#!/usr/bin/env python3
"""Regenerate _similarity.md: papers ranked by shared idea engagement.

Two papers are similar when their stance edges (first-mention/argues-for/
argues-against/nuances/tests) target overlapping idea sets. High similarity
in this map rarely means redundancy — it surfaces twins (same program),
rivals (fighting over the same claims), and founder-successor pairs.

    python3 _scripts/make_similarity.py
"""
import glob
import itertools
import os
import re

VAULT = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
MIN_SHARED = 3      # pairs table
MIN_IDEAS_NN = 4    # nearest-neighbor section: skip papers with tiny sets
TOP_PAIRS = 40

def main() -> None:
    sets = {}
    for p in sorted(glob.glob(os.path.join(VAULT, "papers", "*.md"))):
        stem = os.path.splitext(os.path.basename(p))[0]
        s = set()
        for line in open(p, encoding="utf-8"):
            m = re.match(r"^(first-mention|argues-for|argues-against|nuances|tests)::\s*\[\[([^\]]+)\]\]", line)
            if m:
                s.add(m.group(2))
        if len(s) >= 2:
            sets[stem] = s

    pairs = []
    for a, b in itertools.combinations(sorted(sets), 2):
        inter = sets[a] & sets[b]
        if len(inter) >= MIN_SHARED:
            jac = len(inter) / len(sets[a] | sets[b])
            pairs.append((jac, len(inter), a, b, inter))
    pairs.sort(key=lambda x: (-x[0], -x[1], x[2]))

    # nearest neighbor per paper (>=1 shared idea)
    nn = {}
    for a in sets:
        best = None
        for b in sets:
            if a == b:
                continue
            inter = sets[a] & sets[b]
            if not inter:
                continue
            jac = len(inter) / len(sets[a] | sets[b])
            if best is None or (jac, len(inter)) > (best[0], best[1]):
                best = (jac, len(inter), b)
        if best and len(sets[a]) >= MIN_IDEAS_NN:
            nn[a] = best

    lines = [
        "---", "tags: [meta]", "---", "",
        "# Paper similarity — shared idea engagement",
        "",
        "Generated by `_scripts/make_similarity.py` from stance edges; do not edit.",
        "Jaccard = shared ideas / union of both papers' idea sets. High similarity",
        "here usually means twins (same program), rivals (same disputed claims), or",
        "founder–successor pairs — not redundancy.",
        "",
        f"## Top pairs (≥{MIN_SHARED} shared ideas)",
        "",
        "| J | Shared | Paper | Paper | Shared ideas |",
        "|---|---|---|---|---|",
    ]
    for jac, n, a, b, inter in pairs[:TOP_PAIRS]:
        ideas = ", ".join(f"[[{i}]]" for i in sorted(inter))
        lines.append(f"| {jac:.2f} | {n} | [[{a}]] | [[{b}]] | {ideas} |")

    lines += [
        "",
        f"## Nearest neighbor per paper (papers with ≥{MIN_IDEAS_NN} idea edges)",
        "",
        "| Paper | Nearest neighbor | J | Shared |",
        "|---|---|---|---|",
    ]
    for a in sorted(nn, key=lambda x: (-nn[x][0], x)):
        jac, n, b = nn[a]
        lines.append(f"| [[{a}]] | [[{b}]] | {jac:.2f} | {n} |")

    out = os.path.join(VAULT, "_similarity.md")
    open(out, "w", encoding="utf-8").write("\n".join(lines) + "\n")
    print(f"wrote {min(len(pairs), TOP_PAIRS)} pairs (+{len(nn)} nearest-neighbor rows) to _similarity.md")

if __name__ == "__main__":
    main()