Merge 345263b79d into 08a71feef8

CHANGELOG.md

@@ -5,7 +5,9 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
-<!-- ## [Unreleased] -->
+## [Unreleased]
+### Added
+ - The fourth basic collector, `Histogram`, have been added. ([#10][github-10])
 
 ## [1.1.0] - 2023-11-13
 ### Added
@@ -47,6 +49,7 @@ See [README.md](README.md) for details and documentation.
 
 [github-6]: https://github.com/fredrikekre/Prometheus.jl/pull/6
 [github-7]: https://github.com/fredrikekre/Prometheus.jl/pull/7
+[github-10]: https://github.com/fredrikekre/Prometheus.jl/pull/10
 
 [Unreleased]: https://github.com/fredrikekre/Prometheus.jl/compare/v1.1.0...HEAD
 [1.1.0]: https://github.com/fredrikekre/Prometheus.jl/compare/v1.0.1...v1.1.0

docs/src/index.md

@@ -61,9 +61,9 @@ exposed.
 ## Collectors
 
 This section documents the collectors that are currently supported. This include the "basic"
-collectors ([Counter](@ref), [Gauge](@ref), [Summary](@ref)) as well as some custom
-collectors ([GCCollector](@ref), [ProcessCollector](@ref)). There is also a section on how
-to implement your own collector, see [Custom collectors](@ref).
+collectors ([Counter](@ref), [Gauge](@ref), [Histogram](@ref), [Summary](@ref)) as well as
+some custom collectors ([GCCollector](@ref), [ProcessCollector](@ref)). There is also a
+section on how to implement your own collector, see [Custom collectors](@ref).
 
