from __future__ import annotations

from pathlib import Path
import csv
import json
import matplotlib.pyplot as plt


ROOT = Path(__file__).resolve().parents[1]


def load_csv(name):
    with (ROOT / "data" / name).open("r", encoding="utf-8") as f:
        return list(csv.DictReader(f))


def values(rows, key):
    return [float(r[key]) for r in rows]


def main():
    manifest = json.loads(
        (ROOT / "data" / "manifest.json").read_text(encoding="utf-8")
    )

    supervised = load_csv("supervised.csv")
    baseline = load_csv("baseline_frozen.csv")
    decisions = load_csv("decisions.csv")
    events = load_csv("events.csv")
    sensitivity = load_csv("phi_min_sensitivity.csv")

    out = ROOT / "figures"
    out.mkdir(exist_ok=True)

    # All time-series points are plotted; no thinning/sampling.
    t = values(supervised, "time_s")

    fault_times = sorted({
        float(r["time_s"])
        for r in events
        if r["branch"] == "supervised"
        and r["event"] in ("MOTOR_A_DESTROYED", "ENCODER_DESTROYED")
    })

    phi_min = float(manifest["phi_min"])

    # Figure 1: only CSV resource columns + event times from events.csv.
    plt.figure(figsize=(10, 4.8))
    for key, label in [
        ("rho_motor_a", "Motor A"),
        ("rho_motor_b", "Motor B"),
        ("rho_encoder", "Encoder"),
        ("rho_imu", "IMU"),
        ("rho_can1", "CAN1"),
        ("rho_can2", "CAN2"),
    ]:
        plt.plot(t, values(supervised, key), label=label)

    for tf in fault_times:
        plt.axvline(tf, linestyle="--")

    plt.xlabel("time [s]")
    plt.ylabel("resource operability")
    plt.ylim(-0.05, 1.05)
    plt.title("Resource state from supervised.csv")
    plt.legend(ncol=3)
    plt.grid()
    plt.tight_layout()
    plt.savefig(out / "resources.png", dpi=180)
    plt.close()

    # Figure 2: configuration trajectory from CSV only.
    config_ids = list(manifest["configurations"].keys())
    config_to_y = {cid: i for i, cid in enumerate(config_ids)}
    y = [config_to_y[r["config"]] for r in supervised]

    plt.figure(figsize=(10, 4.2))
    plt.step(t, y, where="post")

    for tf in fault_times:
        plt.axvline(tf, linestyle="--")

    plt.yticks(
        list(range(len(config_ids))),
        [
            f'{cid} {manifest["configurations"][cid]["label"]}'
            for cid in config_ids
        ],
    )
    plt.xlabel("time [s]")
    plt.ylabel("verified configuration")
    plt.title("Configuration trajectory from supervised.csv")
    plt.grid()
    plt.tight_layout()
    plt.savefig(out / "configuration.png", dpi=180)
    plt.close()

    # Figure 3: matched Phi comparison from two CSV logs.
    plt.figure(figsize=(10, 4.8))
    plt.plot(t, values(baseline, "Phi"), label="frozen c0")
    plt.plot(t, values(supervised, "Phi"), label="reconfiguration enabled")
    plt.axhline(phi_min, linestyle=":", label="Phi_min")

    for tf in fault_times:
        plt.axvline(tf, linestyle="--")

    plt.xlabel("time [s]")
    plt.ylabel("mission functionality Phi")
    plt.ylim(-0.02, 1.05)
    plt.title("Mission functionality from generated logs")
    plt.legend()
    plt.grid()
    plt.tight_layout()
    plt.savefig(out / "mission_comparison.png", dpi=180)
    plt.close()

    # Figure 4: matched S comparison.
    plt.figure(figsize=(10, 4.8))
    plt.plot(t, values(baseline, "S"), label="frozen c0")
    plt.plot(t, values(supervised, "S"), label="reconfiguration enabled")

    for tf in fault_times:
        plt.axvline(tf, linestyle="--")

    plt.xlabel("time [s]")
    plt.ylabel("survivability S")
    plt.ylim(-0.02, 1.05)
    plt.title("Survivability from generated logs")
    plt.legend()
    plt.grid()
    plt.tight_layout()
    plt.savefig(out / "survivability_comparison.png", dpi=180)
    plt.close()

    # Figure 5: every selector candidate from decisions.csv.
    labels = [
        f'{float(r["time_s"]):.2f}s {r["candidate"]}'
        for r in decisions
    ]
    S_values = values(decisions, "S")

    plt.figure(figsize=(10, 4.8))
    plt.bar(labels, S_values)
    plt.xlabel("selection event / candidate")
    plt.ylabel("S(c,rho)")
    plt.title("Candidate survivability from decisions.csv")
    plt.xticks(rotation=30, ha="right")
    plt.grid(axis="y")
    plt.tight_layout()
    plt.savefig(out / "candidate_S.png", dpi=180)
    plt.close()

    # Figure 6: state directly from the two execution logs.
    plt.figure(figsize=(10, 4.8))
    plt.plot(t, values(baseline, "x_k"), label="frozen c0")
    plt.plot(t, values(supervised, "x_k"), label="reconfiguration enabled")

    for tf in fault_times:
        plt.axvline(tf, linestyle="--")

    plt.xlabel("time [s]")
    plt.ylabel("plant state x")
    plt.title("Plant state from generated logs")
    plt.legend()
    plt.grid()
    plt.tight_layout()
    plt.savefig(out / "state_comparison.png", dpi=180)
    plt.close()

    # Figure 7: declared Phi_min sensitivity generated by experiment.py.
    x = values(sensitivity, "Phi_min")
    selected_phi = values(sensitivity, "selected_Phi")

    plt.figure(figsize=(10, 4.8))
    plt.plot(x, selected_phi)
    plt.xlabel("Phi_min")
    plt.ylabel("Phi of selected configuration")
    plt.title("Threshold sensitivity from phi_min_sensitivity.csv")
    plt.grid()
    plt.tight_layout()
    plt.savefig(out / "phi_min_sensitivity.png", dpi=180)
    plt.close()


if __name__ == "__main__":
    main()