * Update changelog and comment in blacklisted V7 compat tests. * Update breaking area in changelog.
1 parent ee4fbe4 commit cd3f84c

File tree

26 files changed

+226
-168
lines changed

26 files changed

+226
-168
lines changed

‎docs/changelog/96904.yaml‎

Lines changed: 18 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,18 @@
1+
pr: 96904
2+
summary: Switch TDigestState to use `HybridDigest` by default
3+
area: Aggregations
4+
type: breaking
5+
issues: []
6+
breaking:
7+
title: Switch TDigestState to use `HybridDigest` by default
8+
area: REST API
9+
details: "The default implementation for TDigest in percentile calculations switches\
10+
\ to a new internal implementation offering superior performance (2x-10x speedup),\
11+
\ at a very small accuracy penalty for very large sample populations."
12+
impact: "This change leads to generating slightly different results in percentile\
13+
\ calculations. If the highest possible accuracy is desired, or it's crucial to\
14+
\ produce exactly the same results as in previous versions, one can either set\
15+
\ `execution_hint` to `high_accuracy` in the `tdigest` spec of a given percentile\
16+
\ calculation, or set `search.aggs.tdigest_execution_hint` to `high_accuracy`\
17+
\ in cluster settings to apply to all percentile queries."
18+
notable: true

‎docs/reference/aggregations/metrics/boxplot-aggregation.asciidoc‎

Lines changed: 4 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -55,9 +55,9 @@ The response will look like this:
5555
"load_time_boxplot": {
5656
"min": 0.0,
5757
"max": 990.0,
58-
"q1": 165.0,
58+
"q1": 167.5,
5959
"q2": 445.0,
60-
"q3": 725.0,
60+
"q3": 722.5,
6161
"lower": 0.0,
6262
"upper": 990.0
6363
}
@@ -114,9 +114,9 @@ GET latency/_search
114114
"load_time_boxplot": {
115115
"min": 0.0,
116116
"max": 99.0,
117-
"q1": 16.5,
117+
"q1": 16.75,
118118
"q2": 44.5,
119-
"q3": 72.5,
119+
"q3": 72.25,
120120
"lower": 0.0,
121121
"upper": 99.0
122122
}

‎docs/reference/aggregations/metrics/percentile-aggregation.asciidoc‎

Lines changed: 32 additions & 18 deletions
Original file line numberDiff line numberDiff line change
@@ -54,19 +54,26 @@ percentiles: `[ 1, 5, 25, 50, 75, 95, 99 ]`. The response will look like this:
5454
"aggregations": {
5555
"load_time_outlier": {
5656
"values": {
57-
"1.0": 5.0,
58-
"5.0": 25.0,
59-
"25.0": 165.0,
57+
"1.0": 10.0,
58+
"5.0": 30.0,
59+
"25.0": 170.0,
6060
"50.0": 445.0,
61-
"75.0": 725.0,
62-
"95.0": 945.0,
63-
"99.0": 985.0
61+
"75.0": 720.0,
62+
"95.0": 940.0,
63+
"99.0": 980.0
6464
}
6565
}
6666
}
6767
}
6868
--------------------------------------------------
6969
// TESTRESPONSE[s/\.\.\./"took": $body.took,"timed_out": false,"_shards": $body._shards,"hits": $body.hits,/]
70+
// TESTRESPONSE[s/"1.0": 10.0/"1.0": 9.9/]
71+
// TESTRESPONSE[s/"5.0": 30.0/"5.0": 29.5/]
72+
// TESTRESPONSE[s/"25.0": 170.0/"25.0": 167.5/]
73+
// TESTRESPONSE[s/"50.0": 445.0/"50.0": 445.0/]
74+
// TESTRESPONSE[s/"75.0": 720.0/"75.0": 722.5/]
75+
// TESTRESPONSE[s/"95.0": 940.0/"95.0": 940.5/]
76+
// TESTRESPONSE[s/"99.0": 980.0/"99.0": 980.1/]
7077