 Upstream documentation:
  - <https://prometheus.io/docs/concepts/metric_types/>
@@ -111,6 +111,23 @@ Prometheus.@time
 Prometheus.@inprogress
 ```
 
+### Histogram
+
+Quoting the [upstream
+documentation](https://prometheus.io/docs/concepts/metric_types/#histogram):
+> A histogram samples observations (usually things like request durations or response sizes)
+> and counts them in configurable buckets. It also provides a sum of all observed values.
+
+#### Histogram API reference
+
+```@docs
+Prometheus.Histogram(::String, ::String; kwargs...)
+Prometheus.observe(::Prometheus.Histogram, ::Any)
+```
+```@docs; canonical=false
+Prometheus.@time
+```
+
 ### Summary
 
 Quoting the [upstream

src/Prometheus.jl

@@ -317,6 +317,119 @@ function collect!(metrics::Vector, gauge::Gauge)
 end
 
 
+##########################
+# Histogram <: Collector #
+##########################
+# https://prometheus.io/docs/instrumenting/writing_clientlibs/#histogram
+
+# A histogram SHOULD have the same default buckets as other client libraries.
+# https://github.com/prometheus/client_python/blob/d8306b7b39ed814f3ec667a7901df249cee8a956/prometheus_client/metrics.py#L565
+const DEFAULT_BUCKETS = [
+    .005, .01, .025, .05, .075, .1, .25, .5, .75, 1.0, 2.5, 5.0, 7.5, 10.0, Inf,
+]
+
+mutable struct Histogram <: Collector
+    @const metric_name::String
+    @const help::String
+    @const buckets::Vector{Float64}
+    @atomic _count::Int
+    @atomic _sum::Float64
+    @const bucket_counters::Vector{Threads.Atomic{Int}}
+
+    function Histogram(
+            metric_name::String, help::String; buckets::Vector{Float64}=DEFAULT_BUCKETS,
+            registry::Union{CollectorRegistry, Nothing}=DEFAULT_REGISTRY,
+        )
+        # Make a copy of and verify buckets
+        buckets = copy(buckets)
+        issorted(buckets) || throw(ArgumentError("buckets must be sorted"))
+        length(buckets) > 0 && buckets[end] != Inf && push!(buckets, Inf)
+        length(buckets) < 2 && throw(ArgumentError("must have at least two buckets"))
+        initial_sum = 0.0
+        initial_count = 0
+        bucket_counters = [Threads.Atomic{Int}(0) for _ in 1:length(buckets)]
+        histogram = new(
+            verify_metric_name(metric_name), help, buckets,
+            initial_count, initial_sum, bucket_counters,
+        )
+        if registry !== nothing
+            register(registry, histogram)
+        end
+        return histogram
+    end
+end
+
+"""
+    Prometheus.Histogram(name, help; buckets=DEFAULT_BUCKETS, registry=DEFAULT_REGISTRY)
+
+Construct a Histogram collector.
+
+**Arguments**
+ - `name :: String`: the name of the histogram metric.
+ - `help :: String`: the documentation for the histogram metric.
+
+**Keyword arguments**
+ - `buckets :: Vector{Float64}`: the upper bounds for the histogram buckets. The buckets
+   must be sorted. `Inf` will be added as a last bucket if not already included. The default
+   buckets are `DEFAULT_BUCKETS = $(DEFAULT_BUCKETS)`.
+ - `registry :: Prometheus.CollectorRegistry`: the registry in which to register the
+   collector. If not specified the default registry is used. Pass `registry = nothing` to
+   skip registration.
+
+**Methods**
+ - [`Prometheus.observe`](@ref): add an observation to the histogram.
+ - [`Prometheus.@time`](@ref): time a section and add the elapsed time as an observation.
+"""
+Histogram(::String, ::String; kwargs...)
+
+function metric_names(histogram::Histogram)
+    return (
+        histogram.metric_name * "_count", histogram.metric_name * "_sum",
+        histogram.metric_name,
+    )
+end
+
+"""
+    Prometheus.observe(histogram::Histogram, v)
+
+Add the observed value `v` to the histogram.
+This increases the sum and count of the histogram with `v` and `1`, respectively, and
+increments the counter for all buckets containing `v`.
+"""
+function observe(histogram::Histogram, v)
+    @atomic histogram._count += 1
+    @atomic histogram._sum += v
+    for (bucket, bucket_counter) in zip(histogram.buckets, histogram.bucket_counters)
+        # TODO: Iterate in reverse and break early
+        if v <= bucket
+            Threads.atomic_add!(bucket_counter, 1)
+        end
+    end
+    return nothing
+end
+
+function collect!(metrics::Vector, histogram::Histogram)
+    label_names = LabelNames(("le",))
+    push!(metrics,
+        Metric(
+            "histogram", histogram.metric_name, histogram.help,
+            [
+                Sample("_count", nothing, nothing, @atomic(histogram._count)),
+                Sample("_sum", nothing, nothing, @atomic(histogram._sum)),
+                (
+                    Sample(
+                        nothing, label_names,
