import matplotlib import pandas as pd import pytest matplotlib.use("Agg") from shinka.plots import ( plot_generation_runtime_timeline, plot_normalized_occupancy_over_time, ) from shinka.plots.plot_throughput import ( _compute_occupancy_series, _prepare_pool_runtime_data, ) def _runtime_df() -> pd.DataFrame: return pd.DataFrame( [ { "id": "job-a-main", "source_job_id": "job-a", "is_island_copy": False, "correct": True, "combined_score": 0.9, "timestamp": 10, "generation": 1, "patch_name": "patch-a", "model_name": "model-a", "timeline_lane_mode": "pool_slots", "pipeline_started_at": 0, "sampling_started_at": 0, "sampling_finished_at": 2, "evaluation_started_at": 2, "evaluation_finished_at": 6, "postprocess_started_at": 6, "postprocess_finished_at": 8, "sampling_worker_id": 1, "evaluation_worker_id": 1, "postprocess_worker_id": 1, "sampling_worker_capacity": 2, "evaluation_worker_capacity": 2, "postprocess_worker_capacity": 1, }, { "id": "job-a-copy", "source_job_id": "job-a", "is_island_copy": True, "correct": False, "combined_score": 0.1, "timestamp": 11, "generation": 1, "patch_name": "patch-a-copy", "model_name": "model-a", "timeline_lane_mode": "pool_slots", "pipeline_started_at": 0, "sampling_started_at": 0, "sampling_finished_at": 2, "evaluation_started_at": 2, "evaluation_finished_at": 6, "postprocess_started_at": 6, "postprocess_finished_at": 8, "sampling_worker_id": 1, "evaluation_worker_id": 1, "postprocess_worker_id": 1, "sampling_worker_capacity": 2, "evaluation_worker_capacity": 2, "postprocess_worker_capacity": 1, }, { "id": "job-b", "source_job_id": "job-b", "is_island_copy": False, "correct": True, "combined_score": 0.8, "timestamp": 20, "generation": 2, "patch_name": "patch-b", "model_name": "model-b", "timeline_lane_mode": "pool_slots", "pipeline_started_at": 1, "sampling_started_at": 1, "sampling_finished_at": 3, "evaluation_started_at": 4, "evaluation_finished_at": 8, "postprocess_started_at": 8, "postprocess_finished_at": 10, "sampling_worker_id": 2, "evaluation_worker_id": 2, "postprocess_worker_id": 1, "sampling_worker_capacity": 2, "evaluation_worker_capacity": 2, "postprocess_worker_capacity": 1, }, { "id": "job-missing", "source_job_id": "job-missing", "is_island_copy": False, "correct": True, "combined_score": 0.7, "timestamp": 30, "generation": 3, "patch_name": "patch-missing", "model_name": "model-c", "timeline_lane_mode": "pool_slots", "pipeline_started_at": 2, "sampling_started_at": 2, "sampling_finished_at": 4, "evaluation_started_at": 5, "evaluation_finished_at": 9, "postprocess_started_at": 9, "postprocess_finished_at": None, "sampling_worker_id": 1, "evaluation_worker_id": 1, "postprocess_worker_id": 1, "sampling_worker_capacity": 2, "evaluation_worker_capacity": 2, "postprocess_worker_capacity": 1, }, ] ) def test_prepare_pool_runtime_data_dedupes_rows_and_computes_capacities(): prepared = _prepare_pool_runtime_data(_runtime_df()) assert prepared is not None assert prepared.capacities == {"sampling": 2, "evaluation": 2, "postprocess": 1} assert prepared.lane_labels == [ "Sampling W1", "Evaluation W1", "Postprocess W1", "Sampling W2", "Evaluation W2", ] assert list(prepared.rows["id"]) == ["job-a-main", "job-b"] assert prepared.peaks == {"sampling": 2, "evaluation": 2, "postprocess": 1} def test_prepare_pool_runtime_data_handles_missing_optional_columns(): runtime_df = _runtime_df().drop( columns=["source_job_id", "is_island_copy", "patch_name", "model_name"] ) prepared = _prepare_pool_runtime_data(runtime_df) assert prepared is not None assert list(prepared.rows["id"]) == ["job-a-main", "job-a-copy", "job-b"] assert list(prepared.rows["source_job_id"]) == ["job-a-main", "job-a-copy", "job-b"] assert list(prepared.rows["patch_name"]) == ["unnamed", "unnamed", "unnamed"] assert list(prepared.rows["model_name"]) == ["N/A", "N/A", "N/A"] assert prepared.capacities == {"sampling": 2, "evaluation": 2, "postprocess": 1} def test_compute_occupancy_series_matches_expected_utilization_stats(): prepared = _prepare_pool_runtime_data(_runtime_df()) assert prepared is not None series = _compute_occupancy_series( prepared.rows, start_key="evaluation_started_at", end_key="evaluation_finished_at", capacity=prepared.capacities["evaluation"], ) assert series is not None assert series.total_duration == pytest.approx(6.0) assert series.avg_occupied == pytest.approx(8.0 / 6.0) assert series.utilization_pct == pytest.approx((8.0 / 12.0) * 100.0) assert series.full_occupancy_pct == pytest.approx((2.0 / 6.0) * 100.0) assert series.idle_pct == pytest.approx(0.0) def test_plot_generation_runtime_timeline_uses_deduped_pool_rows(): fig, ax = plot_generation_runtime_timeline(_runtime_df(), title="Runtime Timeline") assert fig is not None assert ax is not None assert [tick.get_text() for tick in ax.get_yticklabels()] == [ "Sampling W1", "Evaluation W1", "Postprocess W1", "Sampling W2", "Evaluation W2", ] assert len(ax.patches) == 6 assert {text.get_text() for text in ax.get_legend().get_texts()} == { "Sampling", "Evaluation", "Postprocess", } assert ax.get_legend()._ncols == 3 assert ax.get_legend()._loc == 9 assert {text.get_fontsize() for text in ax.get_legend().get_texts()} == {10.0} assert ax.get_legend().get_bbox_to_anchor()._bbox.y0 < 0 def test_plot_normalized_occupancy_over_time_adds_reference_line(): fig, ax = plot_normalized_occupancy_over_time( _runtime_df(), title="Normalized Occupancy" ) assert fig is not None assert ax is not None labels = [line.get_label() for line in ax.lines] assert labels == [ "Sampling Occupancy", "Evaluation Occupancy", "Postprocess Occupancy", "100% Capacity", ] assert ax.get_ylim()[1] >= 100 assert list(ax.lines[-1].get_ydata()) == [100, 100] assert ax.get_legend()._ncols == 2 assert ax.get_legend()._loc == 9 assert {text.get_fontsize() for text in ax.get_legend().get_texts()} == {10.0} assert ax.get_legend().get_bbox_to_anchor()._bbox.y0 < 0