fix: rework testing framework

Started as a small tweak `leaves` -> `BenchmarkTools.leaves` (symbol no
longer exported)

Then CI wouldn't pass and now we're here.

On the plus side, the new system:
- use ReTestItems
- Actually creates a fake Example project
- more or less cleans up after it's self

Project.toml

@@ -20,8 +20,19 @@ BSON = "0.3"
 BenchmarkTools = "1"
 CodecZlib = "0.7"
 JSON = "0.21"
+Pkg = "1.9"
 Revise = "3"
+Statistics = "1.9"
 julia = "1.6"
 
 [extras]
+Pkg = "44cfe95a-1eb2-52ea-b672-e2afdf69b78f"
+Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
+ReTestItems = "817f1d60-ba6b-4fd5-9520-3cf149f6a823"
 Revise = "295af30f-e4ad-537b-8983-00126c2a3abe"
+TOML = "fa267f1f-6049-4f14-aa54-33bafae1ed76"
+Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
+UUIDs = "cf7118a7-6976-5b1a-9a39-7adc72f591a4"
+
+[targets]
+test = ["Test", "ReTestItems", "Revise", "UUIDs", "TOML", "Random"]

test/backward_compat_test.jl

@@ -0,0 +1,49 @@
+@testitem "compat" setup=[ExamplePkg] begin
+    using Test
+    using JSON
+    using PkgJogger
+    using CodecZlib
+    using BenchmarkTools
+    using Dates
+
+    # Get Benchmarking results
+    jogger, cleanup = ExamplePkg.create_jogger()
+    b = jogger.benchmark()
+
+    # Save using JSON
+    function save_benchmarks(filename, results::BenchmarkTools.BenchmarkGroup)
+        # Collect system information to save
+        mkpath(dirname(filename))
+        out = Dict(
+            "julia" => PkgJogger.julia_info(),
+            "system" => PkgJogger.system_info(),
+            "datetime" => string(Dates.now()),
+            "benchmarks" => results,
+            "git" => PkgJogger.git_info(filename),
+        )
+
+        # Write benchmark to disk
+        open(GzipCompressorStream, filename, "w") do io
+            JSON.print(io, out)
+        end
+    end
+
+    @testset "Compat *.json.gz" begin
+        f = tempname(; cleanup=false) * ".json.gz"
+        finalizer(rm, f)
+        save_benchmarks(f, b)
+
+        # Check that the deprecated warming is logged
+        local b2
+        b2 = @test_logs (:warn, r"Legacy `\*\.json\.gz` format is deprecated.*") begin
+            jogger.load_benchmarks(f)
+        end
+
+        # Check that benchmarks are still there
+        @test b2 isa Dict
+        @test haskey(b2, "benchmarks")
+        @test b2["benchmarks"] isa BenchmarkTools.BenchmarkGroup
+    end
+
+    cleanup()
+end

test/ci_test.jl

@@ -0,0 +1,11 @@
+@testitem "ci" setup=[ExamplePkg] begin
+    pkgjogger_env = Base.active_project()
+    jogger, cleanup = ExamplePkg.create_jogger()
+    dir = abspath(joinpath(jogger.BENCHMARK_DIR, ".."))
+    env = copy(ENV)
+    env["JULIA_LOAD_PATH"] = join(["@", "@stdlib", pkgjogger_env], Sys.iswindows() ? ";" : ":")
+    cmd = Cmd(`$(Base.julia_cmd()) --startup-file=no -e 'using PkgJogger; PkgJogger.ci()'`; env, dir)
+    p = run(cmd; wait=true)
+    @test success(p)
+    cleanup()
+end

test/judging.jl → test/judge_tests.jl

@@ -1,14 +1,13 @@
+@testitem "judging" setup=[ExamplePkg, BenchmarkTests] begin
 using PkgJogger
 using BenchmarkTools
-using Example
-@jog Example
 
-include("utils.jl")
+jogger, cleanup = ExamplePkg.create_jogger()
 
 function gen_example()
-    results = JogExample.benchmark()
-    filename = JogExample.save_benchmarks(results)
-    dict = JogExample.load_benchmarks(filename)
+    results = jogger.benchmark()
+    filename = jogger.save_benchmarks(results)
+    dict = jogger.load_benchmarks(filename)
     uuid = get_uuid(filename)
     return results, filename, dict, uuid
 end
@@ -25,7 +24,7 @@ new = gen_example()
 old = gen_example()
 
 @testset "JogPkgName.judge($(typeof(n)), $(typeof(o)))" for (n, o) in Iterators.product(new, old)
-    test_judge(JogExample.judge, n, o)
+    test_judge(jogger.judge, n, o)
 end
 
 @testset "PkgJogger.judge($(typeof(n)), $(typeof(o)))" for (n, o) in Iterators.product(new[1:3], old[1:3])
@@ -35,24 +34,25 @@ end
 @testset "Missing Results - $(typeof(n))" for n in new
     @testset "Empty Suite" begin
         # Expect an empty judgement
-        judgement = test_judge(JogExample.judge, n, BenchmarkGroup())
+        judgement = test_judge(jogger.judge, n, BenchmarkGroup())
         isempty(judgement)
     end
     @testset "Missing Benchmark Judgement" begin
         # Get a suite of results to modify
         ref = deepcopy(first(new))
-        ref_leaves = first.(leaves(ref))
+        ref_leaves = first.(BenchmarkTools.leaves(ref))
 
         # Add a new Trial results
-        name, trial = first(leaves(ref))
+        name, trial = first(BenchmarkTools.leaves(ref))
         name[end] = rand()
         ref[name] = deepcopy(trial)
 
         # Expect the extra benchmark to be skipped
-        judgement = test_judge(JogExample.judge, n, ref)
-        judgement_leaves = first.(leaves(judgement))
+        judgement = test_judge(jogger.judge, n, ref)
+        judgement_leaves = first.(BenchmarkTools.leaves(judgement))
         @test Set(judgement_leaves) == Set(ref_leaves)
     end
 end
+cleanup()
 
-cleanup_example()
+end