+                        make_label_values(label_names, (histogram.buckets[i],)),
+                        histogram.bucket_counters[i][],
+                    ) for i in 1:length(histogram.buckets)
+               )...,
+            ]
+        ),
+    )
+    return metrics
+end
+
+
 ########################
 # Summary <: Collector #
 ########################
@@ -357,7 +470,8 @@ Construct a Summary collector.
    skip registration.
 
 **Methods**
- - [`Prometheus.observe`](@ref): add an observation to the summary.
+ - [`Prometheus.observe`](@ref observe(::Summary, ::Any)): add an observation to the
+   summary.
  - [`Prometheus.@time`](@ref): time a section and add the elapsed time as an observation.
 """
 Summary(::String, ::String; kwargs...)
@@ -404,8 +518,8 @@ specific action depends on the type of collector:
 
  - `collector :: Gauge`: set the value of the gauge to the elapsed time
    ([`Prometheus.set`](@ref))
- - `collector :: Summary`: add the elapsed time as an observation
-   ([`Prometheus.observe`](@ref))
+ - `collector :: Histogram` and `collector :: Summary`: add the elapsed time as an
+   observation ([`Prometheus.observe`](@ref))
 
 The expression to time, `expr`, can be a single expression (for example a function call), or
 a code block (`begin`, `let`, etc), e.g.
@@ -431,6 +545,7 @@ end
 
 at_time(gauge::Gauge, v) = set(gauge, v)
 at_time(summary::Summary, v) = observe(summary, v)
+at_time(histogram::Histogram, v) = observe(histogram, v)
 at_time(::Collector, v) = unreachable()
 
 """
@@ -760,6 +875,7 @@ end
 
 prometheus_type(::Type{Counter}) = "counter"
 prometheus_type(::Type{Gauge}) = "gauge"
+prometheus_type(::Type{Histogram}) = "histogram"
 prometheus_type(::Type{Summary}) = "summary"
 prometheus_type(::Type) = unreachable()
 
@@ -782,12 +898,34 @@ function collect!(metrics::Vector, family::Family{C}) where C
             else
                 @assert(child_samples isa Vector{Sample})
                 for child_sample in child_samples
-                    @assert(child_sample.label_values === nothing)
+                    if C === Histogram && (child_sample.label_names !== nothing) && (child_sample.label_values !== nothing)
+                        # TODO: Only allow child samples to be labeled for Histogram
+                        # collectors for now.
+                        @assert(
+                            length(child_sample.label_names.label_names) ==
+                            length(child_sample.label_values.label_values)
+                        )
+                        # TODO: Bypass construtor verifications
+                        merged_names = LabelNames((
+                            label_names.label_names...,
+                            child_sample.label_names.label_names...,
+                        ))
+                        merged_values = LabelValues((
+                            label_values.label_values...,
+                            child_sample.label_values.label_values...,
+                        ))
+                        push!(samples, Sample(child_sample.suffix, merged_names, merged_values, child_sample.value))
+                    else
+                        @assert(
+                            (child_sample.label_names  === nothing) ===
+                            (child_sample.label_values === nothing)
+                        )
                         push!(samples, Sample(child_sample.suffix, label_names, label_values, child_sample.value))
                     end
                 end
             end
         end
+    end
     # Sort samples lexicographically by the labels
     sort!(samples; by = function(x)
         labels = x.label_values

test/runtests.jl

@@ -167,6 +167,76 @@ end
           """
 end
 
+@testset "Prometheus.Histogram" begin
+    # Constructors and implicit registration
+    empty!(Prometheus.DEFAULT_REGISTRY.collectors)
+    c = Prometheus.Histogram("metric_name_histogram", "A histogram.")
+    @test c in Prometheus.DEFAULT_REGISTRY.collectors
+    r = Prometheus.CollectorRegistry()
+    c = Prometheus.Histogram("metric_name_histogram", "A histogram."; registry=r)
+    @test c in r.collectors
+    @test c.buckets == Prometheus.DEFAULT_BUCKETS
+    @test c._count == 0
+    @test c._sum == 0
+    @test all(x -> x[] == 0, c.bucket_counters)
+    @test_throws(
+        Prometheus.ArgumentError("metric name \"invalid-name\" is invalid"),
+        Prometheus.Histogram("invalid-name", "help"),
+    )
+    # Prometheus.observe(...)
+    v1 = 0.9
+    Prometheus.observe(c, v1)
+    @test c._count == 1
+    @test c._sum == v1
+    for (ub, counter, known_count) in zip(c.buckets, c.bucket_counters, [zeros(Int, 9); ones(Int, 6)])
+        @test counter[] == (v1 > ub ? 0 : 1) == known_count
