def download_afc_files()

in afa/app/app.py [0:0]

• 53 lines of code
• 4 McCabe index (conditional complexity)

def download_afc_files():
    """
    """

    df = state["report"]["data"]["df"]

    status_dict = parse_s3_json(state.report["afc"]["status_json_s3_path"])
    s3_export_path = status_dict["s3_export_path"]
    prefix = status_dict["prefix"]
    horiz = state["report"]["afc"]["horiz"]
    freq = state["report"]["afc"]["freq"]

    preds_s3_prefix = \
        f'{s3_export_path}/{prefix}/{prefix}_processed.csv'
    results_s3_prefix = \
        f'{s3_export_path}/{prefix}/accuracy-metrics-values/Accuracy_{prefix}_*.csv'
    backtests_s3_prefix = \
        f'{s3_export_path}/{prefix}/forecasted-values/Forecasts_{prefix}_BacktestExportJob_*.csv'

    _df_preds = wr.s3.read_csv(preds_s3_prefix,
        dtype={"channel": str, "family": str, "item_id": str})

    _preds = []
    for _, dd in _df_preds.groupby(["channel", "family", "item_id"], as_index=False, sort=False):
        dd.sort_values(by="timestamp", ascending=True, inplace=True)
        if dd.shape[0] > horiz:
            dd = dd.iloc[1:,:]
        _preds.append(dd)

    df_preds = pd.concat(_preds)
    df_preds["type"] = "fcast"
    df_preds["timestamp"] = pd.DatetimeIndex(df_preds["timestamp"])


    df_actual = state["report"]["data"].get("df2", None)

    if df_actual is None:
        df_actual = get_df_resampled(df, freq)

    df_preds = df_preds.append(
                df_actual
                .reset_index()
                .rename({"index": "timestamp"}, axis=1)
                .assign(type='actual'))

    df_preds["channel"] = df_preds["channel"].str.upper()
    df_preds["family"] = df_preds["family"].str.upper()
    df_preds["item_id"] = df_preds["item_id"].str.upper()

    freq = FREQ_MAP_PD[state.report["afc"]["freq"]]

    df_results = wr.s3.read_csv(results_s3_prefix,
        dtype={"channel": str, "family": str, "item_id": str})

    df_results[["channel", "family", "item_id"]] = \
        df_results["item_id"].str.split("@@", expand=True)

    df_backtests = \
        wr.s3.read_csv(backtests_s3_prefix)

    df_backtests[["channel", "family", "item_id"]] = \
        df_backtests["item_id"].str.split("@@", expand=True)
    df_backtests["timestamp"] = pd.DatetimeIndex(df_backtests["backtestwindow_end_time"])
    df_backtests["p10"] = np.clip(df_backtests["p10"], 0, None)
    df_backtests["demand"] = np.round(np.clip(df_backtests["p50"], 0, None), 0)
    df_backtests["target_value"] = df_backtests["target_value"].round(0)

    df_backtests = df_backtests[["timestamp", "channel", "family", "item_id",
                                 "demand", "p10", "p90", "target_value"]]
    df_backtests.sort_values(by=["channel", "family", "item_id", "timestamp"],
                             inplace=True)

    return df_preds, df_results, df_backtests