"""
Semantic code duplication: relative change since 2021 baseline.

For each project, the 2021 percentage is the baseline (index = 100).
Every later year is expressed as the percent change relative to that baseline,
stock-chart style: raw 10% -> 15% is shown as +50%, not +5 points.

The "average" line is the geometric mean of the per-project index values for
each year -- the correct center for ratio/index data, treating a project that
doubled and one that halved as symmetric.
"""

import numpy as np
import matplotlib.pyplot as plt

YEARS = [2021, 2022, 2023, 2024, 2025, 2026]

PROJECTS = {
    "gitea":        {2026: 22.38, 2025: 24.17, 2024: 24.65, 2023: 24.19, 2022: 24.39, 2021: 36.54},
    "caffeine":     {2026: 36.11, 2025: 36.00, 2024: 35.64, 2023: 35.61, 2022: 36.18, 2021: 35.74},
    "openmct":      {2026: 26.01, 2025: 26.07, 2024: 26.20, 2023: 25.52, 2022: 23.87, 2021: 17.54},
    "tusky":        {2026: 20.99, 2025: 26.54, 2024: 23.53, 2023: 24.91, 2022: 22.44, 2021: 21.89},
    "paperless-ngx":{2026: 7.49,  2025: 10.32, 2024: 10.36, 2023: 8.56,  2022: 10.84, 2021: 10.36},
    "helix":        {2026: 17.44, 2025: 16.77, 2024: 18.46, 2023: 16.18, 2022: 17.36, 2021: 13.33},
    "immich":       {2026: 24.10, 2025: 29.26, 2024: 28.79, 2023: 28.79, 2022: 15.83, 2021: 15.83},
    "caddy":        {2026: 22.53, 2025: 22.75, 2024: 21.66, 2023: 19.68, 2022: 19.18, 2021: 16.54},
    "subtitleedit": {2026: 47.16, 2025: 52.26, 2024: 52.19, 2023: 52.09, 2022: 51.72, 2021: 51.82},
    "ghostfolio":   {2026: 21.74, 2025: 19.04, 2024: 18.01, 2023: 16.70, 2022: 13.91, 2021: 14.78},
    "zulip":        {2026: 19.74, 2025: 19.23, 2024: 17.04, 2023: 15.37, 2022: 15.39, 2021: 14.39},
    "mihon":        {2026: 21.33, 2025: 20.73, 2024: 21.74, 2023: 21.87, 2022: 27.32, 2021: 30.31},
    "opencast":     {2026: 24.59, 2025: 24.57, 2024: 24.28, 2023: 23.88, 2022: 26.95, 2021: 27.17},
    "outline":      {2026: 25.59, 2025: 23.94, 2024: 22.51, 2023: 22.87, 2022: 24.29, 2021: 42.48},
}


def relative_change(project):
    """Percent change of each year relative to the project's own 2021 value."""
    baseline = project[2021]
    return [(project[y] / baseline - 1.0) * 100.0 for y in YEARS]


# Per-project relative-change series.
series = {name: relative_change(p) for name, p in PROJECTS.items()}

index_matrix = np.array(
    [[PROJECTS[name][y] / PROJECTS[name][2021] for y in YEARS] for name in PROJECTS]
)

# Geometric-mean line, computed on index values (baseline = 1.0) per year,
# then converted back to percent change. exp(mean(log(.))) over projects.
geo_index = np.exp(np.mean(np.log(index_matrix), axis=0))
geo_change = (geo_index - 1.0) * 100.0

# Arithmetic mean for comparison (plain average of index values per year).
arith_index = np.mean(index_matrix, axis=0)
arith_change = (arith_index - 1.0) * 100.0

# --- Plot ---
fig, ax = plt.subplots(figsize=(9, 5))

# Each project: same semi-transparent line. Overlaps darken naturally.
for name, vals in series.items():
    ax.plot(YEARS, vals, color="#1f4e79", alpha=0.3, linewidth=2, zorder=2)

# Average lines: stronger, primary. Geometric is the headline; arithmetic for comparison.
ax.plot(
    YEARS, geo_change,
    color="#c0392b", linewidth=3, zorder=5,
    marker="o", markersize=6,
    label="Average (geometric mean)",
)
ax.plot(
    YEARS, arith_change,
    color="#e08e0b", linewidth=2, zorder=4,
    marker="s", markersize=4, linestyle="--",
    label="Average (arithmetic mean)",
)

# Baseline reference at 0% change.
ax.axhline(0, color="#888888", linewidth=1, linestyle="--", zorder=1)

ax.set_title("Semantic code duplication: change since 2021 baseline", fontsize=13, pad=13)
ax.set_xlabel("Year")
ax.set_ylabel("Change relative to 2021 (%)")
ax.set_xticks(YEARS)
ax.yaxis.set_major_formatter(plt.FuncFormatter(lambda v, _: f"{v:+.0f}%"))
ax.grid(True, axis="y", alpha=0.25)
ax.legend(loc="best")

fig.tight_layout()
fig.savefig("chart.png", dpi=150)