from __future__ import annotations

import csv
from dataclasses import replace
import importlib.util
import math
from pathlib import Path
import subprocess
import sys

import pytest


TOOLS = Path(__file__).resolve().parents[1] / "tools/pim"


@pytest.fixture
def comparison(monkeypatch):
    directory = TOOLS / "pimcomp/compare"
    monkeypatch.syspath_prepend(str(directory))
    spec = importlib.util.spec_from_file_location("comparison_plots", directory / "generate_plots.py")
    module = importlib.util.module_from_spec(spec)
    monkeypatch.setitem(sys.modules, spec.name, module)
    spec.loader.exec_module(module)
    return module


@pytest.fixture
def comparison_csv():
    with (TOOLS / "pimcomp/compare/results_comparison.csv").open(newline="") as stream:
        reader = csv.DictReader(stream)
        return reader.fieldnames, list(reader)


def write_csv(path, fields, rows):
    with path.open("w", newline="") as stream:
        writer = csv.DictWriter(stream, fieldnames=fields)
        writer.writeheader()
        writer.writerows(rows)
    return path


@pytest.mark.parametrize("custom", [False, True])
def test_comparison_plot_cli_input_paths(comparison, tmp_path, monkeypatch, custom):
    expected = tmp_path / "custom.csv" if custom else TOOLS / "pimcomp/compare/results_comparison.csv"
    seen = []

    class InputCaptured(Exception):
        pass

    def capture(path):
        seen.append(path)
        raise InputCaptured

    monkeypatch.setattr(comparison, "load_results", capture)
    monkeypatch.setattr(comparison, "configure_style", lambda: None)
    monkeypatch.setattr(sys, "argv", ["generate_plots.py", *(["--input", str(expected)] if custom else [])])
    with pytest.raises(InputCaptured):
        comparison.main()
    assert seen == [expected]


def test_exact_comparison_configuration_matrix(comparison, comparison_csv, tmp_path):
    fields, rows = comparison_csv
    results = comparison.load_results(write_csv(tmp_path / "results.csv", fields, rows))
    keys = {(row.arch, row.model, row.mode, row.pipeline) for row in results}
    assert keys == {(arch, model, mode, stage) for arch in comparison.ARCHITECTURES
                    for model in comparison.MODELS for mode, stage in comparison.CONFIGURATIONS}


@pytest.mark.parametrize("damage,message", [("missing", "Missing"), ("duplicate", "Duplicate"),
                                             ("arch", "Unexpected"), ("model", "Unexpected"),
                                             ("pipeline", "Unexpected"), ("mode", "Unexpected")])
def test_comparison_rejects_bad_matrix_keys(comparison, comparison_csv, tmp_path, damage, message):
    fields, rows = comparison_csv
    if damage == "missing":
        rows.pop()
    elif damage == "duplicate":
        rows[-1] = dict(rows[-2])  # Still 40 rows and 20 per architecture.
    else:
        field = "raptor_pipeline" if damage == "pipeline" else damage
        rows[-1][field] = {"arch": "arch-c", "model": "other", "pipeline": "3", "mode": "other"}[damage]
    with pytest.raises(ValueError, match=message):
        comparison.load_results(write_csv(tmp_path / "results.csv", fields, rows))


def test_matrix_validation_survives_optimized_python(comparison_csv, tmp_path):
    fields, rows = comparison_csv
    path = write_csv(tmp_path / "duplicate.csv", fields, [*rows[:-1], rows[-2]])
    program = """import sys
from pathlib import Path
sys.path.insert(0, sys.argv[1])
import generate_plots
try:
    generate_plots.load_results(Path(sys.argv[2]))
except ValueError as exc:
    if 'Duplicate configuration' not in str(exc): raise
else:
    raise RuntimeError('Invalid matrix accepted with Python -O')
"""
    subprocess.run([sys.executable, "-O", "-c", program, str(TOOLS / "pimcomp/compare"), str(path)], check=True)


