[For data science] Oreore Jupyter cheat sheet [Jupyter Notebook / Lab]

I usually do ** data analysis and model building ** mainly on Jupyter. Make a note of how many times you check it. (Updated from time to time)

Development

I want to reload the modified .py module

Modules that have already been imported will be cached or will not be updated even if they are imported again normally. ** You can reload (reimport) with the following. ** **

import importlib
importlib.reload(hoge)
#hoge is an imported module

I didn't know this until recently, and I restarted it every time, so it's really eye-opening. Another option is How to auto-update modules with % autoreload.

I want to change the tab name of Jupyter Lab

If you have multiple servers running, you often don't know which tab is the notebook on which server. You can change the tab name as follows.

Do the following within Jupyter

%%javascript
document.title='Jupyter-GPU'

Alternatively, it can be specified at build time.

jupyter lab build --name='Jupyter-GPU'

Reference: https://github.com/jupyterlab/jupyterlab/issues/4422#issuecomment-395962448

Drawing

I want to draw Japanese with matplotlib

If you are not particular about fonts, ** japanize-matplotlib is the quickest way. ** **

 pip install japanize-matplotlib

import matplotlib.pyplot as plt
import japanize_matplotlib

plt.plot([1, 2, 3, 4])
plt.xlabel('The joy of using Japanese easily')
plt.show()

Log edition

I want to debug with pdb

There are two timings to start the debugger.

** Specify breakpoint and start debugger ** Insert the code below

from IPython.core.debugger import Pdb; Pdb().set_trace()

** Debugger starts when a bug occurs ** Applies only to specific cells

#Put it at the beginning of the cell you want to debug
%%debug

Applies to the entire notebook

#Put it somewhere in your notebook
%pdb on
#This is when you want to turn off the bug detection mode
%pdb off 

[Jupyter Notebook / Lab] 3 ways to debug on Jupyter [Pdb]

I want to display a progress bar

I wonder if I can go with this once

from tqdm.auto import tqdm
import numpy as np

#Enclose in tqdm
for i in tqdm(np.arange(1, 100000, 1)):
    #Process here
    pass

• In Jupyter Lab, it may not be displayed properly unless the plug-in is installed. Reference

pandas edition

I want to display the progress bar with apply of pandas

import pandas as pd
import numpy as np
from tqdm.auto import tqdm

# set description
tqdm.pandas(desc="Do this")

# apply
df = pd.DataFrame({'hoge': np.arange(1, 100000, 1)})
df['hoge'] = df['hoge'].progress_apply(lambda x:  x + 1)

• In Jupyter Lab, it may not be displayed properly unless the plug-in is installed. Reference

I want to prevent the display of DataFrame from being omitted

Increase the number of items that can be displayed and the maximum number of characters that can be displayed in one cell.

import pandas as pd
pd.set_option("display.max_colwidth", 500) #500 characters in 1 cell
pd.set_option("display.max_rows", 100) #Can display 100 lines

** * If set_option does not work in JupyterLab, it works well if you display the records below max_rows likedf [: 100](from my personal experience) **

If you want to prevent omission of only a specific cell, do the following (@chik_taks told me!)

with pd.option_context('display.max_colwidth', 200):
    display(df)

I want to output DataFrame in Markdown tabular format

You can output DataFrame as markdown and copy it. I use it soberly

pip install pytablewriter

import pytablewriter
writer = pytablewriter.MarkdownTableWriter()
writer.from_dataframe(df)
writer.write_table()
# | col1 |  col2  |
# |------|--------|
# |hoge1 |line1   |
# |hoge2 |line2   |

Shortcut system

Shortcut to command palette

• Jupyter：Cmd + Option + P
• JupyterLab：Ctrl/Cmd + Shift + C

Reference