algo/v0/collatz_k_scripts/collatz_k_fusion.py

# -*- coding: utf-8 -*-
"""
collatz_k_fusion.py

Construction de clauses de fusion (F) sur un ensemble de classes modulo 2^m.

Sorties:
  - CSV exhaustif
  - Markdown audit (résumé + impact par état)
"""

from __future__ import annotations
from pathlib import Path
import csv
from collections import Counter
from typing import Iterable, Dict, Tuple

from collatz_k_core import prefix_data, fusion_choice_a, delta_F, Nf_F, preimage_m
from collatz_k_utils import write_text


def build_fusion_clauses(
    residues: Iterable[int],
    t: int,
    state_id_of_residue4096: Dict[int, int],
    state_mot7: Dict[int, str],
    out_md: str,
    out_csv: str,
    modulus_power: int,
) -> Tuple[int, int]:
    residues = list(residues)
    hits = []

    A_dist = Counter()
    a_dist = Counter()
    Nf_dist = Counter()
    ymod_dist = Counter()
    tot_state = Counter()
    hit_state = Counter()

    for n0 in residues:
        sid = state_id_of_residue4096[n0 % 4096]
        tot_state[sid] += 1

        pref = prefix_data(n0, t)
        a = fusion_choice_a(pref.y)
        if a is None:
            continue
        dF = delta_F(pref.A, t, a)
        if dF <= 0:
            continue
        Nf = Nf_F(pref.C, pref.A, t, a)
        if n0 < Nf:
            continue
        m = preimage_m(pref.y, a)
        if m >= n0:
            continue

        hit_state[sid] += 1
        A_dist[pref.A] += 1
        a_dist[a] += 1
        Nf_dist[Nf] += 1
        ymod_dist[pref.y % 3] += 1

        hits.append(
            {
                "classe_mod_2^m": n0,
                "m": modulus_power,
                "t": t,
                "a": a,
                "A_t": pref.A,
                "mot_a0..": " ".join(map(str, pref.word)),
                "C_t": pref.C,
                "y": pref.y,
                "y_mod_3": pref.y % 3,
                "DeltaF": dF,
                "Nf": Nf,
                "preimage_m": m,
                "etat_id": sid,
                "base_mod_4096": n0 % 4096,
            }
        )

    out_csv_p = Path(out_csv)
    out_csv_p.parent.mkdir(parents=True, exist_ok=True)
    with out_csv_p.open("w", newline="", encoding="utf-8") as f:
        if hits:
            w = csv.DictWriter(f, fieldnames=list(hits[0].keys()))
            w.writeheader()
            for row in hits:
                w.writerow(row)
        else:
            f.write("")

    lines = []
    lines.append(f"# Clauses de fusion F(t={t}) au palier 2^{modulus_power}")
    lines.append("")
    lines.append("## Introduction")
    lines.append("")
    lines.append("Audit de clauses de fusion (F) sur U (impairs → impairs).")
    lines.append("")
    lines.append("## Résultats globaux")
    lines.append("")
    lines.append(f"- Taille du domaine analysé : {len(residues)}")
    lines.append(f"- Clauses valides : {len(hits)}")
    frac = (len(hits) / len(residues)) if residues else 0.0
    lines.append(f"- Fraction couverte : {frac}")
    lines.append("")
    lines.append("Distribution de y mod 3 :")
    for k in sorted(ymod_dist):
        lines.append(f"- y mod 3 = {k} : {ymod_dist[k]}")
    lines.append("")
    lines.append("Distribution de a :")
    for k in sorted(a_dist):
        lines.append(f"- a = {k} : {a_dist[k]}")
    lines.append("")
    lines.append("Distribution de A_t :")
    for k in sorted(A_dist):
        lines.append(f"- A_t = {k} : {A_dist[k]}")
    lines.append("")
    lines.append("Distribution de N_F (premières valeurs) :")
    for k in sorted(Nf_dist)[:40]:
        lines.append(f"- N_F = {k} : {Nf_dist[k]}")
    if len(Nf_dist) > 40:
        lines.append(f"- (… {len(Nf_dist)-40} valeurs supplémentaires)")
    lines.append("")
    lines.append("## Impact par état (60 états base B12)")
    lines.append("")
    header = ["état_id", "mot_7", "effectif", "touchés", "fraction"]
    lines.append("| " + " | ".join(header) + " |")
    lines.append("| " + " | ".join(["---"] * len(header)) + " |")
    for sid in range(1, 61):
        eff = tot_state.get(sid, 0)
        hit = hit_state.get(sid, 0)
        frac_s = (hit / eff) if eff else 0.0
        mot = state_mot7.get(sid, "")
        lines.append(f"| {sid} | {mot} | {eff} | {hit} | {frac_s} |")
    lines.append("")
    lines.append("## Fichier exhaustif")
    lines.append("")
    lines.append(f"- {out_csv_p.name}")

    write_text(out_md, "\n".join(lines) + "\n")
    return len(hits), len(residues)