#!/usr/bin/env python3
"""
Compares DEVStone benchmark results from multiple DEVS variants
(Classic, Parallel, Dynamic Structure, etc.).

Each CSV must contain:
  model_type, depth, width, run, creation_time, setup_time, simulation_time

Produces:
  - Plots of scalability (depth/width)
  - Overhead composition
  - Speedup ratios
  - A summary CSV table with average speedups
"""

import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
from pathlib import Path

# === CONFIGURATION ===
FILES = {
    # "Classic": "./Results-PyPDEVS/DEVSTONE-classic.csv",
    "Parallel": "./Results-PyPDEVS/DEVSTONE.csv",
    # "Minimal": "./Results-PyPDEVS/DEVSTONE-minimal.csv",
    "Dynamic Structure": "./Results-PyPDEVS/DEVSTONE-dynamic.csv",
    # "DS-C": "./Results-PyPDEVS/DEVSTONE-classic-dynamic.csv",

}
OUTPUT_DIR = Path("./Results-PyPDEVS")
OUTPUT_DIR.mkdir(exist_ok=True)

# === LOAD & MERGE ===
dfs = []
for variant, path in FILES.items():
    df = pd.read_csv(path)
    df["variant"] = variant
    dfs.append(df)
data = pd.concat(dfs, ignore_index=True)

# === PROCESSING ===
data["total_time"] = data["creation_time"] + data["setup_time"] + data["simulation_time"]
agg = (
    data.groupby(["variant", "model_type", "depth", "width"], as_index=False)
    .agg({
        "creation_time": ["mean", "std"],
        "setup_time": ["mean", "std"],
        "simulation_time": ["mean", "std"],
        "total_time": ["mean", "std"]
    })
)
# Flatten multi-level columns
agg.columns = ["_".join(col).rstrip("_") for col in agg.columns.values]

sns.set_theme(style="whitegrid", font_scale=1.2)

# === Scalability with Width ===
for model in agg["model_type"].unique():
    subset = agg[agg["model_type"] == model]
    plt.figure(figsize=(7, 5))
    sns.lineplot(data=subset, x="width", y="simulation_time_mean", hue="variant", marker="o")
    plt.title(f"Scalability with Width – {model}")
    plt.xlabel("Width (# submodels per level)")
    plt.ylabel("Simulation time (s)")
    plt.tight_layout()
    plt.savefig(OUTPUT_DIR / f"scalability_width_{model}.png")
    plt.close()

# === Scalability with Depth ===
for model in agg["model_type"].unique():
    subset = agg[agg["model_type"] == model]
    plt.figure(figsize=(7, 5))
    sns.lineplot(data=subset, x="depth", y="simulation_time_mean", hue="variant", marker="o")
    plt.title(f"Scalability with Depth – {model}")
    plt.xlabel("Depth (levels)")
    plt.ylabel("Simulation time (s)")
    plt.tight_layout()
    plt.savefig(OUTPUT_DIR / f"scalability_depth_{model}.png")
    plt.close()

# === Overhead Composition ===
comp = data.melt(
    id_vars=["variant", "model_type", "depth", "width"],
    value_vars=["creation_time", "setup_time", "simulation_time"],
    var_name="phase",
    value_name="time"
)
plt.figure(figsize=(9, 6))
sns.barplot(data=comp, x="variant", y="time", hue="phase", estimator="mean", errorbar="sd")
plt.title("Average Overhead Composition by Variant")
plt.ylabel("Mean time (s)")
plt.tight_layout()
plt.savefig(OUTPUT_DIR / "overhead_composition.png")
plt.close()

print(f"\nAnalysis complete. Plots and summary saved in: {OUTPUT_DIR}/")