import argparse
import time
import pandas as pd
import plotly.express as px
from google_play_scraper import search, app
from gp_common import Run, checked_run

def collect(queries, fetch_search=search, fetch_app=app, root="outputs/gp_runs"):
    run = Run("publisher-source", settings={"queries": queries, "lang": "ja", "country": "jp", "n_hits": 30}, root=root)
    pairs, states, details = [], [], []
    for q in queries:
        try:
            items = fetch_search(q, lang="ja", country="jp", n_hits=30)
            pairs.extend((q, str(i["appId"])) for i in items if i.get("appId"))
            states.append((q, "response_nonempty" if items else "empty_response", len(items), ""))
        except Exception as exc:
            states.append((q, "failed", 0, type(exc).__name__ + ": " + str(exc)))
        time.sleep(0.5)
    for app_id in sorted({pair[1] for pair in pairs}):
        try:
            d = fetch_app(app_id, lang="ja", country="jp")
            details.append((app_id, "success", d.get("developerId"), d.get("developer"), d.get("score"), d.get("reviews"), d.get("minInstalls"), d.get("genreId"), d.get("genre"), ""))
        except Exception as exc:
            details.append((app_id, "failed", None, None, None, None, None, None, None, type(exc).__name__ + ": " + str(exc)))
        time.sleep(0.5)
    run.csv("query_apps.csv", pd.DataFrame(pairs, columns=["query", "appId"]).drop_duplicates())
    run.csv("search_status.csv", pd.DataFrame(states, columns=["query", "status", "returned_rows", "error"]))
    run.csv("app_details.csv", pd.DataFrame(details, columns=["appId", "status", "developerId", "developer", "score", "reviews", "minInstalls", "genreId", "genre", "error"]))
    return run.finish(unique_target_apps=len(details), search_scope="responses only; exhaustive coverage not verified", partial_failure=any(s[1] == "failed" for s in states) or any(d[1] == "failed" for d in details))

def aggregate(folder, root="outputs/gp_runs"):
    folder, meta = checked_run(folder)
    pairs = pd.read_csv(folder / "query_apps.csv", dtype=str, keep_default_na=False)
    d = pd.read_csv(folder / "app_details.csv", dtype=str, keep_default_na=False)
    if d.appId.duplicated().any() or set(pairs.appId) != set(d.appId):
        raise ValueError("App IDの対応・重複を確認してください。")
    good = d[d.status.eq("success")].copy()
    for col in ["score", "reviews", "minInstalls"]:
        good[col] = pd.to_numeric(good[col], errors="coerce")
    good["publisher"] = ["id:" + i if i else "name:" + n if n else "unknown_app:" + a for i, n, a in zip(good.developerId, good.developer, good.appId)]
    groups = good.groupby("publisher").agg(app_count=("appId", "nunique"), score_valid=("score", "count"),
        score_mean=("score", "mean"), reviews_valid=("reviews", "count"), reviews_sum=("reviews", lambda s: s.sum(min_count=1)),
        installs_valid=("minInstalls", "count"), minimum_installs_sum=("minInstalls", lambda s: s.sum(min_count=1))).reset_index()
    run = Run("publisher-aggregates", folder / "metadata.json", {"source_run": meta["run_id"], "publisher_identity": "developerId else exact name; unknown kept per app"}, root)
    run.csv("publishers.csv", groups)
    good["genreId"] = good["genreId"].replace("", "unknown")
    good["genre"] = good["genre"].replace("", "unknown")
    genres = good.groupby(["publisher", "genreId", "genre"]).agg(app_count=("appId", "nunique")).reset_index()
    run.csv("publisher_genres.csv", genres)
    if not groups.empty:
        run.figure("publishers.html", px.bar(groups, x="app_count", y="publisher", orientation="h", title="検索で取得できた範囲のアプリ数", hover_data=list(groups.columns)))
        run.figure("publisher_genres.html", px.bar(genres, x="publisher", y="app_count", color="genre", title="取得できたアプリの運営会社・ストアジャンル別件数", hover_data=["genreId"]))
    else:
        run.meta["skipped"]["publishers.html"] = "詳細取得成功0件"
    return run.finish(query_count=len(meta["settings"]["queries"]), unique_target_apps=len(d), details_success=len(good),
                      aggregation_apps=int(groups.app_count.sum()), not_market_share=True)

if __name__ == "__main__":
    p = argparse.ArgumentParser(); p.add_argument("mode", choices=["collect", "aggregate"]); p.add_argument("values", nargs="+")
    a = p.parse_args()
    collect(a.values) if a.mode == "collect" else aggregate(a.values[0])