+    end
+    v2 = 10v1
+    Prometheus.observe(c, v2)
+    @test c._count == 2
+    @test c._sum == v1 + v2
+    for (ub, counter, known_count) in zip(c.buckets, c.bucket_counters, [zeros(Int, 9); [1, 1, 1, 1]; [2, 2]])
+        @test counter[] == ((v2 > ub && v1 > ub) ? 0 : v2 > ub ? 1 : 2) == known_count
+    end
+    # Prometheus.collect(...)
+    r = Prometheus.CollectorRegistry()
+    buckets = [1.0, 2.0, Inf]
+    c = Prometheus.Histogram("metric_name_histogram", "A histogram."; buckets=buckets, registry=r)
+    Prometheus.observe(c, 0.5)
+    Prometheus.observe(c, 1.6)
+    metrics = Prometheus.collect(c)
+    @test length(metrics) == 1
+    metric = metrics[1]
+    @test metric.metric_name == c.metric_name
+    @test metric.help == c.help
+    @test length(metric.samples) == length(buckets) + 2
+    s1, s2 = metric.samples[1], metric.samples[2]
+    @test s1.suffix == "_count"
+    @test s2.suffix == "_sum"
+    @test s1.label_values === nothing
+    @test s2.label_values === nothing
+    @test s1.value == 2
+    @test s2.value == 0.5 + 1.6
+    for (ub, counter, sample, known_count) in zip(c.buckets, c.bucket_counters, metric.samples[3:end], [1, 2, 2])
+        @test sample.suffix === nothing
+        @test (sample.label_names::Prometheus.LabelNames{1}).label_names === (:le,)
+        @test (sample.label_values::Prometheus.LabelValues{1}).label_values == (string(ub),)
+        @test sample.value == counter[] == known_count
+    end
+    # Prometheus.expose_metric(...)
+    @test sprint(Prometheus.expose_metric, metric) ==
+          sprint(Prometheus.expose_io, r) ==
+          """
+          # HELP metric_name_histogram A histogram.
+          # TYPE metric_name_histogram histogram
+          metric_name_histogram_count 2
+          metric_name_histogram_sum 2.1
+          metric_name_histogram{le="1.0"} 1
+          metric_name_histogram{le="2.0"} 2
+          metric_name_histogram{le="Inf"} 2
+          """
+end
+
 @testset "Prometheus.LabelNames and Prometheus.LabelValues" begin
     @test_throws(
         Prometheus.ArgumentError("label name \"invalid-label\" is invalid"),
@@ -277,32 +347,46 @@ end
     @test 0.3 > gauge.value > 0.1
 end
 
-@testset "Prometheus.@time summary::Summary" begin
-    summary = Prometheus.Summary("call_time", "Time of calls"; registry=nothing)
-    Prometheus.@time summary sleep(0.1)
-    @test 0.3 > summary._sum > 0.1
-    @test summary._count == 1
-    Prometheus.@time summary let
+@testset "Prometheus.@time collector::$(Collector)" for Collector in (Prometheus.Histogram, Prometheus.Summary)
+    ishist = Collector === Prometheus.Histogram
+    buckets = [1.0, Inf]
+    collector = Collector(
+        "call_time", "Time of calls";
+        (ishist ? (; buckets=buckets) : (;))...,
+        registry=nothing,
+    )
+    Prometheus.@time collector sleep(0.1)
+    @test 0.3 > collector._sum > 0.1
+    @test collector._count == 1
+    ishist && @test (x->x[]).(collector.bucket_counters) == [1, 1]
+    Prometheus.@time collector let
         sleep(0.1)
     end
-    @test 0.4 > summary._sum > 0.2
-    @test summary._count == 2
-    Prometheus.@time summary f() = sleep(0.1)
+    @test 0.4 > collector._sum > 0.2
+    @test collector._count == 2
+    ishist && @test (x->x[]).(collector.bucket_counters) == [2, 2]
+    Prometheus.@time collector f() = sleep(0.1)
     @sync begin
         @async f()
         @async f()
     end
-    @test 0.7 > summary._sum > 0.4
-    @test summary._count == 4
-    Prometheus.@time summary function g()
+    @test 0.7 > collector._sum > 0.4
+    @test collector._count == 4
+    ishist && @test (x->x[]).(collector.bucket_counters) == [4, 4]
+    Prometheus.@time collector function g()
         sleep(0.1)
     end
     @sync begin
         @async g()
         @async g()
     end
-    @test 0.9 > summary._sum > 0.6
-    @test summary._count == 6
+    @test 0.9 > collector._sum > 0.6
+    @test collector._count == 6
+    ishist && @test (x->x[]).(collector.bucket_counters) == [6, 6]
+    if ishist
+        Prometheus.@time collector sleep(1.1)
+        @test (x->x[]).(collector.bucket_counters) == [6, 7]
+    end
 end
 
 @testset "Prometheus.@inprogress gauge::Gauge" begin
@@ -353,11 +437,71 @@ end
     @test_throws Prometheus.UnreachableError Prometheus.at_inprogress_exit(Coll())
 end
 
-@testset "Prometheus.Family{Summary}" begin
+# @testset "Prometheus.Family{$Collector}" for Collector in (Histogram, Summary)
+#     r = Prometheus.CollectorRegistry()
+#     c = Prometheus.Family{Prometheus.Histogram}(
+#         "http_request_time", "Time to process requests.",
+#         ("endpoint", "status_code");
+#         buckets = [1.,],
+#         registry = r,
+#     )
+#     @test c in r.collectors