@pytest.mark.parametrize("field,value", [("pimcomp_latency_ms", "999"),
    ("pimcomp_throughput_samples_s", "999"), ("pimcomp_power_mw", "999"),
    ("pimcomp_energy_pj", "999"), ("pimcomp_functional_validation", "FAIL")])
def test_repeated_pimcomp_batch_results_must_agree(comparison, comparison_csv, tmp_path, field, value):
    fields, rows = comparison_csv
    row = next(row for row in rows if row["arch"] == "arch-a" and row["model"] == "vgg8"
               and row["raptor_pipeline"] == "4")
    row[field] = value
    with pytest.raises(ValueError, match="inconsistent shared PIMCOMP batch results"):
        comparison.load_results(write_csv(tmp_path / "results.csv", fields, rows))


@pytest.mark.parametrize("status,expected", [("PASS", ["1.0"]),
    ("FAIL", ["1.0", "INVALID"]), ("SKIPPED", ["1.0", "SKIPPED"]),
    ("NA", ["1.0", "NA"])])
@pytest.mark.parametrize("direction", [1, -1])
def test_value_label_validation_text(comparison, status, expected, direction):
    fig, ax = comparison.plt.subplots()
    try:
        comparison.value_label(ax, 0, 1, "1.0", status, direction)
        assert [text.get_text() for text in ax.texts] == expected
        assert ax.texts[-1].get_color() == (comparison.ERROR_COLOR if status == "FAIL" else comparison.TEXT_COLOR)
        if status != "PASS":
            value, validation = ax.texts
            assert value.xy == validation.xy
            assert value.get_position()[1] == validation.get_position()[1] == 0
            assert value.get_position()[0] == -validation.get_position()[0] != 0
    finally:
        comparison.plt.close(fig)


@pytest.mark.parametrize("values", [(1, 2), (-1, -2), (-1, 2), (0, 0)])
def test_value_label_limits_are_tight_shared_and_unclipped(comparison, values):
    fig, axes = comparison.plt.subplots(1, 2, sharey=True, figsize=(10, 4.4), layout="constrained")
    try:
        for ax, value in zip(axes, values):
            ax.set_ylim(-10, 10)
            ax.bar(0, value)
            direction = -1 if value < 0 else 1
            comparison.value_label(ax, 0, value + direction * 0.05, f"{value:+.2f}%", "FAIL", direction)
        comparison.fit_value_labels(fig)
        fig.canvas.draw()
        renderer = fig.canvas.get_renderer()
        assert axes[0].get_ylim() == axes[1].get_ylim()
        lower, upper = axes[0].get_ylim()
        assert upper - lower < 20
        assert lower <= min(0, *values) and upper >= max(0, *values)
        assert axes[0].get_window_extent(renderer).height == pytest.approx(axes[1].get_window_extent(renderer).height)
        for ax in axes:
            bounds = ax.get_window_extent(renderer)
            for text in ax.texts:
                extent = text.get_window_extent(renderer)
                assert bounds.y0 <= extent.y0 <= extent.y1 <= bounds.y1
    finally:
        comparison.plt.close(fig)


@pytest.mark.parametrize("statuses,expected", [
    (("PASS", "PASS", "PASS"), "PASS"),
    (("FAIL", "PASS", "PASS"), "FAIL"),
    (("PASS", "FAIL", "PASS"), "FAIL"),
    (("PASS", "PASS", "FAIL"), "FAIL"),
    (("PASS", "SKIPPED", "NA"), "SKIPPED"),
    (("PASS", "PASS", "NA"), "N/A"),
])
def test_compact_validation_combines_pipeline_statuses(comparison, tmp_path, monkeypatch, statuses, expected):
    rows = comparison.load_results(TOOLS / "pimcomp/compare/results_comparison.csv")
    stages = dict(zip((2, 4, 8), statuses))
    rows = [replace(row, raptor_validation=stages[row.pipeline])
            if row.arch == "arch-a" and row.model == "vgg8" and row.mode == "throughput"
            else row for row in rows]
    captured = []
    monkeypatch.setattr(comparison, "save_figure", lambda fig, name, path: captured.append((fig, name)))
    comparison.draw_validation(rows, tmp_path, compact=True)
    fig, name = captured[0]
    try:
        assert name == "validation_status_compact"
        assert fig.get_size_inches() == pytest.approx((8.5, 3.3))
        for index, ax in enumerate(fig.axes):
            assert [label.get_text() for label in ax.get_xticklabels(minor=True)] == [
                "Raptor", "PIMCOMP", "Raptor", "PIMCOMP"]
            assert [label.get_text() for label in ax.get_yticklabels()] == ([
                "VGG-8", "ResNet-18", "ResNet-34", "GoogLeNet", "YOLO11n"] if index == 0 else [])
            assert len(ax.patches) == len(ax.texts) == 4 * len(comparison.MODELS)
            assert [label.get_text() for label in ax.get_xticklabels()] == ["Latency", "Throughput"]
            assert all(text.get_fontsize() == 12 and text.get_weight() == "normal" for text in ax.texts)
            assert all(text.get_color() == "#1D2530" for text in ax.texts)
        ax = fig.axes[0]
        throughput_x = ax.xaxis.get_minorticklocs()[2]
        assert [text.get_text() for text in ax.texts if text.get_position() == (throughput_x, 0)] == [expected]
    finally:
        comparison.plt.close(fig)