7178
As you can see, the aggregation will return a calculated value for each percentile
7279
in the default range. If we assume response times are in milliseconds, it is
@@ -128,38 +135,45 @@ Response:
128135
"values": [
129136
{
130137
"key": 1.0,
131-
"value": 5.0
138+
"value": 10.0
132139
},
133140
{
134141
"key": 5.0,
135-
"value": 25.0
142+
"value": 30.0
136143
},
137144
{
138145
"key": 25.0,
139-
"value": 165.0
146+
"value": 170.0
140147
},
141148
{
142149
"key": 50.0,
143150
"value": 445.0
144151
},
145152
{
146153
"key": 75.0,
147-
"value": 725.0
154+
"value": 720.0
148155
},
149156
{
150157
"key": 95.0,
151-
"value": 945.0
158+
"value": 940.0
152159
},
153160
{
154161
"key": 99.0,
155-
"value": 985.0
162+
"value": 980.0
156163
}
157164
]
158165
}
159166
}
160167
}
161168
--------------------------------------------------
162169
// TESTRESPONSE[s/\.\.\./"took": $body.took,"timed_out": false,"_shards": $body._shards,"hits": $body.hits,/]
170+
// TESTRESPONSE[s/"value": 10.0/"value": 9.9/]
171+
// TESTRESPONSE[s/"value": 30.0/"value": 29.5/]
172+
// TESTRESPONSE[s/"value": 170.0/"value": 167.5/]
173+
// TESTRESPONSE[s/"value": 445.0/"value": 445.0/]
174+
// TESTRESPONSE[s/"value": 720.0/"value": 722.5/]
175+
// TESTRESPONSE[s/"value": 940.0/"value": 940.5/]
176+
// TESTRESPONSE[s/"value": 980.0/"value": 980.1/]
163177

164178
==== Script
165179

@@ -203,13 +217,13 @@ GET latency/_search
203217
"aggregations": {
204218
"load_time_outlier": {
205219
"values": {
206-
"1.0": 0.5,
207-
"5.0": 2.5,
208-
"25.0": 16.5,
220+
"1.0": 0.99,
221+
"5.0": 2.95,
222+
"25.0": 16.75,
209223
"50.0": 44.5,
210-
"75.0": 72.5,
211-
"95.0": 94.5,
212-
"99.0": 98.5
224+
"75.0": 72.25,
225+
"95.0": 94.05,
226+
"99.0": 98.01
213227
}
214228
}
215229
}

‎docs/reference/aggregations/pipeline/moving-percentiles-aggregation.asciidoc‎

Lines changed: 7 additions & 7 deletions
Original file line numberDiff line numberDiff line change
@@ -97,7 +97,7 @@ And the following may be the response:
9797
"doc_count": 3,
9898
"the_percentile": {
9999
"values": {
100-
"1.0": 150.0,
100+
"1.0": 151.0,
101101
"99.0": 200.0
102102
}
103103
}
@@ -108,13 +108,13 @@ And the following may be the response:
108108
"doc_count": 2,
109109
"the_percentile": {
110110
"values": {
111-
"1.0": 10.0,
112-
"99.0": 50.0
111+
"1.0": 10.4,
112+
"99.0": 49.6
113113
}
114114
},
115115
"the_movperc": {
116116
"values": {
117-
"1.0": 150.0,
117+
"1.0": 151.0,
118118
"99.0": 200.0
119119
}
120120
}
@@ -125,13 +125,13 @@ And the following may be the response:
125125
"doc_count": 2,
126126
"the_percentile": {
127127
"values": {
128-
"1.0": 175.0,
129-
"99.0": 200.0
128+
"1.0": 175.25,
129+
"99.0": 199.75
130130
}
131131
},
132132
"the_movperc": {
133133
"values": {
134-
"1.0": 10.0,
134+
"1.0": 11.6,
135135
"99.0": 200.0
136136
}
137137
}

