David Ashpole
9d52bde6d6
Use Set hash in Distinct (2nd attempt) ( #7175 )
...
Re-opening https://github.com/open-telemetry/opentelemetry-go/pull/5028
with new benchmarks. For cases with 10 attributes, this reduces the
overhead of metric measurements by ~80-90% (depending on lock
contention). It introduces a small probability of collision for
attribute sets in the metrics SDK. For an instrument with 1 million
different attribute sets, the probability of a collision is
approximately 2 * 10^-8. For a more "normal" cardinality of 1000 on an
instrument, it is approximately 2 * 10^-17.
```
goos: linux
goarch: amd64
pkg: go.opentelemetry.io/otel/attribute
cpu: Intel(R) Xeon(R) CPU @ 2.20GHz
│ main.txt │ hash.txt │
│ sec/op │ sec/op vs base │
EquivalentMapAccess/Empty-24 32.01n ± 2% 10.12n ± 4% -68.37% (p=0.002 n=6)
EquivalentMapAccess/1_string_attribute-24 106.25n ± 2% 10.01n ± 5% -90.58% (p=0.002 n=6)
EquivalentMapAccess/10_string_attributes-24 826.250n ± 1% 9.982n ± 11% -98.79% (p=0.002 n=6)
EquivalentMapAccess/1_int_attribute-24 106.65n ± 2% 10.13n ± 3% -90.50% (p=0.002 n=6)
EquivalentMapAccess/10_int_attributes-24 833.25n ± 2% 10.04n ± 5% -98.80% (p=0.002 n=6)
geomean 190.3n 10.06n -94.72%
```
Parallel benchmarks:
```
goos: linux
goarch: amd64
pkg: go.opentelemetry.io/otel/sdk/metric
cpu: Intel(R) Xeon(R) CPU @ 2.20GHz
│ main24.txt │ new24.txt │
│ sec/op │ sec/op vs base │
SyncMeasure/NoView/Int64Counter/Attributes/0-24 288.4n ± 13% 267.0n ± 16% ~ (p=0.180 n=6)
SyncMeasure/NoView/Int64Counter/Attributes/1-24 372.7n ± 24% 303.3n ± 6% -18.61% (p=0.002 n=6)
SyncMeasure/NoView/Int64Counter/Attributes/10-24 1862.5n ± 11% 302.2n ± 6% -83.77% (p=0.002 n=6)
SyncMeasure/NoView/Float64Counter/Attributes/0-24 288.2n ± 5% 291.8n ± 14% ~ (p=0.589 n=6)
SyncMeasure/NoView/Float64Counter/Attributes/1-24 374.8n ± 22% 326.2n ± 15% -12.98% (p=0.002 n=6)
SyncMeasure/NoView/Float64Counter/Attributes/10-24 1984.0n ± 10% 277.9n ± 15% -85.99% (p=0.002 n=6)
SyncMeasure/NoView/Int64UpDownCounter/Attributes/0-24 286.8n ± 13% 279.4n ± 14% ~ (p=0.818 n=6)
SyncMeasure/NoView/Int64UpDownCounter/Attributes/1-24 415.4n ± 14% 309.5n ± 11% -25.47% (p=0.002 n=6)
SyncMeasure/NoView/Int64UpDownCounter/Attributes/10-24 1923.0n ± 19% 294.1n ± 17% -84.71% (p=0.002 n=6)
SyncMeasure/NoView/Float64UpDownCounter/Attributes/0-24 284.9n ± 5% 271.6n ± 11% ~ (p=0.240 n=6)
SyncMeasure/NoView/Float64UpDownCounter/Attributes/1-24 382.9n ± 23% 295.7n ± 13% -22.78% (p=0.002 n=6)
SyncMeasure/NoView/Float64UpDownCounter/Attributes/10-24 1787.0n ± 28% 289.2n ± 12% -83.81% (p=0.002 n=6)
SyncMeasure/NoView/Int64Histogram/Attributes/0-24 283.4n ± 8% 269.9n ± 9% ~ (p=0.589 n=6)
