%matplotlib notebook showing a blank histogram
Question:
In my Jupyter notebook I am now using %matplotlib notebook instead of %matplotlib inline, it’s awesome that I can now interact with my plots on Jupyter. However, when I try to make an histogram I get a blank plot:
If I use %matplotlib inline everything works fine:
What’s going on?
Answers:
Seeing that my comment above has indeed helped someone to solve the problem I will post it as an answer.
The problem occurs if you switch from %matplotlib inline to %matplotlib notebook without restarting the kernel.
Switching from %matplotlib notebook to %matplotlib inline works fine.
So the solution is to either restart the kernel or start a new notebook.
It seems that in some cases it helps to repeat the setting of the notebook backend, i.e. call it twice like
%matplotlib notebook
%matplotlib notebook
An analysis for why that is can be found in this comment
The answer is not necessarily to restart the entire kernel.
If you reload the matplotlib module, it will work, too. Provided you use Python 3.6 like me, and you have import matplotlib.pyplot as plt like me:
from importlib import reload
reload(plt)
%matplotlib notebook
It does the trick. Yes it is still a hack. At least this is an independent codecell you can use in the middle of the notebook. Switching back via %matplotlib inline is not a problem.
You can also remove once imported names from the sys.modules list, then they get imported again when you call the import again.
import sys
sys.modules.pop('matplotlib')
from matplotlib import pyplot as plt
In many cases, that’s a less good idea. But it might sometimes be the only straw to hold on.
The issue seems to be an interaction between switching:
%matplotlib notebook
%matplotlib inline
and using the figure “power button” on interactive plots:
Solution: If you pressed the button and are getting blank plots, restart Jupyter and don’t use the power button again if you need to switch between notebook and inline
On a new notebook if you start by using %matplotlib notebook then you can switch between that an inline without issue.
I confirm that both restarting of the kernel and the running of the code do the trick!
I was able to fix it by downgrading matplotlib to 3.1.3:
conda install matplotlib=3.1.3
I was running version 3.3.2 and had this same issue. I was not switching between %matplotlib inline and %matplotlib notebook, and it did not matter if I placed %matplotlib notebook before or after importing
For me, it also appeared after switching from ‘inline’ to ‘notebook’ but restarting the kernel didn’t work. I had to go through file > close and halt, and then either reopen it or restart the kernel if the page is still open.
In my Jupyter notebook I am now using %matplotlib notebook instead of %matplotlib inline, it’s awesome that I can now interact with my plots on Jupyter. However, when I try to make an histogram I get a blank plot:
If I use %matplotlib inline everything works fine:
What’s going on?
Seeing that my comment above has indeed helped someone to solve the problem I will post it as an answer.
The problem occurs if you switch from %matplotlib inline to %matplotlib notebook without restarting the kernel.
Switching from %matplotlib notebook to %matplotlib inline works fine.
So the solution is to either restart the kernel or start a new notebook.
It seems that in some cases it helps to repeat the setting of the notebook backend, i.e. call it twice like
%matplotlib notebook
%matplotlib notebook
An analysis for why that is can be found in this comment
The answer is not necessarily to restart the entire kernel.
If you reload the matplotlib module, it will work, too. Provided you use Python 3.6 like me, and you have import matplotlib.pyplot as plt like me:
from importlib import reload
reload(plt)
%matplotlib notebook
It does the trick. Yes it is still a hack. At least this is an independent codecell you can use in the middle of the notebook. Switching back via %matplotlib inline is not a problem.
You can also remove once imported names from the sys.modules list, then they get imported again when you call the import again.
import sys
sys.modules.pop('matplotlib')
from matplotlib import pyplot as plt
In many cases, that’s a less good idea. But it might sometimes be the only straw to hold on.
The issue seems to be an interaction between switching:
%matplotlib notebook
%matplotlib inline
and using the figure “power button” on interactive plots:
Solution: If you pressed the button and are getting blank plots, restart Jupyter and don’t use the power button again if you need to switch between notebook and inline
On a new notebook if you start by using %matplotlib notebook then you can switch between that an inline without issue.
I confirm that both restarting of the kernel and the running of the code do the trick!
I was able to fix it by downgrading matplotlib to 3.1.3:
conda install matplotlib=3.1.3
I was running version 3.3.2 and had this same issue. I was not switching between %matplotlib inline and %matplotlib notebook, and it did not matter if I placed %matplotlib notebook before or after importing
For me, it also appeared after switching from ‘inline’ to ‘notebook’ but restarting the kernel didn’t work. I had to go through file > close and halt, and then either reopen it or restart the kernel if the page is still open.