How to efficiently load custom data

When performing history request on custom data I notice that it is seemly going through all the data that is in the dataset it is searching through. The same goes for using the data in a backtest - the algorithm seems to iterate through all rows of the dataset until it gets to the date I have set as starting date. The data is sorted by date so it should easily use an algorithm to find the rows I am asking for. How can I define the custom data in a matter so it quickly finds the correct rows during a History request?

I am also unable to fetch the n latest bars through qb.History(symbol, 10), only by date qb.History(symbol, start=datetime(2021, 1, 3), end=datetime(2021, 1, 7)). 

I load my custom data like this:

class MESData(PythonData):
    def GetSource(self, config, date, isLiveMode):
        source = os.path.join(Globals.DataFolder, "custom/future/minute/mes.csv") 
        subscription_data_source = SubscriptionDataSource(source, SubscriptionTransportMedium.LocalFile, FileFormat.Csv)
        return subscription_data_source

    def Reader(self, config: SubscriptionDataConfig, line, date, isLiveMode):
        if not line.strip():
            return None
        
        data = MESData()
        data.Symbol = config.Symbol
        

        # Example Line Format:
        # Date              Open    High    Low     Close   Volume 
        # 2023-05-14 18:01	4132.25	4133.50	4131.25	4132.00	501.0
        splitted = line.split(',')
        
        # Add data
        data.EndTime = datetime.strptime(splitted[0], "%Y-%m-%d %H:%M:%S")
        data.Time = data.EndTime - timedelta(minutes=1)
        data.Value = float(splitted[4])      
        
        data["Open"] = float(splitted[1])
        data["High"] = float(splitted[2])
        data["Low"] = float(splitted[3])
        data["Close"] = float(splitted[4])
        data["Volume"] = float(splitted[5])
                
        
        return data