RuntimeError: Indicator "OBV(C,V)" errored with exception: input array type is not double

[ghost]

Hello,

I am having the following problem with my code

from backtesting import Backtest, Strategy
from talib import ADX, PLUS_DI, MINUS_DI,OBV,SMA
from backtesting.test import GOOG,EURUSD

class ADX_Strategy(Strategy):

    def init(self):
        self.High = self.data.High
        self.Close = self.data.Close
        self.Low = self.data.Low
        self.Volume = self.data.Volume

        self.ADX = self.I(ADX, self.High,self.Low,self.Close)
        self.PLUS_DI = self.I(PLUS_DI,self.High,self.Low,self.Close)
        self.MINUS_DI = self.I(MINUS_DI,self.High,self.Low,self.Close)
        self.OBV = self.I(OBV,self.Close,self.Volume)

    def next(self):

        print("Test")
        pass`

And it is throwing the following exception in the run() call:

ipython-input-40-d7cee23cbdc2> in init(self)
     10         self.PLUS_DI = self.I(PLUS_DI,self.High,self.Low,self.Close)
     11         self.MINUS_DI = self.I(MINUS_DI,self.High,self.Low,self.Close)
---> 12         self.OBV = self.I(OBV,self.Close,self.Volume)
     13 
     14     def next(self):

~/anaconda3/envs/Mamani_1_0_0/lib/python3.8/site-packages/backtesting/backtesting.py in I(self, func, name, plot, overlay, color, scatter, *args, **kwargs)
    126             value = func(*args, **kwargs)
    127         except Exception as e:
--> 128             raise RuntimeError(f'Indicator "{name}" errored with exception: {e}')
    129 
    130         if isinstance(value, pd.DataFrame):

RuntimeError: Indicator "OBV(C,V)" errored with exception: input array type is not double

Any idea what could be wrong?

Thanks,
Carlos Mamani

[ghost]

Edit:
I had to convert the Volume Series type to Double, for some reason it fails when I call the OBV indicator through the backtesting.Indicator class, but If I call the OBV class directly from the talib module it doesn't fail.

[kernc]

FWIK, ta-lib functions all expect array.astype(float) as their inputs when of np.ndarray type. The reason you see it works with plain OBV(df.Close, df.Volume) is that ta-lib does this on-the-fly pd.Series unwrapping.

You can, therefore, call:

        self.OBV = self.I(OBV, self.Close, self.Volume.astype(float))

or, using the .s accessor:

        self.OBV = self.I(OBV, self.Close, self.Volume.s)