@pytest.mark.parametrize("rates,stage", [((300, 200, 100), 2), ((100, 300, 200), 4),
    ((100, 200, 300), 8), ((math.nan, 200, 300), 8), ((300, 300, 200), 2),
    ((math.nan, math.nan, math.nan), 2)])
def test_compact_speedups_select_best_pipeline(comparison, rates, stage):
    template = comparison.load_results(TOOLS / "pimcomp/compare/results_comparison.csv")[0]
    rows = [replace(template, latency_speedup=0.5)]
    rows += [replace(template, mode="throughput", pipeline=depth, raptor_throughput=rate,
                     pimcomp_throughput=100, raptor_validation="FAIL")
             for depth, rate in zip((2, 4, 8), rates)]
    latency, throughput = comparison.select_compact_speedups(rows)
    assert latency is rows[0] and latency.speedup == 0.5
    assert throughput.pipeline == stage and throughput.raptor_validation == "FAIL"


@pytest.mark.parametrize("diverging", [False, True])
@pytest.mark.parametrize("metric,ylabel", [("speedup", "Speedup (x)"),
    ("energy", "Energy-per-inference advantage (x)"), ("power", "Average power advantage (x)")])
def test_compact_metric_bars_and_units(comparison, tmp_path, monkeypatch, diverging, metric, ylabel):
    from matplotlib.colors import to_rgba

    template = comparison.load_results(TOOLS / "pimcomp/compare/results_comparison.csv")[0]
    rows = [replace(template, arch=arch, mode=mode, pipeline=stage, latency_speedup=0.5,
                    energy=0.5 if mode == "latency" else 3, power=0.5 if mode == "latency" else 3,
                    raptor_throughput=300, pimcomp_throughput=100, pimcomp_validation="FAIL")
            for arch in comparison.ARCHITECTURES for mode, stage in (("latency", 1), ("throughput", 4))]
    captured = []
    monkeypatch.setattr(comparison, "save_figure", lambda fig, name, path: captured.append(fig))
    comparison.draw_metric(rows, metric, ylabel, "compact", tmp_path,
                           diverging=diverging, compact=True)
    fig = captured[0]
    try:
        assert fig.get_size_inches() == pytest.approx((8.5, 4.4))
        expected_legend = {
            "speedup": ["Single-inference latency", "Best-case throughput"],
            "energy": ["Single-inference energy", "Best-case throughput energy"],
            "power": ["Single-inference power", "Best-case throughput power"],
        }
        assert [text.get_text() for text in fig.legends[0].get_texts()] == expected_legend[metric]
        for index, ax in enumerate(fig.axes):
            latency, throughput = (container.patches[0] for container in ax.containers)
            lower, upper = ax.get_xlim()
            assert latency.get_x() - lower == pytest.approx(0.4)
            assert upper - throughput.get_x() - throughput.get_width() == pytest.approx(0.4)
            assert latency.get_x() + latency.get_width() == throughput.get_x()
            assert latency.get_facecolor() == to_rgba(comparison.METRIC_STYLES["latency"][0])
            assert throughput.get_facecolor() == to_rgba(comparison.METRIC_STYLES["throughput"][0])
            assert latency.get_edgecolor() == to_rgba(comparison.METRIC_STYLES["latency"][1])
            assert throughput.get_edgecolor() == to_rgba(comparison.METRIC_STYLES["throughput"][1])
            assert not latency.get_hatch() and throughput.get_hatch() == "//"
            assert latency.get_height() == (-1 if diverging else 0.5)
            assert throughput.get_height() == (2 if diverging else 3)
            assert ax.get_ylabel() == (ylabel if index == 0 else "")
            assert bool(ax.get_yticklabels()) == (index == 0)
            assert all("×" not in label.get_text() for label in ax.get_yticklabels())
            assert [text.get_text() for text in ax.texts] == [
                "2.00×" if diverging else "0.50×", "P. INVALID", "3.00×", "P. INVALID"]
    finally:
        comparison.plt.close(fig)