SyncMeasure/NoView/Int64Histogram/Attributes/1-24 300.7n ± 8% 270.1n ± 15% -10.16% (p=0.026 n=6)
SyncMeasure/NoView/Int64Histogram/Attributes/10-24 1046.8n ± 24% 299.2n ± 16% -71.42% (p=0.002 n=6)
SyncMeasure/NoView/Float64Histogram/Attributes/0-24 264.3n ± 12% 295.9n ± 5% +11.93% (p=0.026 n=6)
SyncMeasure/NoView/Float64Histogram/Attributes/1-24 321.0n ± 8% 269.4n ± 11% -16.09% (p=0.002 n=6)
SyncMeasure/NoView/Float64Histogram/Attributes/10-24 1052.2n ± 10% 274.6n ± 5% -73.90% (p=0.002 n=6)
geomean 540.0n 287.7n -46.72%
```
Single-threaded benchmarks:
```
goos: linux
goarch: amd64
pkg: go.opentelemetry.io/otel/sdk/metric
cpu: Intel(R) Xeon(R) CPU @ 2.20GHz
│ main1.txt │ new1.txt │
│ sec/op │ sec/op vs base │
SyncMeasure/NoView/Int64Counter/Attributes/0 130.95n ± 1% 97.99n ± 21% -25.17% (p=0.002 n=6)
SyncMeasure/NoView/Int64Counter/Attributes/1 300.8n ± 7% 104.6n ± 3% -65.21% (p=0.002 n=6)
SyncMeasure/NoView/Int64Counter/Attributes/10 1646.0n ± 2% 105.8n ± 2% -93.58% (p=0.002 n=6)
SyncMeasure/NoView/Float64Counter/Attributes/0 132.65n ± 1% 99.28n ± 4% -25.16% (p=0.002 n=6)
SyncMeasure/NoView/Float64Counter/Attributes/1 295.4n ± 3% 107.7n ± 3% -63.54% (p=0.002 n=6)
SyncMeasure/NoView/Float64Counter/Attributes/10 1620.0n ± 1% 109.6n ± 4% -93.23% (p=0.002 n=6)
SyncMeasure/NoView/Int64UpDownCounter/Attributes/0 132.85n ± 80% 99.34n ± 1% -25.22% (p=0.002 n=6)
SyncMeasure/NoView/Int64UpDownCounter/Attributes/1 300.4n ± 1% 106.0n ± 1% -64.71% (p=0.002 n=6)
SyncMeasure/NoView/Int64UpDownCounter/Attributes/10 1622.0n ± 1% 105.8n ± 1% -93.48% (p=0.002 n=6)
SyncMeasure/NoView/Float64UpDownCounter/Attributes/0 134.90n ± 51% 99.16n ± 4% -26.49% (p=0.002 n=6)
SyncMeasure/NoView/Float64UpDownCounter/Attributes/1 312.4n ± 34% 107.8n ± 2% -65.51% (p=0.002 n=6)
SyncMeasure/NoView/Float64UpDownCounter/Attributes/10 1613.0n ± 23% 106.1n ± 1% -93.43% (p=0.002 n=6)
SyncMeasure/NoView/Int64Histogram/Attributes/0 103.50n ± 17% 88.53n ± 1% -14.46% (p=0.002 n=6)
SyncMeasure/NoView/Int64Histogram/Attributes/1 199.50n ± 16% 95.44n ± 2% -52.16% (p=0.002 n=6)
SyncMeasure/NoView/Int64Histogram/Attributes/10 878.70n ± 2% 95.78n ± 2% -89.10% (p=0.002 n=6)
SyncMeasure/NoView/Float64Histogram/Attributes/0 108.55n ± 54% 88.45n ± 1% -18.51% (p=0.002 n=6)
SyncMeasure/NoView/Float64Histogram/Attributes/1 257.30n ± 14% 95.05n ± 2% -63.06% (p=0.002 n=6)
SyncMeasure/NoView/Float64Histogram/Attributes/10 882.70n ± 18% 96.28n ± 1% -89.09% (p=0.002 n=6)
geomean 355.2n 100.3n -71.77%
```
---------
Co-authored-by: Tyler Yahn <MrAlias@users.noreply.github.com >
Co-authored-by: Robert Pająk <pellared@hotmail.com >
2025-09-16 16:04:50 -04:00
Tyler Yahn
59563f7ae4
Do not use the user-defined empty set when comparing sets. ( #7357 )
...
