Skip plotting NaN and inf values in python

Question

Is there a way to skip plotting NaN and Inf values in python?

Say I have something like:

and I want to plot m1 vs m2/m3 so there are cases I am dividing by zero, so when I go to plot it throws a ValueError: Axis limits cannot be NaN or Inf

What is the solution? Thank you.

(I am using dataframes in pandas where I go m4 = df['m2'] / df['m3'] and go to plot m1 vs m4)

Please note: there are no inf or Nan Values in the dataframe itself, but they occur after manipulation.

Possible duplicate of [dropping infinite values from dataframes in pandas?](https://stackoverflow.com/questions/17477979/dropping-infinite-values-from-dataframes-in-pandas) — Chris, Sep 29 '19 at 03:55
It's not, as that removes it from the data frame itself. This is plotting after manipulation — sci-guy, Sep 29 '19 at 04:03

Valdi_Bo · Answer 1 · 2019-09-29T05:36:49.077

According to what tools provides us Pandas, the easiest solution is to:

replace Inf values (both positive and negative) with NaN,
drop NaN values.

They your Series will contain neither Infs nor NaNs.

So, generating m4, expand your instruction to:

m4 = (df['m2'] / df['m3']).replace([np.inf, -np.inf], np.nan).dropna()

getting (in your case):

0    2.500000
1    0.666667
2    0.750000
4    0.375000
6    0.750000
dtype: float64

(no values for key 3 and 5), which can be plotted as you wish.

Edit

I noticed that (at least in Pandas v. 0.25) a DataFrame scatter plot can be generated even with Inf or NaN values in y column.

I did the following experiment:

Created m4 without removing Inf / Nan values, but setting its name (will be needed in a moment in join):
```
m4 = df.m2.divide(df.m3).rename('m4')
```
Replaced the second occurrence of Inf with NaN:
```
m4.iat[5] = np.nan
```
so now it contains both Inf and NaN.

Generated the plot:

df.join(m4).plot.scatter(x='m1', y='m4');

In effect, I got the picture, without any errors.

Apparently, any Inf / NaN values are in this case silently dropped and only after that the picture is generated.

score 0 · Answer 2 · answered Sep 29 '19 at 05:32

What you can do is to yield the rows where all its elements are finite.

import pandas as pd, numpy as np 
m1 = [4 ,3 ,4 ,2 ,4 ,3 ,2]
m2 = [5 ,2 ,3 ,5 ,3 ,4 ,3]
m3 = [2, 3, 4, 0, 8, 0, 4]
df = pd.DataFrame({'m1':m1,'m2':m2,'m3':m3})

Say m4 = m2/m3:

df['m4']= df.m2/df.m3

...and MAY CONTAIN infinite elements. To solve this problem, use NumPy’s isfinite on your data frame df and then use .all(1), which returns TRUE if all cells in a row are finite:

df[np.isfinite(df).all(1)]

For visualization purposes, please see the attached screenshots.

After this, you can now plot m1 vs. m4 without worrying about the infinite values:

Hope this helps!

Skip plotting NaN and inf values in python

2 Answers2

Edit