#!/usr/bin/env python3
"""Recreate the five manuscript figures from the declared model and saved ledger.

These are synthetic mathematical calculations, not experimental micrographs.
Run verify.py first if results/verification.json is missing.
"""
from pathlib import Path
import json
import numpy as np
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt
from model import comb_profile, optimal_action, cubic_constant, diffusion_comb

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

def finish(fig, name):
    fig.tight_layout()
    fig.savefig(ROOT/'figures'/name,dpi=200)
    plt.close(fig)

def main():
    (ROOT/'figures').mkdir(exist_ok=True)
    ledger=ROOT/'results'/'verification.json'
    if not ledger.exists():
        raise FileNotFoundError('Run python code/verify.py before plotting.')
    report=json.loads(ledger.read_text(encoding='utf-8'))
    theta=np.linspace(0,2*np.pi,4097)
    fig=plt.figure(figsize=(7.5,4.3))
    for m in (2,4,8):
        plt.plot(theta/(2*np.pi),comb_profile(m,optimal_action(m,.5),theta),
                 label=f'{m} sequential gates')
    plt.xlabel('Position / original optical fringe period')
    plt.ylabel('Retained molecular fraction')
    plt.title('A finer pattern from sequential chemical memory')
    plt.ylim(-.025,.57); plt.legend()
    finish(fig,'comb_profiles.png')

    orders=np.arange(2,129)
    fig=plt.figure(figsize=(7.5,4.3))
    for q in (.1,.5,.9):
        dose=[m*optimal_action(int(m),q) for m in orders]
        plt.loglog(orders,dose,label=f'Midpoint fraction q = {q}')
    plt.xlabel('Pitch multiplication factor m');plt.ylabel('Minimum marking action U')
    plt.title('Exact finite-order frontier; cubic asymptotic growth');plt.legend()
    finish(fig,'dose_frontier.png')

    fig=plt.figure(figsize=(7.5,4.3))
    for q in (.1,.5,.9):
        ratios=[optimal_action(int(m),q)/m**2 for m in orders]
        plt.semilogx(orders,ratios,label=f'q = {q}; limit {cubic_constant(q):.6f}')
    plt.xlabel('Pitch multiplication factor m');plt.ylabel('Minimum marking action / m³')
    plt.title('Convergence to the sharp cubic constants');plt.legend()
    finish(fig,'cubic_constants.png')

    fig=plt.figure(figsize=(7.5,4.3))
    for d in (0.,.0001,.001):
        result=diffusion_comb(6,optimal_action(6,.5),d,n=768)
        plt.plot(result['theta']/(2*np.pi),result['survival'],label=f'Dθ = {d:g}')
    plt.xlabel('Position / original optical fringe period');plt.ylabel('Retained molecular fraction')
    plt.title('Diffusion fills dark nodes and exposes the pulse order');plt.legend()
    finish(fig,'diffusion_profiles.png')

    hazards=report['hazard_examples']
    fig=plt.figure(figsize=(7.5,4.3))
    plt.loglog([r['factor_gates'] for r in hazards],[r['error'] for r in hazards],
               'o-',label='Observed error on a 24³ grid')
    plt.loglog([r['factor_gates'] for r in hazards],[r['bound'] for r in hazards],
               's--',label='Whole-domain analytic upper bound')
    plt.xlabel('Number of patterned factor gates');plt.ylabel('Absolute conversion-fraction error')
    plt.title('General 3D compilation is constructive, not inexpensive');plt.legend()
    finish(fig,'universal_compiler_cost.png')
    print('Five figures recreated in',ROOT/'figures')

if __name__=='__main__':
    main()