‎modules/aggregations/build.gradle‎

Lines changed: 11 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -43,7 +43,17 @@ tasks.named("yamlRestTestV7CompatTransform").configure { task ->
4343
task.skipTest("search.aggregation/20_terms/numeric profiler", "The profiler results aren't backwards compatible.")
4444
task.skipTest("search.aggregation/210_top_hits_nested_metric/top_hits aggregation with sequence numbers", "#42809 the use nested path and filter sort throws an exception")
4545
task.skipTest("search.aggregation/370_doc_count_field/Test filters agg with doc_count", "Uses profiler for assertions which is not backwards compatible")
46-
task.skipTest("search.aggregation/420_percentile_ranks_tdigest_metric/filtered", "Uses t-digest library which is not backwards compatible")
46+
47+
// In 8.9.0, the default t-digest algorithm changed from AVL-tree-based to hybrid, combining a sorted array of samples with a merging
48+
// implementation. This change leads to slight different percentile results, compared to previous versions.
49+
task.skipTest("search.aggregation/180_percentiles_tdigest_metric/Basic test", "Hybrid t-digest produces different results.")
50+
task.skipTest("search.aggregation/180_percentiles_tdigest_metric/Non-keyed test", "Hybrid t-digest produces different results.")
51+
task.skipTest("search.aggregation/180_percentiles_tdigest_metric/Only aggs test", "Hybrid t-digest produces different results.")
52+
task.skipTest("search.aggregation/180_percentiles_tdigest_metric/Explicit Percents test", "Hybrid t-digest produces different results.")
53+
task.skipTest("search.aggregation/180_percentiles_tdigest_metric/Metadata test", "Hybrid t-digest produces different results.")
54+
task.skipTest("search.aggregation/180_percentiles_tdigest_metric/Filtered test", "Hybrid t-digest produces different results.")
55+
task.skipTest("search.aggregation/420_percentile_ranks_tdigest_metric/filtered", "Hybrid t-digest produces different results.")
56+
4757
task.addAllowedWarningRegex("\\[types removal\\].*")
4858
}
4959

‎modules/aggregations/src/yamlRestTest/resources/rest-api-spec/test/aggregations/percentiles_tdigest_metric.yml‎

Lines changed: 27 additions & 24 deletions
Original file line numberDiff line numberDiff line change
@@ -46,6 +46,9 @@ setup:
4646

4747
---
4848
"Basic test":
49+
- skip:
50+
version: " - 8.8.99"
51+
reason: t-digest forked in 8.9.0
4952
- do:
5053
search:
5154
rest_total_hits_as_int: true
@@ -61,21 +64,21 @@ setup:
6164
- match: { hits.total: 4 }
6265
- length: { hits.hits: 4 }
6366

64-
- match: { aggregations.percentiles_int.values.1\.0: 1.0 }
65-
- match: { aggregations.percentiles_int.values.5\.0: 1.0 }
66-
- match: { aggregations.percentiles_int.values.25\.0: 26.0 }
67+
- match: { aggregations.percentiles_int.values.1\.0: 2.5 }
68+
- match: { aggregations.percentiles_int.values.5\.0: 8.5 }
69+
- match: { aggregations.percentiles_int.values.25\.0: 38.5 }
6770
- match: { aggregations.percentiles_int.values.50\.0: 76.0 }
68-
- match: { aggregations.percentiles_int.values.75\.0: 126.0 }
69-
- match: { aggregations.percentiles_int.values.95\.0: 151.0 }
70-
- match: { aggregations.percentiles_int.values.99\.0: 151.0 }
71+
- match: { aggregations.percentiles_int.values.75\.0: 113.5 }
72+
- match: { aggregations.percentiles_int.values.95\.0: 143.5 }
73+
- match: { aggregations.percentiles_int.values.99\.0: 149.5 }
7174