Fix #7356
2025-09-15 09:28:12 -07:00
Tyler Yahn
4623d0a989
Fix minor grammatical error in Distinct docs ( #7203 )
2025-08-18 12:24:26 +02:00
Mikhail Mazurskiy
5e1c62a2d5
Modernize ( #7089 )
...
Use
https://pkg.go.dev/golang.org/x/tools/gopls/internal/analysis/modernize
to update code to new style.
---------
Co-authored-by: Flc゛ <four_leaf_clover@foxmail.com >
Co-authored-by: Damien Mathieu <42@dmathieu.com >
2025-07-29 10:19:11 +02:00
Robert Pająk
1737ab8666
docs: unify doc comments for functions returning bool ( #7064 )
...
Fixes https://github.com/open-telemetry/opentelemetry-go/issues/7063
If also fixes Go Doc comment for `SpanID.IsEmpty`.
The pattern is based on the way the Go standard library documents
functions returning a boolean.
2025-07-23 07:58:50 +02:00
Lijingfeng
a7e83aace9
simplify conversions from slice to array ( #5818 )
...
Go 1.17 introduced an expression for convertion from slice to array, so
codes can be simplified.
Ref:
https://tip.golang.org/ref/spec#Conversions_from_slice_to_array_or_array_pointer
> Converting a slice to an array yields an array containing the elements
of the underlying array of the slice.
2024-09-16 09:17:59 +02:00
Mikhail Mazurskiy
1fe2f03854
Deprecate Sortable ( #4734 )
...
* Deprecate Sortable
* Remove redundant checks
* Move code to a non-deprecated func
* Apply suggestions from code review
Co-authored-by: Tyler Yahn <MrAlias@users.noreply.github.com >
* Mention individual deprecated function
* Update attribute/set.go
Co-authored-by: Robert Pająk <pellared@hotmail.com >
* Add BenchmarkNewSet
---------
Co-authored-by: Tyler Yahn <MrAlias@users.noreply.github.com >
Co-authored-by: Robert Pająk <pellared@hotmail.com >
2024-03-11 08:31:19 -07:00
Tyler Yahn
fe9bab54b7
Use Distinct instead of Set for map keys ( #5027 )
2024-03-06 11:11:16 +01:00
Robert Pająk
7dea232a46
[chore] Simplify the license header ( #4987 )
2024-02-29 07:05:28 +01:00
Alex Boten
8778c38832
docs: minor update to docstring ( #4833 )
...
Signed-off-by: Alex Boten <aboten@lightstep.com >
2024-01-19 15:18:16 +01:00
Tyler Yahn
deddec38ac
Optimize (attribute.Set).Filter for no filtered case ( #4774 )
...
* Optimize Set.Filter for no filtered case
When all elements of the Set are kept during a call to Filter, do not
allocate a new Set and the dropped attributes slice. Instead, return the
immutable Set and nil.
To achieve this the functionality of filterSet is broken down into a
more generic filteredToFront function.
* Apply suggestions from code review
Co-authored-by: Robert Pająk <pellared@hotmail.com >
* Rename run to benchFn based on review feedback
---------
Co-authored-by: Robert Pająk <pellared@hotmail.com >
2024-01-08 07:49:45 -08:00
Tyler Yahn
69611bd01a
Switch Stream back to having an AttributeFilter field and add New*Filter functions ( #4444 )
...
* Add allow/deny attr filters
* Revert "Replace `Stream.AttributeFilter` with `AllowAttributeKeys` (#4288 )"
This reverts commit 1633c74aeacheung.zhy.csu@gmail.com >
2023-08-25 07:39:43 -07:00
Tyler Yahn
a25f4fe7ef
Guard zero value Set methods ( #3957 )
...
