%i:!SrSSKrSSKJs Jr SSKrSSKrSSK r SSK J r J r SSK r \RRSS5 \RRSS5 \RRSS 5 \RRS S 5 \RRS S 5 \RRS S 5 \RRSS 5 \RRSS5 SSKJr \"5 SSKJr SSKJrJrJrJr \ R2R4r\R8R;\R8R=\5S5r S\!S\"S-4Sjr#S\!S\"SS4Sjr$S\!S\ S\!4Sjr%S*S\&S \&4S!jjr'S"\&S\(\&4S#jr)S$\&S%\(\&S"\&4S&jr*\ RVS'5r,S(r-S)r.g)+a5Performance regression tests for rating recompute. Usage: # Run benchmarks and save baseline: pytest tests/test_performance_recompute.py --benchmark-only --benchmark-json=tests/benchmark_baseline.json # Run benchmarks and compare against saved baseline: pytest tests/test_performance_recompute.py --benchmark-only --benchmark-compare=tests/benchmark_baseline.json # Run tests without benchmarking: pytest tests/test_performance_recompute.py -m "not slow" # Run including benchmarks: pytest tests/test_performance_recompute.py -v N)date timedelta ELO_SCALE_Cz400.0 ELO_K_BASEz24.0INITIAL_RATINGz1500.0 ENABLE_DRIVERfalseENABLE_TRAINERENABLE_ADJUSTMENTS ENABLE_RD TAB_MOCK_MODEtrue)reload_settings)recompute_ratings)HorseRepositoryMeetingRepositoryRaceRepositoryStarterRepositoryzbenchmark_baseline.jsonkeyreturnc[RR[5(dg[ [5n[ R "U5nSSS5 WRS/5nUH>nURSS5nXP:XdMURS05RSS5s $ g!,(df  Ne=f![ R[[4a gf=f)zLoad a benchmark baseline value from JSON file if it exists. Reads the pytest-benchmark JSON output format and finds the mean execution time for the benchmark with the given test name. N benchmarksnamestatsmean) ospathexistsBENCHMARK_BASELINE_FILEopenjsonloadgetJSONDecodeErrorKeyError TypeError)rfdatarbrs E/root/tipsharks/tipsharks-elo-api/tests/test_performance_recompute.py_load_baseliner,9s 77>>1 2 2  ) *a99Q:CC#"C# test_name current_meanc [U5nUbUS:aX- nSoCU:oU(d[R"SU4SX445S[R"5;d[R "U5(a[R "U5OS[R "U5S.-n[R"SUS US -S S US -S S US- S-SS3 5S-SU0-n[[R"U55eS=pTggg)a4Check that the current benchmark result is within 10% of the saved baseline. Args: test_name: Name of the test function (matches the "name" field in baseline JSON) current_mean: Current mean execution time in seconds Raises: AssertionError: If current run is more than 10% slower than baseline Nrg?<z%(py0)s < %(py3)sslowdown_ratiopy0py3z$Performance regression detected for z: i.2fzms vs baseline zms (dz.1fz% slower, limit is 10%) >assert %(py5)spy5) r, @pytest_ar_call_reprcompare @py_builtinslocals_should_repr_global_name _saferepr_format_assertmsgAssertionError_format_explanation)r-r.baseliner3 @py_assert2 @py_assert1 @py_format4 @py_format6s r+_check_benchmark_baselinerJNsi(H1 %0 $ $    ~   6     ~    ~        29+Rd"3'x$s6KL!#s*3//F H      !