python array object cache

def get_cache_parquet(file_path, da):
    if os.path.isfile(file_path):
        df = pd.read_parquet(file_path)
        if df.iloc[0]['max_da'] == da: return df
    return []
    
class NumpyEncoder(json.JSONEncoder):
    def default(self, obj):
        if isinstance(obj, np.ndarray):
            return obj.tolist()
        return json.JSONEncoder.default(self, obj)
    
def fund_ind(market, order, limit):
    sql_str = "SELECT MAX(da) FROM fund_indicators"
    items = dbfetchallDict(market, sql_str, "db_fund")
    da = items[0]['max']
    file_path = 'fund_summary.parquet'
    df = get_cache_parquet(file_path, da)
    if len(df) > 0:
        nav_list = []
        is_desc = False if order == 'desc' else True
        df = df.sort_values(['score'],ascending=is_desc)
        for col in df.columns:
            if df[col].dtype == object:
                if col == 'nav':
                    nav_list = df['nav'].tolist()
                    df['nav'] = df['code']
        res = df.to_dict(orient='records')
        for i in range(len(res)):
            res[i]['nav'] = json.loads(json.dumps(nav_list[i], cls=NumpyEncoder))
        return res
    sql_str = f"""
    SELECT M.code, REPLACE(M.code, ' Equity', '') AS ticker, M.name, E.score, MaxDates.max_da, P.ar, P.sr, P.mdd, 
    ROUND(CAST(P.sr/2 AS numeric), 2) as sr,
    (
    SELECT JSON_AGG(JSON_BUILD_OBJECT('da', da::text, 'beta_equity', beta_equity))
    FROM stock_ind
    WHERE code = e.code
    AND da BETWEEN NOW() - INTERVAL '180 days' AND NOW()
    ) AS nav
    
    INNER JOIN (
	SELECT code, MAX(da) AS max_da
	FROM stock_ind_e
	WHERE code like '00%%' group by code) AS MaxDates     
	ON E.code = MaxDates.code AND E.da = MaxDates.max_da
    AND M.name IS NOT NULL
    AND E.da = (SELECT MAX(da) FROM indicators_e)
    ORDER BY E.score {order}
    LIMIT {limit};
    """
    market = "jp"
    items = dbfetchallDict(market, sql_str, "fund_ind")
    df = pd.DataFrame(items)
    df.to_parquet(file_path)
    return items