Bug Report:

[zbig01]

Bug when using box_size = 'atr' in pnf charts in mplfinance : FutureWarning: Series.getitem treating keys as positions is deprecated

OS Win 11 Pro 24H2 Desktop
IDE Visual Studio Code 1.96.2
Python 3.12.8
mplfinance 0.12.10b0

When using the 'atr' box_size parameter in creating pnf charts in mplfinance the _utils.py script throws this warning

\mplfinance_utils.py:129: FutureWarning: Series.getitem treating keys as positions is deprecated. In a future version, integer keys will always be treated as labels (consistent with DataFrame behavior). To access a value by position, use ser.iloc[pos] high = highs[i]
\mplfinance_utils.py:130: FutureWarning: Series.getitem treating keys as positions is deprecated. In a future version, integer keys will always be treated as labels (consistent with DataFrame behavior). To access a value by position, use ser.iloc[pos] low = lows[i]
_utils.py:131: FutureWarning: Series.getitem treating keys as positions is deprecated. In a future version, integer keys will always be treated as labels (consistent with DataFrame behavior). To access a value by position, use ser.iloc[pos] close_prev = closes[i-1]

This is the responsible function:

def _calculate_atr(atr_length, highs, lows, closes):
    """Calculate the average true range
    atr_length : time period to calculate over
    all_highs : list of highs
    all_lows : list of lows
    all_closes : list of closes
    """
    if atr_length < 1:
        raise ValueError("Specified atr_length may not be less than 1")
    elif atr_length >= len(closes):
        raise ValueError("Specified atr_length is larger than the length of the dataset: " + str(len(closes)))
    atr = 0
    for i in range(len(highs)-atr_length, len(highs)):
        high = highs[i]
        low = lows[i]
        close_prev = closes[i-1]
        tr = max(abs(high-low), abs(high-close_prev), abs(low-close_prev))
        atr += tr
    return atr/atr_length

This warning is suppressed when you modify the code like so

high = highs.iloc[i]
low = lows.iloc[i]
close_prev = closes.iloc[i-1]

This however creates a new problem if you subsequently run a Renko chart, you get this error

LocalCache\local-packages\Python312\site-pa
high = highs.iloc[i]
^^^^^^^^^^
AttributeError: 'numpy.ndarray' object has no attribute 'iloc'

To fix this I created a small helper function

def safe_indexing(data, index):
    if isinstance(data, pd.Series):
        return data.iloc[index]
    else:
        return data[index]

and placed it in _utils.py
and updated _utils.py

high = highs[i]
low = lows[i]
close_prev = closes[i-1]

to

high = safe_indexing(highs, i)
low = safe_indexing(lows, i)
close_prev = safe_indexing(closes, i-1)

Now both charts render with no messages

[DanielGoldfarb]

@zbig01

Thanks for reporting this. And your solution is a good one. However I am wondering (knowing some of the history of this package) if the real issue is this: Once upon a time the values passed into _calculate_atr(atr_length, highs, lows, closes) for highs, lows, and closes were always lists. And then when I re-wrote the pnf code, I passed in pd.Series instead.

If so, then maybe the more appropriate solution will be to fix the pnf rewrite to pass in the values as lists. I will take a look.

[zbig01]

@DanielGoldfarb

Thanks for your comment. I realize my fix was more of a hack than a fix. Your suggestion of using lists will likely be a more elegant solution.

Cheers and Happy holidays.