Why if x!=np.nan is not working in the lambda method?

Asked Jan 23 '22 at 18:48

Active Jan 23 '22 at 18:55

Viewed 389 times

Given that i have a below datadrame:

dt = pd.DataFrame({"Cabin" : ["Hello","Morning",np.NAN,"BBC"]})

I wish to apply the following process on the Cabin column:

dt["Cabin"].map(lambda x: x[0] if x!=np.nan else np.nan )

But, i am facing with the below error:

TypeError: 'float' object is not subscriptable

I can understand the if x!=np.nan is not working, but i don't know how to fix it ?

The expected output is as following

expected_output = pd.DataFrame({"Cabin" : ["H","M",np.NAN,"B"]})

edited Jan 23 '22 at 18:52

asked Jan 23 '22 at 18:48

Jeff

7,767
28
85
138

2

What's the expected outcome? – Mureinik Jan 23 '22 at 18:50
I think the root problem here is that you're expecting `np.NAN != np.NAN` to be `False` so that `x[0]` won't be evaluated when `x` is nan. – Brian61354270 Jan 23 '22 at 18:51
The lambda will always return `np.nan` since `np.nan != np.nan`. – BrokenBenchmark Jan 23 '22 at 18:51
2

*NOTHING* is equal to `nan` - not even `nan`. Numpy has a `np.isnan()` function for performing this test. – jasonharper Jan 23 '22 at 18:51
Of course numpy.NAN is not subscriptable – Seabyte Jan 23 '22 at 18:52
Its the Nan inequality problem addressed in several areas like https://stackoverflow.com/questions/10034149/why-is-nan-not-equal-to-nan. – tdelaney Jan 23 '22 at 18:54
From the [documentation](https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.Series.map.html): instead of explicitly checking for nan yourself, just pass `na_action='ignore'`. – Karl Knechtel Jan 23 '22 at 18:55
@jasonharper i need to check if it is not `nan`. I tried the `lambda x: x[0] if !np.isnan(x) else np.nan` but it has syntax error !!! – Jeff Jan 23 '22 at 18:56
What other values can be in this column? I see string and Nan, can there be other things like non-Nan float? – tdelaney Jan 23 '22 at 18:57
1

@Jeff while the linked dupes answer your actual question, you're over complicating this based on your example data - just use `df['Cabin'].str[0]` – Jon Clements Jan 23 '22 at 18:57
1

Try `if not np.isnan(x)` – tdelaney Jan 23 '22 at 18:57

Why if x!=np.nan is not working in the lambda method?

0 Answers0