"""Project 109: likes by feature for #gamedev posts, compared WITHIN follower bands. Reads posts.json, writes counts.json.
Main metric: median likes, plus share of posts with 5+ likes. Also an author-balanced check (each author counts once:
their median), so a few accounts posting 20x a day can't drive the result."""
import json, re, statistics as st
from collections import Counter, defaultdict
from datetime import datetime

import sys
METRIC = sys.argv[1] if len(sys.argv) > 1 else "others"  # "others" = likes minus the 7 mass-likers (masslikes.py)
D = json.load(open("posts.json", encoding="utf-8")); P = D["posts"]
BANDS = ["<100", "100-999", "1k-9.9k", "10k+"]
def band(f): return "<100" if f < 100 else "100-999" if f < 1000 else "1k-9.9k" if f < 10000 else "10k+"
def ntags(p): return len(p["tags"])
def tagband(n): return "1-2 tags" if n <= 2 else "3-5 tags" if n <= 5 else "6+ tags"
def hour(p): return datetime.fromisoformat(p["time"].replace("Z", "+00:00")).hour
def hblock(p):
    h = hour(p); return f"{h//6*6:02d}-{h//6*6+5:02d} UTC"
def steam(p): return any("store.steampowered.com" in l or "s.team" in l for l in p["links"])
def itch(p): return any("itch.io" in l for l in p["links"])
def lang(p): return (p["langs"] or ["?"])[0][:2]
def length(p):
    n = len(p["text"]); return "short <100" if n < 100 else "mid 100-250" if n <= 250 else "long >250"
WISH = re.compile(r"\bwishlist", re.I)
AI = re.compile(r"\b(ai|genai|gen ai|ai-generated|midjourney|stable diffusion|chatgpt|llm)\b", re.I)
NOAI = re.compile(r"(#?no ?ai|#noai|not ai|ai[- ]free|human[- ]made|#humanmade|handmade)", re.I)
Q = re.compile(r"\?")

FEATS = {
    "media": lambda p: p["media"],
    "hashtags": lambda p: tagband(ntags(p)),
    "time posted (UTC)": hblock,
    "length": length,
    "Steam link": lambda p: "yes" if steam(p) else "no",
    "itch.io link": lambda p: "yes" if itch(p) else "no",
    "says 'wishlist'": lambda p: "yes" if WISH.search(p["text"]) else "no",
    "asks a question (?)": lambda p: "yes" if Q.search(p["text"]) else "no",
    "says it's human-made / no AI": lambda p: "yes" if NOAI.search(p["text"]) else "no",
    "language": lambda p: lang(p) if lang(p) in ("en", "ja", "es", "pt", "fr", "de") else "other",
    "bot label": lambda p: "yes" if p["bot"] else "no",
}

def stats(s):
    if not s: return None
    L = [p[METRIC] for p in s]
    return {"n": len(s), "median": st.median(L), "p5": round(sum(x >= 5 for x in L) / len(L), 3),
            "zero": round(sum(x == 0 for x in L) / len(L), 3), "mean": round(st.mean(L), 1)}

def author_median(s):
    by = defaultdict(list)
    for p in s: by[p["did"]].append(p[METRIC])
    return {"authors": len(by), "median_of_author_medians": st.median([st.median(v) for v in by.values()]) if by else None}

out = {"metric": METRIC, "window": D["window"], "n": len(P), "authors": len({p["did"] for p in P}), "base": {}, "features": {}}
print(f"N={len(P)} posts, {out['authors']} authors, window {D['window']}")
for b in BANDS:
    s = [p for p in P if band(p["followers"]) == b]; out["base"][b] = (stats(s) or {}) | author_median(s)
    print(b, out["base"][b])
c = Counter(p["handle"] for p in P); out["top_posters"] = c.most_common(10)
print("top posters", c.most_common(10))
print("share of posts from top 10 accounts", round(sum(n for _, n in c.most_common(10)) / len(P), 3))
for name, fn in FEATS.items():
    vals = sorted({fn(p) for p in P})
    out["features"][name] = {}
    print(f"\n== {name}")
    for v in vals:
        row = {"all": stats([p for p in P if fn(p) == v]) | author_median([p for p in P if fn(p) == v])}
        for b in BANDS:
            row[b] = stats([p for p in P if fn(p) == v and band(p["followers"]) == b])
        out["features"][name][v] = row
        cells = "  ".join(f"{b}: med {row[b]['median']:g} 5+ {row[b]['p5']:.0%} (n{row[b]['n']})" if row[b] else f"{b}: -" for b in BANDS)
        print(f"  {v:14} all n{row['all']['n']:4} med {row['all']['median']:g} authmed {row['all']['median_of_author_medians']:g} 5+ {row['all']['p5']:.0%} | {cells}")
json.dump(out, open(f"counts_{METRIC}.json", "w", encoding="utf-8"), indent=1, ensure_ascii=False)