How to convert pandas dataframe to line object?

Tony 0

Hi all,
I'm trying to port an indicator implemented by the FinTa library. The indicator receives as input the pandas dataframe in the format ohlc and returns two variables per each candle that analyzes in this format:

                              WT1.       WT2.

datetime
2019-08-27 04:00:00+00:00 NaN NaN
2019-08-27 05:00:00+00:00 116.666667 NaN
2019-08-27 06:00:00+00:00 81.045764 NaN
2019-08-27 07:00:00+00:00 23.298112 NaN
2019-08-27 08:00:00+00:00 43.091340 66.025471
... ... ...

This is the code I have used to integrate this indicator:

class WTO(bt.Indicator):
lines = (('wt1'), ('wt2'),)

def __init__(self):
    self.addminperiod(10)
    ohlc = self.data._dataname[['open', 'high', 'low', 'close']]
    
    wto = ft.TA.WTO(ohlc, channel_lenght = 7,
                    average_lenght = 21)
    print(wto)
    # print(wto['WT1.'].to_list())
    # print(wto['WT2.'].to_list())
    self.lines.wt1 = wto['WT1.'].values
    self.lines.wt2 = wto['WT2.'].values

The interpreter gives me the following error:

Traceback (most recent call last):
File "backtrade2.py", line 101, in <module>
cerebro.run()
File "/home/antonino/.local/lib/python3.6/site-packages/backtrader/cerebro.py", line 1127, in run
runstrat = self.runstrategies(iterstrat)
File "/home/antonino/.local/lib/python3.6/site-packages/backtrader/cerebro.py", line 1217, in runstrategies
strat = stratcls(*sargs, **skwargs)
File "/home/antonino/.local/lib/python3.6/site-packages/backtrader/metabase.py", line 88, in call
_obj, args, kwargs = cls.doinit(_obj, *args, **kwargs)
File "/home/antonino/.local/lib/python3.6/site-packages/backtrader/metabase.py", line 78, in doinit
_obj.init(*args, **kwargs)
File "backtrade2.py", line 27, in init
self.my_wto = WTO(self.datas[0])
File "/home/antonino/.local/lib/python3.6/site-packages/backtrader/indicator.py", line 53, in call
return super(MetaIndicator, cls).call(*args, **kwargs)
File "/home/antonino/.local/lib/python3.6/site-packages/backtrader/metabase.py", line 88, in call
_obj, args, kwargs = cls.doinit(_obj, *args, **kwargs)
File "/home/antonino/.local/lib/python3.6/site-packages/backtrader/metabase.py", line 78, in doinit
_obj.init(*args, **kwargs)
File "backtrade2.py", line 22, in init
self.lines.wt1 = wto['WT1.'].values
File "/home/antonino/.local/lib/python3.6/site-packages/backtrader/lineseries.py", line 79, in set
value = value(0)
TypeError: 'numpy.ndarray' object is not callable

The error is clearly caused by the fact that backtrader expects data in a different format. It should be a line object. Is there any good way to convert my data in that format. I can work around this issue by calculating the indicator outside backtrader and then feed it as input data, but this sounds like a dirty workaround.

Thanks for your help.

run-out

@tony-0 I hope I'm following you correctly, but if so, you should preprocess your pandas ohlc and the indicators before loading data into your backtest. You can load a pandas dataframe using this and this

Tony 0

@run-out That is definitely an option, but I would like to integrate the Wave Trend Oscillator indicator from the FinTa (https://pypi.org/project/finta/) library directly within my strategy. Is that possible?

run-out

@tony-0 I think a good option then would be to create a custom indicator. I have used pandas inside custom indicators to so some heavy lifting. Have a look here and see if this helps.

Tony 0

@run-out I'm trying to create it as a custom indicator but I don't know how to convert the FinTa indicator's output in a line object.
This is my code

class WTO(bt.Indicator):
    lines = (('wt1'), ('wt2'),)
    
    def __init__(self):
        self.addminperiod(10)
        ohlc = self.data._dataname[['open', 'high', 'low', 'close']]
        
        wto = ft.TA.WTO(ohlc, channel_lenght = 7,
                        average_lenght = 21)
        print(wto)
        self.lines.wt1 = wto['WT1.'].values
        self.lines.wt2 = wto['WT2.'].values

run-out

@tony-0 Have you checked to see if you are getting values here:

self.lines.wt1 = wto['WT1.'].values
self.lines.wt2 = wto['WT2.'].values

Assuming you are, you should then be getting lines for your indicator. However, I think you have to take out the braces in your lines definition :

lines = ('wt1', 'wt2',)

Let us know if that works, I'm interested to know.

Tony 0

@run-out
Yes, if I print those to variables I get:

[         nan 116.66666667  81.04576412 ...  -3.12979911   6.40985569
  13.60683347]
[         nan          nan          nan ... -16.15140739  -7.0198421
   1.50132138]

I tried to remove the braces but that didn't change. I still get the same error.

Traceback (most recent call last):
  File "backtrade2.py", line 126, in <module>
    cerebro.run()
  File "/home/antonino/.local/lib/python3.6/site-packages/backtrader/cerebro.py", line 1127, in run
    runstrat = self.runstrategies(iterstrat)
  File "/home/antonino/.local/lib/python3.6/site-packages/backtrader/cerebro.py", line 1217, in runstrategies
    strat = stratcls(*sargs, **skwargs)
  File "/home/antonino/.local/lib/python3.6/site-packages/backtrader/metabase.py", line 88, in __call__
    _obj, args, kwargs = cls.doinit(_obj, *args, **kwargs)
  File "/home/antonino/.local/lib/python3.6/site-packages/backtrader/metabase.py", line 78, in doinit
    _obj.__init__(*args, **kwargs)
  File "backtrade2.py", line 41, in __init__
    self.my_wto = WTO(self.datas[0])
  File "/home/antonino/.local/lib/python3.6/site-packages/backtrader/indicator.py", line 53, in __call__
    return super(MetaIndicator, cls).__call__(*args, **kwargs)
  File "/home/antonino/.local/lib/python3.6/site-packages/backtrader/metabase.py", line 88, in __call__
    _obj, args, kwargs = cls.doinit(_obj, *args, **kwargs)
  File "/home/antonino/.local/lib/python3.6/site-packages/backtrader/metabase.py", line 78, in doinit
    _obj.__init__(*args, **kwargs)
  File "backtrade2.py", line 36, in __init__
    self.lines.wt1 = wto['WT1.'].values 
  File "/home/antonino/.local/lib/python3.6/site-packages/backtrader/lineseries.py", line 79, in __set__
    value = value(0)
TypeError: 'numpy.ndarray' object is not callable

run-out

@tony-0 Do you mind if I ask why you don't want to pre-process this?