+#     @test length(c.children) == 0
+#     # Prometheus.inc(...)
+#     l1 = ("/foo/", "200")
+#     l2 = ("/bar/", "404")
+#     @test Prometheus.labels(c, l1) === Prometheus.labels(c, l1)
+#     @test Prometheus.labels(c, l2) === Prometheus.labels(c, l2)
+#     @test length(c.children) == 2
+#     @test Prometheus.labels(c, l1)._count == 0
+#     @test Prometheus.labels(c, l1)._sum == 0
+#     @test Prometheus.labels(c, l2)._count == 0
+#     @test Prometheus.labels(c, l2)._sum == 0
+#     Prometheus.observe(Prometheus.labels(c, l1), 1.2)
+#     Prometheus.observe(Prometheus.labels(c, l2), 2.1)
+#     @test Prometheus.labels(c, l1)._count == 1
+#     @test Prometheus.labels(c, l1)._sum == 1.2
+#     @test Prometheus.labels(c, l2)._count == 1
+#     @test Prometheus.labels(c, l2)._sum == 2.1
+#     Prometheus.observe(Prometheus.labels(c, l1), 3.4)
+#     Prometheus.observe(Prometheus.labels(c, l2), 4.3)
+#     @test Prometheus.labels(c, l1)._count == 2
+#     @test Prometheus.labels(c, l1)._sum == 4.6
+#     @test Prometheus.labels(c, l2)._count == 2
+#     @test Prometheus.labels(c, l2)._sum == 6.4
+#     # Prometheus.collect(...)
+#     metrics = Prometheus.collect(c)
+#     @test length(metrics) == 1
+#     metric = metrics[1]
+#     @test metric.metric_name == c.metric_name
+#     @test metric.help == c.help
+#     @test length(metric.samples) == 4
+#     s1, s2, s3, s4 = metric.samples
+#     @test s1.label_values.label_values == s2.label_values.label_values == ("/bar/", "404")
+#     @test s3.label_values.label_values == s4.label_values.label_values == ("/foo/", "200")
+#     @test s1.value == 2   # _count
+#     @test s2.value == 6.4 # _sum
+#     @test s3.value == 2   # _count
+#     @test s4.value == 4.6 # _sum
+#     # Prometheus.expose_metric(...)
+#     @test sprint(Prometheus.expose_metric, metric) ==
+#           sprint(Prometheus.expose_io, r) ==
+#           """
+#           # HELP http_request_time Time to process requests.
+#           # TYPE http_request_time summary
+#           http_request_time_count{endpoint="/bar/",status_code="404"} 2
+#           http_request_time_sum{endpoint="/bar/",status_code="404"} 6.4
+#           http_request_time_count{endpoint="/foo/",status_code="200"} 2
+#           http_request_time_sum{endpoint="/foo/",status_code="200"} 4.6
+#           """
+# end
+
+@testset "Prometheus.Family{$Collector}" for Collector in (Prometheus.Histogram, Prometheus.Summary)
     r = Prometheus.CollectorRegistry()
-    c = Prometheus.Family{Prometheus.Summary}(
+    c = Prometheus.Family{Collector}(
         "http_request_time", "Time to process requests.",
         ("endpoint", "status_code");
+        (Collector === Prometheus.Histogram ? (; buckets = [2.0, Inf]) : (;))...,
         registry = r,
     )
     @test c in r.collectors
@@ -390,6 +534,30 @@ end
     metric = metrics[1]
     @test metric.metric_name == c.metric_name
     @test metric.help == c.help
+    if Collector === Prometheus.Histogram
+        buckets = Prometheus.labels(c, l1).buckets
+        @test length(buckets) == length(Prometheus.labels(c, l2).buckets)
+        @test length(metric.samples) == 2 * (length(buckets) + 2)
+        # _count and _sum samples
+        s1, s2, s5, s6 = metric.samples[[1, 2, 5, 6]]
+        @test s1.label_values.label_values == s2.label_values.label_values == ("/bar/", "404")
+        @test s5.label_values.label_values == s6.label_values.label_values == ("/foo/", "200")
+        @test s1.value == 2   # _count
+        @test s2.value == 6.4 # _sum
+        @test s5.value == 2   # _count
+        @test s6.value == 4.6 # _sum
+        # {le} samples
+        for (ls, subrange) in ((l1, 7:8), (l2, 3:4))
+            for (ub, counter, sample) in zip(buckets, Prometheus.labels(c, ls).bucket_counters, metric.samples[subrange])
+                @test sample.suffix === nothing
+                @test (sample.label_names::Prometheus.LabelNames{3}).label_names ===
+                      (:endpoint, :status_code, :le)
+                @test (sample.label_values::Prometheus.LabelValues{3}).label_values ==
+                      (ls..., string(ub))
+                @test sample.value == counter[]
+            end
+        end
+    else # Collector === Prometheus.Summary
         @test length(metric.samples) == 4
         s1, s2, s3, s4 = metric.samples
         @test s1.label_values.label_values == s2.label_values.label_values == ("/bar/", "404")
@@ -409,6 +577,7 @@ end
               http_request_time_count{endpoint="/foo/",status_code="200"} 2
               http_request_time_sum{endpoint="/foo/",status_code="200"} 4.6
               """
+    end
 end
 
 @testset "Label types for Prometheus.Family{C}" begin