How to split a ipython notebook

Question

My Jupyter notebooks is getting long, which makes it difficult to navigate.

I want to save each chapter (Cel starting with Heading 1) to a different file. How can I do that? Cut and paste of multiple cells between notebooks seems not possible.

Answer 1

This is the method I use - it is a little awkward, but it works:

1. Make multiple copies of the master notebook using File->Make Copy from the menu. make one copy for each chapter you want to extract.
2. Rename the copy for each chapter: e.g. rename "master-copy0" to "Chapter 1".
3. Delete each cells that don't belong to Chapter 1 - for example using 'dd' in command mode.
4. Save the abbreviated file.
5. Repeat steps 3 and 4 for each chapter.

I believe that the developers may be working on a better solution for a future release.

Answer 2

The easiest way might be to edit the .ipnb file in a text editor. Below I listed the content of a very simple notebook.

The notebook looks like

Chapter 1

In [1]: 1+1

Out[1]: 2

Chapter 2

In [2]: 2+2

Out[2]: 4

To take out chapter 1 and place it behind chapter 2, this is what you can do

1. Search for "level": 1
2. You find { "cell_type": "heading", "level": 1, "metadata": {}, "source": [ "Chapter 1" ] }, and { "cell_type": "heading", "level": 1, "metadata": {}, "source": [ "Chapter 2" ] },
3. Move everything from the start of the first search result, just below the end of the second search result
4. Pay attention to commas

You can manipulate multiple notebooks in a simlar fashion.

This is the .ipnb file for the example

{
 "metadata": {
  "name": "",
  "signature": ""
 },
 "nbformat": 3,
 "nbformat_minor": 0,
 "worksheets": [
  {
   "cells": [
    {
     "cell_type": "heading",
     "level": 1,
     "metadata": {},
     "source": [
      "Chapter 1"
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "1+1"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "metadata": {},
       "output_type": "pyout",
       "prompt_number": 1,
       "text": [
        "2"
       ]
      }
     ],
     "prompt_number": 1
    },
    {
     "cell_type": "heading",
     "level": 1,
     "metadata": {},
     "source": [
      "Chapter 2"
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "2+2"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "metadata": {},
       "output_type": "pyout",
       "prompt_number": 2,
       "text": [
        "4"
       ]
      }
     ],
     "prompt_number": 2
    }
   ],
   "metadata": {}
  }
 ]
}

Answer 3

A notebook file is json format, so I get all data as JSON format and split it into several files automatically.

This code is what I made.

The code seems to be complex, but it is simple if you just check it for a while and this is an example of a separate file, http://www.fun-coding.org/DS&AL4-1.html which I also transformed as HTML after I split it.

import json
from pprint import pprint
import re

def notebook_spliter(FILENAME, chapter_num):

    with open(FILENAME + '.ipynb') as data_file:    
        data = json.load(data_file)

    copy_cell, chapter_in = list(), False

    regx = re.compile("## [0-9]+\. ")
    for num in range(len(data['cells'])):
        if chapter_in and data['cells'][num]['cell_type'] != 'markdown':
            copy_cell.append(data['cells'][num])
        elif data['cells'][num]['cell_type'] == 'markdown':
            regx_result = regx.match(data['cells'][num]['source'][0])

            if regx_result:
                print (regx_result.group())
                regx2 = re.compile("[0-9]+")
                regx2_result = regx2.search(regx_result.group())
                if regx2_result:
                    print (int(regx2_result.group()))
                    if chapter_in == False:
                        if chapter_num == int(regx2_result.group()):
                            chapter_in = True
                            copy_cell.append(data['cells'][num])
                    else:
                        if chapter_num != int(regx2_result.group()):
                            break
            elif chapter_in:
                copy_cell.append(data['cells'][num])

    copy_data["cells"] = copy_cell
    copy_data["metadata"] = data["metadata"]
    copy_data["nbformat"] = data["nbformat"]
    copy_data["nbformat_minor"] = data["nbformat_minor"]
    with open(FILENAME + '-' + str(chapter_num) + '.ipynb', 'w') as fd:
        json.dump(copy_data, fd, ensure_ascii=False)

This is a function to check chapter numbers in a notebook file. I added chapter number to the notebook file with '## 1. chapter name' in markdown cell, so just check ## digit. pattern with regular expression.

Then, next code is to copy data of cells into this chapter number, and save the only copied cells and others(metadata, nbformat, and nbformat_minor) to separate file.

copy_data = dict()
FILENAME = 'DS&AL1' 
CHAPTERS = list()
with open(FILENAME + '.ipynb') as data_file:    
    data = json.load(data_file)

for num in range(len(data['cells'])):
    if data['cells'][num]['cell_type'] == 'markdown':
        regx_result = regx.match(data['cells'][num]['source'][0])

        if regx_result:
            regx2 = re.compile("[0-9]+")
            regx2_result = regx2.search(regx_result.group())
            if regx2_result:
                CHAPTERS.append(int(regx2_result.group()))
print (CHAPTERS)

for chapternum in CHAPTERS:
    notebook_spliter(FILENAME, chapternum)

Answer 4

2023 Update

Some years later, luckily there is a library that can do such things for you:

pip install nbmanips
nb select has_html_tag h1 | nb split -s nb.ipynb

• The first part of the command (nb select has_html_tag h1) will tell nbmanips on which cells to perform the split.

• The second part (nb split -s nb.ipynb) will split the notebook based on the piped selection. The -s flag tells nbmanips to use the selection instead of a cell index.

my source: https://towardsdatascience.com/split-your-jupyter-notebooks-in-2-lines-of-code-de345d647454

the library: https://pypi.org/project/nbmanips/