72-
- match: { aggregations.percentiles_double.values.1\.0: 1.0 }
73-
- match: { aggregations.percentiles_double.values.5\.0: 1.0 }
74-
- match: { aggregations.percentiles_double.values.25\.0: 26.0 }
75+
- match: { aggregations.percentiles_double.values.1\.0: 2.5 }
76+
- match: { aggregations.percentiles_double.values.5\.0: 8.5 }
77+
- match: { aggregations.percentiles_double.values.25\.0: 38.5 }
7578
- match: { aggregations.percentiles_double.values.50\.0: 76.0 }
76-
- match: { aggregations.percentiles_double.values.75\.0: 126.0 }
77-
- match: { aggregations.percentiles_double.values.95\.0: 151.0 }
78-
- match: { aggregations.percentiles_double.values.99\.0: 151.0 }
79+
- match: { aggregations.percentiles_double.values.75\.0: 113.5 }
80+
- match: { aggregations.percentiles_double.values.95\.0: 143.5 }
81+
- match: { aggregations.percentiles_double.values.99\.0: 149.5 }
7982

8083
- do:
8184
search:
@@ -97,21 +100,21 @@ setup:
97100
- match: { hits.total: 4 }
98101
- length: { hits.hits: 4 }
99102

100-
- match: { aggregations.percentiles_int.values.1\.0: 1.0 }
101-
- match: { aggregations.percentiles_int.values.5\.0: 1.0 }
102-
- match: { aggregations.percentiles_int.values.25\.0: 26.0 }
103+
- match: { aggregations.percentiles_int.values.1\.0: 2.5 }
104+
- match: { aggregations.percentiles_int.values.5\.0: 8.5 }
105+
- match: { aggregations.percentiles_int.values.25\.0: 38.5 }
103106
- match: { aggregations.percentiles_int.values.50\.0: 76.0 }
104-
- match: { aggregations.percentiles_int.values.75\.0: 126.0 }
105-
- match: { aggregations.percentiles_int.values.95\.0: 151.0 }
106-
- match: { aggregations.percentiles_int.values.99\.0: 151.0 }
107+
- match: { aggregations.percentiles_int.values.75\.0: 113.5 }
108+
- match: { aggregations.percentiles_int.values.95\.0: 143.5 }
109+
- match: { aggregations.percentiles_int.values.99\.0: 149.5 }
107110

108-
- match: { aggregations.percentiles_double.values.1\.0: 1.0 }
109-
- match: { aggregations.percentiles_double.values.5\.0: 1.0 }
110-
- match: { aggregations.percentiles_double.values.25\.0: 26.0 }
111+
- match: { aggregations.percentiles_double.values.1\.0: 2.5 }
112+
- match: { aggregations.percentiles_double.values.5\.0: 8.5 }
113+
- match: { aggregations.percentiles_double.values.25\.0: 38.5 }
111114
- match: { aggregations.percentiles_double.values.50\.0: 76.0 }
112-
- match: { aggregations.percentiles_double.values.75\.0: 126.0 }
113-
- match: { aggregations.percentiles_double.values.95\.0: 151.0 }
114-
- match: { aggregations.percentiles_double.values.99\.0: 151.0 }
115+
- match: { aggregations.percentiles_double.values.75\.0: 113.5 }
116+
- match: { aggregations.percentiles_double.values.95\.0: 143.5 }
117+
- match: { aggregations.percentiles_double.values.99\.0: 149.5 }
115118

116119

117120
---

‎server/src/main/java/org/elasticsearch/search/aggregations/metrics/MedianAbsoluteDeviationAggregator.java‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -92,7 +92,7 @@ public void collect(int doc, long bucket) throws IOException {
9292

9393
TDigestState valueSketch = valueSketches.get(bucket);
9494
if (valueSketch == null) {
95-
valueSketch = TDigestState.create(compression);
95+
valueSketch = TDigestState.create(compression, executionHint);
9696
valueSketches.set(bucket, valueSketch);
9797
}
9898

‎server/src/main/java/org/elasticsearch/search/aggregations/metrics/TDigestState.java‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -38,7 +38,7 @@ protected enum Type {
3838
SORTING;
3939

4040
static Type defaultValue() {
41-
return AVL_TREE;
41+
return HYBRID;
4242
}
4343

