"""Publish the aggregate v17.01 replay as a public Trackio snapshot."""

import argparse
import csv
import json
import os
import shutil
import tempfile
from pathlib import Path

from huggingface_hub import CommitOperationAdd, HfApi


PROJECT = "trl-active-groups"
RUN = "v17.01-replay"
METRICS = [
    "train/frac_active_groups",
    "train/frac_reward_zero_std",
    "train/frac_reward_all_zero",
    "train/frac_reward_all_one",
    "train/reward",
    "train/reward_std",
]
COUNT_KEYS = [
    "scorable_groups",
    "active_groups",
    "zero_std_groups",
    "all_zero_groups",
    "all_one_groups",
    "unscorable_groups",
]


def main():
    parser = argparse.ArgumentParser(description=__doc__)
    parser.add_argument("--space-id", required=True, help="Your public Hugging Face Space ID.")
    args = parser.parse_args()
    root = Path(__file__).resolve().parents[1]
    provenance = json.loads((root / "data/provenance.json").read_text())
    with (root / "data/v17.01-intervals.csv").open() as handle:
        intervals = list(csv.DictReader(handle))

    with tempfile.TemporaryDirectory(prefix="trl-active-groups-") as tmp:
        os.environ["TRACKIO_DIR"] = str(Path(tmp) / "trackio")
        os.environ["TRACKIO_STORAGE_MODE"] = "sqlite"
        os.environ["TRACKIO_PLOT_ORDER"] = ",".join(METRICS)
        os.environ.pop("TRACKIO_SPACE_ID", None)
        os.environ.pop("TRACKIO_SERVER_URL", None)
        import trackio
        from trackio.frontend_config import BUNDLED_FRONTEND_DIR

        trackio.init(
            project=PROJECT,
            name=RUN,
            group="Real rewards by generation model version",
            config=provenance,
            auto_log_gpu=False,
            auto_log_cpu=False,
            embed=False,
        )
        for row in intervals:
            counts = {key: int(row[key]) for key in COUNT_KEYS}
            n = counts["scorable_groups"]
            assert n > 0
            assert counts["active_groups"] + counts["zero_std_groups"] == n
            assert counts["all_zero_groups"] + counts["all_one_groups"] == counts["zero_std_groups"]
            metrics = {f"train/{key}": value for key, value in counts.items()}
            metrics.update(
                {
                    "train/frac_active_groups": counts["active_groups"] / n,
                    "train/frac_reward_zero_std": counts["zero_std_groups"] / n,
                    "train/frac_reward_all_zero": counts["all_zero_groups"] / n,
                    "train/frac_reward_all_one": counts["all_one_groups"] / n,
                    "train/reward": float(row["reward"]),
                    "train/reward_std": float(row["reward_std"]),
                }
            )
            trackio.log(metrics, step=int(row["model_version"]))
        trackio.finish()

        frontend = Path(tmp) / "frontend"
        shutil.copytree(BUNDLED_FRONTEND_DIR, frontend)
        index = frontend / "index.html"
        html = index.read_text().replace("Trackio Dashboard", "TRL active group metrics")
        banner = (
            '<div style="padding:12px 20px;background:#e8f5f1;color:#174a40;'
            'font:14px system-ui;border-bottom:1px solid #b5d9cf">'
            "<strong>v17.01 reward replay</strong> · 101,760 real groups · "
            "grouped by generation model version · "
            '<a href="https://huggingface.co/spaces/' + args.space_id + '/blob/main/README.md" '
            'target="_blank" rel="noopener">Methodology</a> · '
            '<a href="https://github.com/huggingface/trl-internal/pull/317" '
            'target="_blank" rel="noopener">PR #317</a></div>'
        )
        defaults = (
            "<script>const u=new URL(location.href);"
            'if(!u.searchParams.has("metrics")){'
            'u.searchParams.set("project",' + json.dumps(PROJECT) + ");"
            'u.searchParams.set("metrics",' + json.dumps(",".join(METRICS)) + ");"
            'history.replaceState(null,"",u);}</script>'
        )
        index.write_text(html.replace("<body>", "<body>" + banner + defaults))
        trackio.sync(
            project=PROJECT,
            space_id=args.space_id,
            bucket_id=f"{args.space_id}-bucket",
            sdk="static",
            private=False,
            frontend_dir=frontend,
        )
        paths = [root / "README.md", Path(__file__), *sorted((root / "data").glob("*"))]
        HfApi().create_commit(
            repo_id=args.space_id,
            repo_type="space",
            operations=[
                CommitOperationAdd(path_in_repo=path.relative_to(root).as_posix(), path_or_fileobj=str(path))
                for path in paths
            ],
            commit_message="Refresh group fraction demo and publish the reproducible aggregate inputs",
        )
    print(f"https://huggingface.co/spaces/{args.space_id}")


if __name__ == "__main__":
    main()