@pytest.mark.parametrize("missing_energy", [False, True])
def test_compact_metrics_use_speedup_pipeline_selection(comparison, monkeypatch, missing_energy):
    rows = comparison.load_results(TOOLS / "pimcomp/compare/results_comparison.csv")
    cases = {2: (100, 9, 8), 4: (300, math.nan if missing_energy else 0.5, 0.6), 8: (200, 7, 6)}
    rows = [replace(row, raptor_throughput=cases[row.pipeline][0],
                    energy=cases[row.pipeline][1], power=cases[row.pipeline][2])
            if row.arch == "arch-a" and row.model == "vgg8" and row.mode == "throughput"
            else row for row in rows]
    captured = []

    def capture(selected, field, ylabel, basename, output_dir, diverging=False, compact=False):
        if compact:
            captured.append((selected, field, basename, diverging))

    monkeypatch.setattr(comparison, "load_results", lambda path: rows)
    for name in ("configure_style", "draw_validation", "draw_pipeline_effect", "draw_absolute"):
        monkeypatch.setattr(comparison, name, lambda *args, **kwargs: None)
    monkeypatch.setattr(comparison, "draw_metric", capture)
    monkeypatch.setattr(sys, "argv", ["generate_plots.py"])
    comparison.main()
    assert {(field, basename, diverging) for _, field, basename, diverging in captured} == {
        (metric, f"{metric}_compact_{suffix}", diverging)
        for metric in ("speedup", "energy", "power")
        for suffix, diverging in (("comparison", False), ("diverging", True))}
    assert len(captured) == 6
    selected = captured[0][0]
    assert all(plotted is selected for plotted, _, _, _ in captured)
    throughput = next(row for row in selected if row.arch == "arch-a" and row.model == "vgg8"
                      and row.mode == "throughput")
    assert throughput.pipeline == 4 and throughput.power == 0.6
    assert math.isnan(throughput.energy) if missing_energy else throughput.energy == 0.5


def test_compiler_palette_and_grayscale_patterns(comparison, tmp_path, monkeypatch):
    from matplotlib.colors import to_rgba

    assert comparison.COMPILER_STYLES == {
        "raptor": ("#B3D5CA", "#4B7C6A"), "pimcomp": ("#E1BC8B", "#856A44")}
    assert comparison.METRIC_STYLES == {
        "latency": ("#B3CCE3", "#456B90"), "throughput": ("#B3D5CA", "#4B7C6A")}
    assert comparison.VALIDATION_STYLES["PASS"] == ("#B3D5CA", "#4B7C6A")
    assert comparison.TEXT_COLOR == "#222222"
    assert comparison.NETWORK_STYLES["yolo11n"] == ("#FF9494", "#C52020")
    assert comparison.NETWORK_STYLES["googlenet"] == ("#D1C0E0", "#77558F")
    rows = comparison.load_results(TOOLS / "pimcomp/compare/results_comparison.csv")
    selected = [replace(row, power=2 if row.arch == "arch-a" else 0.5)
                for row in rows if row.model == "vgg8" and row.mode == "latency"]
    captured = []
    monkeypatch.setattr(comparison, "save_figure", lambda fig, *args: captured.append(fig))
    comparison.draw_metric(selected, "power", "Average power advantage (x)", "power", tmp_path, diverging=True)
    comparison.draw_absolute(selected, "latency", "Single-inference latency (ms)", "latency", tmp_path)
    try:
        for ax, compiler in zip(captured[0].axes, ("raptor", "pimcomp")):
            bar = ax.containers[0].patches[0]
            assert bar.get_facecolor() == to_rgba(comparison.COMPILER_STYLES[compiler][0])
            assert bar.get_edgecolor() == to_rgba(comparison.COMPILER_STYLES[compiler][1])
            assert not bar.get_hatch()
        for ax in captured[1].axes:
            assert [container.patches[0].get_hatch() for container in ax.containers] == ["", ""]
    finally:
        for fig in captured:
            comparison.plt.close(fig)