- meeting_id meeting_datevenuecV[R"UUUR5USS.5$)NH)meetingrrcategory)rupsert isoformat)sessionrLrMrNs r+_create_test_meetingrVes2  # #! **,   rK race_numberdistancec lSUS--n[R"UURUUSSSUSS3S.5$) N mobilepacez 2025-05-06T02dz :00:00+12:00)rWrX start_typegaitadvertised_start_string)rrSid)rUrQrWrXhours r+_create_test_racerdqsM b !D   & ")4T#Jl'K    rKcountc/n[U5H3nSU-n[R"XSU35 URU5 M5 U$)N(#Horse_)rangerrSappend)rUre horse_idsihorse_ids r+_create_test_horsesrnsJI 5\!8wF1#,?" rKrace_idrkc SSKnURU[U[U555n[ US5H.upg[ R "UUSUS- 3USUS- 3UUS.US9 M0 g)z;Create test starters with sequential placings (1 = winner).rNr8rhrg)rrm horse_name runner_numberbarrier)placing)randomsampleminlen enumeraterrS)rUrorkreruselectedrlrms r+_create_test_startersr{sx}}YE3y>(BCH 1-     D 12$ &x$&78!"   .rKc ## SnSnSnSn[SSS5n/SQn[X5nUR5 [U5HnU[ US-S 9-n [ US U3XU[ U5-5n UR5 [U5H?n [X U S-S U S --5n UR5 [X RXs5 MA M UR5 Un U[ US-S 9-nX4v g 7f)zCreate 75 races with starters for benchmark. Uses the same db_session as the test, so data is visible. All changes are rolled back after the test via db_session fixture. rZ2ir8)Auckland Christchurch WellingtonHamiltonDunedin)days perf_meeting_r9N) rrnflushrirrVrxrdr{rb) db_session num_meetingsraces_per_meetinghorses_per_race total_horses base_datevenuesrkmrMrQrrace date_fromdate_tos r+fixture_races_datarsLOLT1a I NF$J=I<  9!a%#88 & -s+\!c&k/;R  ()A$Z!a%SAQRD     !*ggy R*!I))9::G  sC:C<cTUup4U"[XU5nSoeU:ow(d[R"SU4SXV45S[R"5;d[R "U5(a[R "U5OS[R "U5S.-n[R"S5S-SU0-n [[R"U 55eS =pv[US 5(a:URb,URRS 5n U b [S U 5 g g g g ) a3Benchmark recompute_ratings on a fixed dataset of ~75 races. This test uses pytest-benchmark to measure execution time. The benchmark is run multiple times and statistics are collected. After the benchmark runs, compares the result against a saved baseline if one exists in tests/benchmark_baseline.json. r>z%(py0)s > %(py3)sresultr40recompute_ratings should create rating snapshotsr:r;Nrrtest_recompute_performance)rr<r=r>r?r@rArBrCrDhasattrrr$rJ) benchmarkrrrrrrFrGrHrImean_vals r+rrs,I (* IFIA:IIII6IIIIII6IIII6IIIAIIIIIIIIIIIy'""y'B??&&v.   %&BH M (C"rKcUup#[R"5n[XU5n[R"5U- nSouU:o(d[R"SU4SXW45S[ R "5;d[R"U5(a[R"U5OS[R"U5S.-n [R"S5S-SU 0-n [[R"U 55eS =pS nXg:o(d[R"S U4S Xg45S [ R "5;d[R"U5(a[R"U5OS [R"U5S.-n [R"SUSS35S-SU 0-n [[R"U 55eS =pg )zAssert recompute completes within a reasonable time threshold. This test runs recompute once and measures wall-clock time. It serves as a quick sanity check that the operation stays under the 5-second threshold for 50-100 races. rrrrr4rr:r;Ng@r0r2elapsedzRecompute took r7z#s, expected < 5.0s for 50-100 races) time perf_counterrr<r=r>r?r@rArBrCrD) rrrrstartrrrFrGrHrIs r+!test_recompute_baseline_thresholdrss,I    E zg >F!E)GIA:IIII6IIIIII6IIII6IIIAIIIIIIIIIIIJ JJ8I8IJJJCI6JJ1I1IJJJ@I JJJ@I JJJ8I8IJ  %HIJJJ6I6IJJrK)r)/__doc__builtinsr>_pytest.assertion.rewrite assertionrewriter<r"rrdatetimerrpytestenviron setdefaultpackages.core.common.settingsrpackages.core.ratings.recomputer"packages.core.storage.repositoriesrrrrmarkslow pytestmarkrjoindirname__file__r strfloatr,rJrVintrdlistrnr{fixturerrrrKr+rs  $  mW- lF+ &1 ow/ &0 *G4 k7+ ov.:=[[   '',,GGOOH8 *  E d . c  c  S C S  C DI c .$$XN,JrK