Why do I get a ‘FutureWarning’ with pandas.concat?

Does anyone meet this FutureWarning when you use Tiingo+pandas_datareader?

The warning is like:

python3.8/site-packages/pandas_datareader/tiingo.py:234: FutureWarning: In a future version of pandas all arguments of concat except for the argument 'objs' will be keyword-only
    return pd.concat(dfs, self._concat_axis)

I think this warning does not impact my accessing to tiingo data, as I can get all the data I want. I just want to understand if there is any risk with my current code :

my python3                -  3.8.5, 
Python 3.8.5, pytest-6.2.4, py-1.10.0, pluggy-0.13.1
pandas_datareader version -  0.10.0
pandas version            -  1.3.2

I also tested my code with a real ‘futureVersion’ of python: 3.9.6 (comparing with python 3.8.5). To my suprise, I didn’t get any warning or error:

platform win32 -- Python 3.9.6, pytest-6.2.4, py-1.10.0, pluggy-0.13.1

Any advice is appreciated.

Answer

Most function parameters in python are “positional or keyword” arguments.

I.e. if I have this function:

def do_something(x, y):
    pass

Then I can either call it like this, using positional arguments:

do_something(1, 2)

Or like this, using keyword arguments:

do_something(x=1, y=2)

Or like this, using a mixture of the two (but note that you’re not allowed to have a positional argument after a keyword argument):

do_something(1, y=2)

But you can also define functions with positional-only or keyword-only parameters

Say I have this other function:

def do_something_else(x, /, y, *, z):
    pass

In this function, I’ve marked x as being positional-only, because it comes before the /. And I’ve marked z as being keyword-only, because it comes after the *. y is a positional-or-keyword parameter, as it comes after the / but before the *. This means that these two attempts to call the function will fail: the first one because z is being called as a positional argument, and the second because x is being called as a keyword argument:

do_something_else(1, 2, 3)       # will fail!
do_something_else(x=1, y=2, z=3) # will fail!

These two attempts, however, will both succeed — y is still a positional-or-keyword parameter.

do_something_else(1, 2, z=3)     # fine
do_something_else(1, y=2, z=3)   # fine

The `FutureWarning` message.

The FutureWarning message has nothing to do with the version of python you’re using, but has everything to do with the version of pandas that you’re using. Pandas is a third-party library, not a part of the python core, so the pandas version you’re using is a completely different thing to the python version you’re using.

The warning is letting you know that currently, you’re fine to write pd.concat(dfs, self._concat_axis), but that they’re planning on changing the definition of the function in a future version of pandas so that all arguments except for objs will be keyword-only. I.e., after they make this change, pd.concat(dfs, self._concat_axis) will raise an error, and you will have to write pd.concat(dfs, axis=self._concat_axis) instead. They are most likely considering making this change because calling functions with keyword arguments is often clearer and more readable for other people.