def test_tick_rotation_changes_only_overlapping_labels(comparison):
    fig, ax = comparison.plt.subplots(figsize=(4, 3), layout="constrained")
    try:
        ax.set_xlim(-1, 6)
        ax.set_xticks([0, 0.5, 5], ["Single-inference", "Throughput", "8"])
        comparison.fit_x_tick_labels(fig)
        assert [text.get_rotation() for text in ax.get_xticklabels()] == [330, 330, 0]
        assert [text.get_ha() for text in ax.get_xticklabels()] == ["left", "left", "center"]
    finally:
        comparison.plt.close(fig)


def test_all_plot_terminology(comparison, monkeypatch):
    from matplotlib.path import Path as PlotPath
    from matplotlib.text import Text
    from matplotlib.transforms import Affine2D

    plotter = comparison
    labels = set()

    def capture(fig, name, output):
        fig.canvas.draw()
        if name != "raptor_pipeline_throughput":
            assert fig.axes[1].get_ylabel() == ""
            assert fig.axes[1].get_yticklabels() == []
        renderer = fig.canvas.get_renderer()
        for ax in fig.axes:
            for container in ax.containers:
                for bar in container.patches:
                    center = bar.get_x() + bar.get_width() / 2
                    bounds = bar.get_window_extent(renderer)
                    for text in ax.texts:
                        if hasattr(text, "xy") and text.xy[0] == center:
                            extent = text.get_window_extent(renderer)
                            assert bounds.x0 <= extent.x0 <= extent.x1 <= bounds.x1, (name, text.get_text())
        for legend in fig.legends:
            bounds = legend.get_window_extent(renderer)
            assert fig.bbox.x0 <= bounds.x0 <= bounds.x1 <= fig.bbox.x1, name
        tick_paths = []
        for ax in fig.axes:
            for minor in (False, True):
                for text in ax.get_xticklabels(minor=minor):
                    if not text.get_visible() or not text.get_text():
                        continue
                    angle = text.get_rotation()
                    assert angle in (0, 330), (name, text.get_text(), angle)
                    bounds = text.get_window_extent(renderer)
                    text.set_rotation(0)
                    unrotated = text.get_window_extent(renderer)
                    text.set_rotation(angle)
                    path = PlotPath.unit_rectangle().transformed(
                        Affine2D().scale(unrotated.width, unrotated.height).rotate_deg(angle))
                    rotated = path.get_extents()
                    path = path.transformed(Affine2D().translate(bounds.x0 - rotated.x0, bounds.y0 - rotated.y0))
                    assert not any(path.intersects_path(other, filled=True) for other in tick_paths), name
                    tick_paths.append(path)
                    if text.get_text() in comparison.MODEL_NAMES.values():
                        assert angle == 0, (name, text.get_text())
        heights = [ax.get_window_extent(renderer).height for ax in fig.axes]
        assert max(heights) - min(heights) < 1, name
        for ax in fig.axes:
            bounds = ax.get_window_extent(renderer)
            for text in ax.texts:
                extent = text.get_window_extent(renderer)
                assert bounds.y0 - 1 <= extent.y0 <= extent.y1 <= bounds.y1 + 1, (name, text.get_text())
                assert fig.bbox.x0 <= extent.x0 <= extent.x1 <= fig.bbox.x1, (name, text.get_text())
        labels.update(text.get_text() for text in fig.findobj(Text))
        plotter.plt.close(fig)

