"""Project 101: reply rates by feature, within follower bands. Reads posts.json, writes features.json + prints tables.
'answered' = at least one top-level reply by someone other than the author (self-threads don't count)."""
import json, re
from collections import defaultdict
D = json.load(open("posts.json", encoding="utf-8")); P = D["posts"]
def band(f): return "<100" if f < 100 else "100-999" if f < 1000 else "1000+"
def feats(p):
    t = p["text"]; tl = t.lower()
    tags = len(re.findall(r"#\w+", t))
    return {
        "question": "?" in t,
        "asks_reader": bool(re.search(r"\b(you|your|anyone|what do|how do you|would you|thoughts|feedback|which)\b", tl)) and "?" in t,
        "image/video": p["image"], "link": p["link"],
        "number": bool(re.search(r"\d", t)),
        "first person story (I/my)": bool(re.search(r"\b(i|i'm|i've|my)\b", tl)),
        "money ($/MRR/revenue)": bool(re.search(r"\$|mrr|revenue|sales|paying|customers", tl)),
        "launch/promo words": bool(re.search(r"\b(launch|launched|check out|try it|sign up|waitlist|link in|free trial)\b", tl)),
        "mentions AI": bool(re.search(r"\b(ai|llm|gpt|claude|agent)\b", tl)),
        "5+ hashtags": tags >= 5,
        "short (<120 chars)": len(t) < 120, "long (>250 chars)": len(t) > 250,
        "bot label": p["bot"],
    }
ok = [p for p in P if p.get("others") is not None]
for p in ok: p["answered"] = p["others"] > 0
out = {"n": len(ok), "base": {}, "features": {}}
print(f"posts with known repliers: {len(ok)} of {len(P)}")
for b in ["<100", "100-999", "1000+"]:
    s = [p for p in ok if band(p["followers"]) == b]
    r = sum(p["answered"] for p in s) / max(1, len(s))
    out["base"][b] = {"n": len(s), "answered": round(r, 3)}
    print(f"band {b:8} n={len(s):3} answered {r:.0%}  median likes {sorted(p['likes'] for p in s)[len(s)//2] if s else 0}")
allr = sum(p["answered"] for p in ok) / len(ok); print(f"ALL answered {allr:.0%}")
names = list(feats(ok[0]).keys())
print(f"\n{'feature':28} {'with: n / answered':>20} {'without':>16}  per band with-vs-without")
for f in names:
    w = [p for p in ok if feats(p)[f]]; wo = [p for p in ok if not feats(p)[f]]
    rate = lambda s: sum(p["answered"] for p in s) / len(s) if s else float("nan")
    per = []
    for b in ["<100", "100-999", "1000+"]:
        bw = [p for p in w if band(p["followers"]) == b]; bwo = [p for p in wo if band(p["followers"]) == b]
        per.append(f"{b}:{rate(bw):.0%}({len(bw)})/{rate(bwo):.0%}({len(bwo)})" if bw and bwo else f"{b}:-")
    out["features"][f] = {"with_n": len(w), "with": round(rate(w), 3) if w else None, "without_n": len(wo), "without": round(rate(wo), 3) if wo else None, "bands": per}
    print(f"{f:28} {len(w):4} / {rate(w):4.0%}      {len(wo):4} / {rate(wo):4.0%}   {' '.join(per)}")
# who answers: bots among repliers
reps = [r for p in ok for r in p.get("other_replies", [])]
print(f"\nreplies read: {len(reps)}, by bot-labelled accounts: {sum(r['bot'] for r in reps)}")
# concentration: authors posting many times
from collections import Counter
c = Counter(p["handle"] for p in P)
print("most active authors:", c.most_common(6))
json.dump(out, open("features.json", "w"), indent=1)