[Pandas-dev] Series.apply "regression" and raw assert

[Marc Garcia]

Thanks for reporting this Pål.

I think it's probably better if you open an issue for this. I think it'll
be easier to keep of all the discussion easily accessible, and not get this
lost.

The first thing that comes to my mind is is if callable(F()) returns True.
I guess it does, then I wonder why assert callable(f) in the pandas code, I
guess f is not the original parameter. Maybe you can have a look at that.

Also, maybe git blame on that assert line points you to the breaking
change, and the PR and issue give you an idea of why the change, and
whether this was intentional.


On Thu, Jun 30, 2022, 17:35 Pål Grønås Drange <paal.drange at gmail.com> wrote:

> Consider the following MWE
>
> ```
> import pandas as pd
> print(pd.__version__)
> df = pd.DataFrame({"x": [1]})
>
> class F:
>     def __iter__(self):
>         yield 0
>     def __call__(self, x):
>         return x
>
> f = F()
> df["x"].apply(lambda e: f(e))  # ok
> df["x"].apply(f.__call__)  # ok
> df["x"].apply(f)  # AssertionError: f not callable
> ```
>
> In Pandas < 1.3, the three calls all return the same value, namely a
> Series containing only [1].
>
> Starting from Pandas 1.3, the last of the three calls raises an
> AssertionError.
>
> This is a backwards incompatible change that happened in 1.3.
>
> However,
> 1. I have realized that I should use Series.map rather than apply.  This
> fixes my issue.
> 2. The documentation for Series.apply says that it expects a "Python
> function" or "Numpy ufunc".  Apparently, it doesn't support "Python
> callables".
>
> The only output from Pandas is "AssertionError", and in the code
> (pandas/core/apply.py:1046 [1106 in trunk]), it simply says "assert
> callable(f)":
>
>
> ```
> # string, list-like, and dict-like are entirely handled in super
> assert callable(f)
> ```
>
> This is weird, since _my f_ clearly is callable.  Would it be possible
> to at least add a string to the assert to make it easier for users to
> understand what's going on?
>
>
> Here is the output:
>
> ```
> (e39) [~]$ python applytest.py
> 1.2.5
> (e39) [~]$ pip install -q pandas==1.3
> (e39) [~]$ python applytest.py
> 1.3.0
> Traceback (most recent call last):
>   File "/home/p/applytest.py", line 14, in <module>
>     df["x"].apply(f)  # AssertionError: f not callable
>   File "/home/p/e39/lib/python3.9/site-packages/pandas/core/series.py",
> line 4356, in apply
>     return SeriesApply(self, func, convert_dtype, args, kwargs).apply()
>   [...]
>   File "/home/p/e39/lib/python3.9/site-packages/pandas/core/apply.py",
> line 1046, in agg
>     assert callable(f)
> AssertionError
> ```
>
> Best regards,
> Pål GD
> _______________________________________________
> Pandas-dev mailing list
> Pandas-dev at python.org
> https://mail.python.org/mailman/listinfo/pandas-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mail.python.org/pipermail/pandas-dev/attachments/20220630/77d9d1b7/attachment.html>