    monkeypatch.setattr(plotter, "save_figure", capture)
    monkeypatch.setattr(sys, "argv", ["generate_plots.py"])
    plotter.main()
    text = "\n".join(labels)
    for old in ("Pimcomp", "ARCHITECTURE", "VGG8", "RESNET", "GOOGLENET", "YOLO11N",
                "Single inference", "Best case", "Latency, 1 stage", "Energy per inference advantage"):
        assert old not in text
    assert {"VGG-8", "ResNet-18", "ResNet-34", "GoogLeNet"} <= labels
    assert "Arch-A" in text and "Arch-B" in text
    assert {"Raptor", "PIMCOMP", "YOLO11n", "Single-inference", "2 stages", "4 stages", "8 stages",
            "Single-inference latency (ms)", "Throughput (inferences/s)", "Energy per inference (mJ)",
            "Energy-per-inference advantage (x)", "Derived as 1 / latency",
            "Throughput measurement"} <= labels


@pytest.mark.parametrize("width,one_row", [(12, True), (4.5, False), (3, False)])
def test_legend_uses_available_width_and_balances_rows(comparison, width, one_row):
    from collections import Counter
    from matplotlib.patches import Patch

    fig = comparison.plt.figure(figsize=(width, 4), layout="constrained")
    try:
        handles = [Patch(label=f"Entry {index}") for index in range(7)]
        comparison.add_legend(fig, handles)
        fig.canvas.draw()
        renderer = fig.canvas.get_renderer()
        legend = fig.legends[0]
        counts = Counter(round(text.get_window_extent(renderer).y0, 3)
                         for text in legend.get_texts() if text.get_visible())
        assert sum(counts.values()) == len(handles)
        assert (len(counts) == 1) == one_row
        assert max(counts.values()) - min(counts.values()) <= 1
        bounds = legend.get_window_extent(renderer)
        assert 0 <= bounds.x0 <= bounds.x1 <= fig.bbox.width
    finally:
        comparison.plt.close(fig)


@pytest.mark.parametrize("diverging", [False, True])
def test_advantage_throughput_groups_touch(comparison, tmp_path, monkeypatch, diverging):
    rows = comparison.load_results(TOOLS / "pimcomp/compare/results_comparison.csv")
    selected = [row for row in rows if row.model in ("vgg8", "resnet18")]
    captured = []
    monkeypatch.setattr(comparison, "save_figure", lambda fig, *args: captured.append(fig))
    comparison.draw_metric(selected, "energy", "Energy-per-inference advantage (x)", "energy", tmp_path,
                           diverging=diverging)
    fig = captured[0]
    try:
        for ax in fig.axes:
            bars = [container.patches[0] for container in ax.containers]
            lower, upper = ax.get_xlim()
            assert bars[0].get_x() - lower == pytest.approx(0.4)
            assert upper - bars[-1].get_x() - bars[-1].get_width() == pytest.approx(0.4)
            for index, (left, right) in enumerate(zip(bars, bars[1:])):
                edge = left.get_x() + left.get_width()
                if index % 4 in (1, 2):
                    assert edge == pytest.approx(right.get_x())
                else:
                    assert edge < right.get_x()
    finally:
        comparison.plt.close(fig)


def test_grouped_absolute_plot_scheduling_labels(comparison, tmp_path, monkeypatch):
    rows = comparison.load_results(TOOLS / "pimcomp/compare/results_comparison.csv")
    captured = []
    monkeypatch.setattr(comparison, "save_figure", lambda fig, *args: captured.append(fig))
    comparison.draw_absolute(rows, "energy", "Energy per inference (mJ)", "energy", tmp_path, scale=1e-9)
    fig = captured[0]
    try:
        for ax in fig.axes:
            assert {text.get_text() for text in ax.get_xticklabels(minor=True)} == {"Latency", "Throughput"}
        labels = [text.get_text() for text in fig.legends[0].get_texts()]
        assert "1 stage (latency)" in labels and "Single-inference" not in labels
    finally:
        comparison.plt.close(fig)


