#!/usr/bin/env python3
"""Recomputes every number in "The Early-Entry Reply Protocol" (2026-09 edition)
from reply-protocol-2026-09.csv. Standard library only.

    python3 reply-protocol-2026-09.py [path/to/reply-protocol-2026-09.csv]

Engagement rate per stance is pooled: total engagements / total impressions of the
replies in that stance (X's definition, applied to the group). Other per-reply
figures are plain means. Profile visits are per 1,000 impressions, pooled.
"""
import csv
import statistics
import sys

STANCES = [
    ("topic_extension", "Topic extension"),
    ("shared_views", "Shared views"),
    ("different_views", "Different views"),
]


def load(path):
    with open(path, newline="", encoding="utf-8") as f:
        rows = list(csv.DictReader(f))
    for r in rows:
        for k in ("impressions", "engagements", "likes", "reposts", "bookmarks", "profile_visits"):
            r[k] = int(r[k])
        r["view_rank_percentile"] = int(r["view_rank_percentile"]) if r["view_rank_percentile"] else None
    return rows


def summary(rows):
    imp = sum(r["impressions"] for r in rows)
    return {
        "n": len(rows),
        "impressions_total": imp,
        "impressions_mean": imp / len(rows),
        "engagement_rate": sum(r["engagements"] for r in rows) / imp,
        "likes_mean": statistics.mean(r["likes"] for r in rows),
        "bookmarks_total": sum(r["bookmarks"] for r in rows),
        "bookmarks_mean": statistics.mean(r["bookmarks"] for r in rows),
        "profile_visits_total": sum(r["profile_visits"] for r in rows),
        "profile_visits_per_1k": 1000 * sum(r["profile_visits"] for r in rows) / imp,
    }


def main():
    path = sys.argv[1] if len(sys.argv) > 1 else "reply-protocol-2026-09.csv"
    rows = load(path)
    print(f"Replies: {len(rows)} ({rows[0]['posted_at']} to {rows[-1]['posted_at']})\n")

    by = {key: summary([r for r in rows if r["stance"] == key]) for key, _ in STANCES}
    every = summary(rows)
    print(f"{'Stance':<17}{'n':>3}{'Eng. rate':>11}{'Impr./reply':>13}{'Likes':>8}{'Bookmarks':>11}{'Visits/1k':>11}")
    for key, label in STANCES:
        s = by[key]
        print(
            f"{label:<17}{s['n']:>3}{s['engagement_rate']:>10.2%}{s['impressions_mean']:>13,.0f}"
            f"{s['likes_mean']:>8.2f}{s['bookmarks_mean']:>11.2f}{s['profile_visits_per_1k']:>11.2f}"
        )
    s = every
    print(
        f"{'All replies':<17}{s['n']:>3}{s['engagement_rate']:>10.2%}{s['impressions_mean']:>13,.0f}"
        f"{s['likes_mean']:>8.2f}{s['bookmarks_mean']:>11.2f}{s['profile_visits_per_1k']:>11.2f}"
    )

    te, dv = by["topic_extension"], by["different_views"]
    print(f"\nEngagement-rate gap, topic extension vs different views: {te['engagement_rate'] / dv['engagement_rate']:.1f}x")
    print(f"Impressions gap, topic extension vs different views: {te['impressions_mean'] / dv['impressions_mean']:.1f}x")
    print(f"Bookmarks: {te['bookmarks_total']} of {every['bookmarks_total']} went to topic-extension replies")
    print(f"Impressions: {every['impressions_total']:,}; profile visits: {every['profile_visits_total']}")

    ranked = [r["view_rank_percentile"] for r in rows if r["view_rank_percentile"] is not None]
    print(f"\nThread ranking ({len(ranked)} replies with a live parent post):")
    print(f"  mean view-rank percentile: {statistics.mean(ranked):.1f}")
    print(f"  most-viewed reply in its thread: {sum(p == 100 for p in ranked)} ({sum(p == 100 for p in ranked) / len(ranked):.0%})")
    print(f"  above the thread's median reply: {sum(p > 50 for p in ranked)} ({sum(p > 50 for p in ranked) / len(ranked):.0%})")
    for key, label in STANCES:
        p = [r["view_rank_percentile"] for r in rows if r["stance"] == key and r["view_rank_percentile"] is not None]
        print(f"  {label}: {statistics.mean(p):.1f} (n={len(p)})")


if __name__ == "__main__":
    main()