Co-authored-by: Chester Cheung <cheung.zhy.csu@gmail.com >
2023-04-04 08:06:10 -07:00
Tyler Yahn
b62eb2ca88
Pool sortables used to create attribute sets ( #3832 )
...
* Pool sortables used to create attribute sets
* Move sync pool to attribute pkg
* Add change to changelog
* Fix comment
* Apply suggestions from code review
Co-authored-by: Peter Liu <lpfvip2008@gmail.com >
* Update sdk/metric/instrument.go
Co-authored-by: Robert Pająk <pellared@hotmail.com >
* Update comment based on feedback
* Apply feedback
---------
Co-authored-by: Peter Liu <lpfvip2008@gmail.com >
Co-authored-by: Robert Pająk <pellared@hotmail.com >
2023-03-13 11:19:28 -07:00
Tyler Yahn
1f5b159161
Use already enabled revive linter and add depguard ( #2883 )
...
* Refactor golangci-lint conf
Order settings alphabetically.
* Add revive settings to golangci conf
* Check blank imports
* Check bool-literal-in-expr
* Check constant-logical-expr
* Check context-as-argument
* Check context-key-type
* Check deep-exit
* Check defer
* Check dot-imports
* Check duplicated-imports
* Check early-return
* Check empty-block
* Check empty-lines
* Check error-naming
* Check error-return
* Check error-strings
* Check errorf
* Stop ignoring context first arg in tests
* Check exported comments
* Check flag-parameter
* Check identical branches
* Check if-return
* Check increment-decrement
* Check indent-error-flow
* Check deny list of go imports
* Check import shadowing
* Check package comments
* Check range
* Check range val in closure
* Check range val address
* Check redefines builtin id
* Check string-format
* Check struct tag
* Check superfluous else
* Check time equal
* Check var naming
* Check var declaration
* Check unconditional recursion
* Check unexported return
* Check unhandled errors
* Check unnecessary stmt
* Check unnecessary break
* Check waitgroup by value
* Exclude deep-exit check in example*_test.go files
2022-05-19 15:15:07 -05:00
Tyler Yahn
a8ea3dbb46
Replace use of old term label with attribute ( #2790 )
...
* Replace use of old term label with attribute
The specification has unified on the term attribute to describe
key-value pairs. There exist still many hold-overs of the use of the
term label. This updates those uses or deprecates exported types and
functions in favor of renamed forms.
* fix infinite recursion
* Remove backticks from attribute set docs
* Remove LabelFilterSelector entirely
* Remove Metadata.Labels instead of deprecate
* Update changelog with public changes
* Revert OC err msg
2022-04-18 07:31:31 -07:00
Aaron Clawson
f24f52aa06
Debug Logging for sdk/trace ( #2500 )
...
* Debug Logging for sdk/trace
* Fixes spelling, adds marshaling to attribute sets
Co-authored-by: Aaron Clawson <MadVikingGod@users.noreply.github.com >
Co-authored-by: Anthony Mirabella <a9@aneurysm9.com >
2022-01-10 19:58:01 -05:00
Gustavo Silva Paiva
2bd4840c30
remove Set.Encoded(Encoder) enconding cache ( #1855 )
...
* remove Set.Encoded(Encoder) enconding cache
Co-authored-by: Tyler Yahn <MrAlias@users.noreply.github.com >
2021-04-29 11:28:04 -07:00
Punya Biswal
ecf65d7968
Rename otel/label -> otel/attribute ( #1541 )
...
* Rename otel/label -> otel/attr
Leave the imported name alone, to avoid a large diff and conflicts
* Better import comment
* Update CHANGELOG.md
Co-authored-by: Tyler Yahn <MrAlias@users.noreply.github.com >
* otel/attr -> otel/attribute
* Missed the changelog entry
* Get rid of import renaming
* Merge remaining conflicts
Co-authored-by: Tyler Yahn <MrAlias@users.noreply.github.com >
Co-authored-by: Anthony Mirabella <a9@aneurysm9.com >
2021-02-18 12:59:37 -05:00
