PerformanceWarning: DataFrame is highly fragmented

[mikeperalta1]

Which version are you running? The lastest version is on Github. Pip is for major releases.
0.3.2b0

Upgrade.
I appear to be running the latest (?)

Describe the bug
When adding technical indicators to an existing data frame, I receive the following warning:

PerformanceWarning: DataFrame is highly fragmented.  This is usually the result of calling `frame.insert` many times, which has poor performance.  Consider using pd.concat instead.  To get a de-fragmented frame, use `newframe = frame.copy()`

These warnings are reduced when I implement the suggestion of newframe = frame.copy(), but not eliminated. I believe fragmentation is happening during pandas-ta calls, but I can't tell for sure because there is no trace.

To Reproduce

I start with a DataFrame populated with just the right columns for pandas-ta. Then add technical indicators like so:

for n in self.__SMA_INTERVALS:
	
	key = "SMA_{}".format(n)
	
	df[key] = ta.sma(df["Close"], length=n)
	df[key] = df[key].fillna(0)
	
	df = df.copy()

I'm not sure whether the warnings come from the SMA indicator. I'm adding SMA, EMA, MACD, STOCH, RSI, ADX, CCI, AROON, BBANDS, the ta.cmf one (Chaikin's AD??), and OBV

Expected behavior
No warnings or excessive fragmentation during the computations.

Screenshots
N/A

Additional context
pandas: 1.3.0
pandas-ta: 0.3.2b0
python: 3.9.6
numpy: 1.21.0

Thanks for using Pandas TA!

[twopirllc]

Hello @mikeperalta1,

I am aware of the PerformanceWarning (PW), but it's just a warning and not a bug. It was never a warning in prior versions of Pandas. So you can downgrade to an earlier Pandas version, or you can suppress the warning. In fact the current development version does suppress the PW until a permanent fix is in place.

from warnings import simplefilter
simplefilter(action="ignore", category=pd.errors.PerformanceWarning)

It occurs when you are appending a Pandas DataFrame to an existing Pandas DataFrame. They recommend using pd.concat() to quickly combine two or more DataFrames. It is not because of an internal calculation of an indicator.

In Pandas TA, it occurs in the internal method _append of the main Pandas TA DataFrame extension class when trying to append the resultant (aroon, bbands, et al) DataFrame to current DataFrame and almost never when appending a resultant (sma, rsi, et al) Series to the current DataFrame.

@pd.api.extensions.register_dataframe_accessor("ta")
class AnalysisIndicators(BasePandasObject):
    # ...
    def _append(self, result=None, **kwargs) -> None:
        # ...

Furthermore, I have already tried to use pd.concat() and frame.copy() in the _append() method to make the PW disappear to no avail. In fact, it would not append the resultant DataFrame to the current DataFrame as expected like it currently does in the for loop:

pandas-ta/pandas_ta/core.py

Lines 416 to 418 in 1deb755

	else:
	for i, column in enumerate(result.columns):
	df[column] = result.iloc[:, i]

:

which is generating this PW. Now this is either a Pandas bug or I am coding pd.concat incorrectly. 🤷🏼‍♂️ I am open to contributions to fix this Issue as it obviously a concern. 😎

Hope this helps!

Kind Regards,
KJ

[mikeperalta1]

@twopirllc Thank you for that thorough explanation! I suppose for now I will simply mute the warning.

[LucianPopaLVP]

any fix to this issue? i have the same problem :(

[twopirllc]

@LucianPopaLVP,

Which Pandas TA version are you running? It has already been silenced in version 0.3.14b. If you are concating yourself, then you need to apply the simplefilter for your situation.

As you know, it's under Pandas care so stay abreast with the discussion in Pandas Issue 42477

Kind Regards,
KJ

[ClaoLinda]

What if you are not even using the "import pandas as pd" and you still get the error?

[SkyisjustTheBeginning]

Well , If you dont import pandas as pd , You could try this -

from warnings import simplefilter
simplefilter(action="ignore", category=pd.errors.PerformanceWarning)