def test_pipeline_plot_connects_derived_and_measured_rates(comparison, tmp_path, monkeypatch):
    rows = comparison.load_results(TOOLS / "pimcomp/compare/results_comparison.csv")
    captured = []
    monkeypatch.setattr(comparison, "save_figure", lambda fig, *args: captured.append(fig))
    comparison.draw_pipeline_effect(rows, tmp_path)
    fig = captured[0]
    try:
        assert fig.get_size_inches() == pytest.approx((13, 4.3))
        assert fig._supylabel.get_text() == "Throughput (inferences/s)"
        for index, (ax, (arch, model)) in enumerate(zip(fig.axes, ((arch, model) for arch in comparison.ARCHITECTURES
                                                                 for model in comparison.MODELS))):
            selected = [row for row in rows if row.arch == arch and row.model == model]
            measured_line, derived_marker, *measured_markers = ax.lines
            assert measured_line.get_ydata()[0] == pytest.approx(1000 / selected[0].raptor_latency)
            assert list(measured_line.get_ydata()[1:]) == [row.raptor_throughput for row in selected[1:]]
            assert derived_marker.get_marker() == "s"
            assert derived_marker.get_ydata()[0] == pytest.approx(1000 / selected[0].raptor_latency)
            assert derived_marker.get_color() == comparison.NETWORK_STYLES[model][1]
            assert all(marker.get_marker() == "o" for marker in measured_markers)
            assert measured_line.get_color() == comparison.NETWORK_STYLES[model][1]
            assert [label.get_text() for label in ax.get_xticklabels()] == (
                ["1", "2", "4", "8"] if arch == "arch-b" else [])
            assert ax.get_xlabel() == ""
            assert ax.get_title() == (comparison.MODEL_NAMES[model] if arch == "arch-a" else "")
            assert ax.title.get_weight() == "normal"
            if model == comparison.MODELS[-1]:
                assert ax.get_ylabel() == comparison.ARCH_NAMES[arch]
                assert ax.yaxis.get_label_position() == "right"
                assert ax.yaxis.label.get_rotation() == 270
                assert ax.yaxis.label.get_weight() == "bold"
            else:
                assert ax.get_ylabel() == ""
            assert all(label.get_rotation() == 0 for label in ax.get_xticklabels())
            lower, upper = ax.get_ylim()
            values = measured_line.get_ydata()
            assert 0 < lower < min(values) <= max(values) < upper
            assert ax.get_yticklabels()
            assert ax.get_ylim() == fig.axes[index % len(comparison.MODELS)].get_ylim()
        assert [text.get_text() for text in fig.legends[0].get_texts()] == [
            "Derived as 1 / latency", "Throughput measurement"]
        assert len({ax.lines[0].get_color() for ax in fig.axes[:len(comparison.MODELS)]}) == len(comparison.MODELS)
    finally:
        comparison.plt.close(fig)


def test_plotter_exports_png_svg_and_pdf(comparison, tmp_path):
    comparison.configure_style()
    rows = comparison.load_results(TOOLS / "pimcomp/compare/results_comparison.csv")
    selected = [replace(row, raptor_validation="FAIL", pimcomp_latency=math.nan)
                for row in rows if row.model == "vgg8" and row.mode == "latency"]
    comparison.draw_absolute(selected, "latency", "Latency (ms)", "comparison-smoke", tmp_path)
    comparison.draw_validation(rows, tmp_path, compact=True)
    compact_names = []
    for metric in ("speedup", "energy", "power"):
        for suffix, diverging in (("comparison", False), ("diverging", True)):
            name = f"{metric}_compact_{suffix}"
            compact_names.append(name)
            comparison.draw_metric(comparison.select_compact_speedups(rows), metric, "Advantage (x)",
                                   name, tmp_path, diverging=diverging, compact=True)
    for name in ("comparison-smoke", "validation_status_compact", *compact_names):
        assert (tmp_path / f"png/{name}.png").read_bytes().startswith(b"\x89PNG\r\n\x1a\n")
        svg = (tmp_path / f"svg/{name}.svg").read_text()
        assert "<svg" in svg
        assert (tmp_path / f"pdf/{name}.pdf").read_bytes().startswith(b"%PDF-")
    assert "N/A" in (tmp_path / "svg/comparison-smoke.svg").read_text()
    assert "P. INVALID" in (tmp_path / "svg/comparison-smoke.svg").read_text()