"""Generate manuscript tables and figures directly from released raw summaries."""
from pathlib import Path
import json
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt
ROOT=Path(__file__).resolve().parents[1];M=ROOT/'manuscript';F=M/'figures';F.mkdir(exist_ok=True)
def read(name):return json.loads((ROOT/'results'/name).read_text())
h=read('heredity_summary.json')
fig,ax=plt.subplots(figsize=(7.2,3.6))
labels=['Literal programs','Hereditary classes','Current signature classes']
values=[h['literal_programs'],h['hereditary_classes'],h['static_evaluation_classes']]
bars=ax.barh(labels,values)
for b,n in zip(bars,values):ax.text(n+1200,b.get_y()+b.get_height()/2,f'{n:,}',va='center',fontsize=11)
ax.set_xlim(0,98000);ax.set_xlabel('Number of distinguishable classes under each contract')
ax.invert_yaxis();ax.spines[['top','right']].set_visible(False);fig.tight_layout();fig.savefig(F/'v4_catalog.pdf');plt.close(fig)
c=read('corridor_summary.json')['aggregates'];labels={'raw':'Raw simulation','selfloop_skip':'Self-loop skipping','compiled':'Compiled exit'}
rows=[]
for key in ['raw','selfloop_skip','compiled']:
    x=c[key];rows.append(f"{labels[key]} & {x['setup_cpu']:.6f} & {x['execution_cpu']:.6f} & {x['total_cpu']:.6f} & {x['random_draws']:,} \\\\")
(M/'tables_v4_corridor.tex').write_text(r'''\begin{table}[htbp]\centering\small
\begin{tabular}{lrrrr}\toprule
Method & Setup CPU & Execution CPU & Total CPU & Draws\\\midrule
'''+ '\n'.join(rows)+r'''
\bottomrule\end{tabular}
\caption{Cold-per-batch corridor comparison. CPU values are seconds summed over the same 12,288 completed episodes per method. Setup is charged in every batch.}
\end{table}
''')
r=read('reinvestment_summary.json')['aggregate'];rows=[]
for key in ['raw','selfloop_skip','compiled']:
    x=r[key];rows.append(f"{labels[key]} & {x['mean_candidates']:,.1f} & {x['mean_best_utility']:.9f} & {x['mean_regret']:.7f} \\\\")
(M/'tables_v4_reinvestment.tex').write_text(r'''\begin{table}[htbp]\centering\small
\begin{tabular}{lrrr}\toprule
Method & Mean candidates & Mean best utility & Mean regret\\\midrule
'''+ '\n'.join(rows)+r'''
\bottomrule\end{tabular}
\caption{Setup-paid synthetic search: 20 seeds and 0.08 CPU seconds per method and seed. Regret is measured against the unit upper bound, not a proved attained optimum.}
\end{table}
''')
print('Generated one plot and two tables from recorded JSON.')