4444
static Type valueForHighAccuracy() {

‎server/src/test/java/org/elasticsearch/search/aggregations/metrics/TDigestPercentilesAggregatorTests.java‎

Lines changed: 7 additions & 7 deletions
Original file line numberDiff line numberDiff line change
@@ -83,8 +83,8 @@ public void testSomeMatchesSortedNumericDocValues() throws IOException {
8383
}, tdigest -> {
8484
assertEquals(7L, tdigest.state.size());
8585
assertEquals(7L, tdigest.state.centroidCount());
86-
assertEquals(4.5d, tdigest.percentile(75), 0.0d);
87-
assertEquals("4.5", tdigest.percentileAsString(75));
86+
assertEquals(4.0d, tdigest.percentile(75), 0.0d);
87+
assertEquals("4.0", tdigest.percentileAsString(75));
8888
assertEquals(2.0d, tdigest.percentile(50), 0.0d);
8989
assertEquals("2.0", tdigest.percentileAsString(50));
9090
assertEquals(1.0d, tdigest.percentile(22), 0.0d);
@@ -107,14 +107,14 @@ public void testSomeMatchesNumericDocValues() throws IOException {
107107
assertEquals(tdigest.state.centroidCount(), 7L);
108108
assertEquals(8.0d, tdigest.percentile(100), 0.0d);
109109
assertEquals("8.0", tdigest.percentileAsString(100));
110-
assertEquals(6.98d, tdigest.percentile(88), 0.0d);
111-
assertEquals("6.98", tdigest.percentileAsString(88));
110+
assertEquals(4.0d, tdigest.percentile(75), 0.0d);
111+
assertEquals("4.0", tdigest.percentileAsString(75));
112112
assertEquals(1.0d, tdigest.percentile(33), 0.0d);
113113
assertEquals("1.0", tdigest.percentileAsString(33));
114114
assertEquals(1.0d, tdigest.percentile(25), 0.0d);
115115
assertEquals("1.0", tdigest.percentileAsString(25));
116-
assertEquals(0.0d, tdigest.percentile(1), 0.0d);
117-
assertEquals("0.0", tdigest.percentileAsString(1));
116+
assertEquals(0.06d, tdigest.percentile(1), 0.0d);
117+
assertEquals("0.06", tdigest.percentileAsString(1));
118118
assertTrue(AggregationInspectionHelper.hasValue(tdigest));
119119
});
120120
}
@@ -135,7 +135,7 @@ public void testQueryFiltering() throws IOException {
135135
assertEquals(4L, tdigest.state.centroidCount());
136136
assertEquals(2.0d, tdigest.percentile(100), 0.0d);
137137
assertEquals(1.0d, tdigest.percentile(50), 0.0d);
138-
assertEquals(0.5d, tdigest.percentile(25), 0.0d);
138+
assertEquals(0.75d, tdigest.percentile(25), 0.0d);
139139
assertTrue(AggregationInspectionHelper.hasValue(tdigest));
140140
});
141141

‎server/src/test/java/org/elasticsearch/search/aggregations/metrics/TDigestStateTests.java‎

Lines changed: 3 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -158,6 +158,8 @@ public void testFactoryMethods() {
158158

159159
assertEquals(fast, anotherFast);
160160
assertEquals(accurate, anotherAccurate);
161+
assertNotEquals(fast, accurate);
162+
assertNotEquals(anotherFast, anotherAccurate);
161163
}
162164

163165
private static TDigestState writeToAndReadFrom(TDigestState state, TransportVersion version) throws IOException {
@@ -195,6 +197,7 @@ public void testSerialization() throws IOException {
195197
assertEquals(serialized, state);
196198

197199
TDigestState serializedBackwardsCompatible = writeToAndReadFrom(state, TransportVersion.V_8_500_000);
200+
assertNotEquals(serializedBackwardsCompatible, state);
198201
assertEquals(serializedBackwardsCompatible, backwardsCompatible);
199202
}
200203
}

0 commit comments

Comments
 (0)