why would opening a jupyter notebook get "'outputs' were unexpected" error?

Question

I am trying download the code on this post.

download seems to be successful though, when i open the notebook, an error shows up.

Notebook validation failed

An invalid notebook may not function properly. The validation error was:

Notebook validation failed: Additional properties are not allowed ('execution_count', 'outputs' were unexpected):

why would opening a jupyter notebook get "'outputs' were unexpected" error?

is there a method i could use to examine which part of the notebook is invalid?

Answer 1

A Jupyter notebook is actually stored as a JSON file, so you can just open it in a text editor. Each cell is a JSON object that gets decoded to a Python dict.

The reason you get this error is that cell_type is markdown, which signifies that this is a Markdown cell.

Since Markdown cells are rendered instead of executed, it doesn't make sense for them to have the outputs and execution_count keys, which apply only to code cells.

You could probably write a simple script to examine your Jupyter notebook; something like this:

import json

valid_keys = ['cell_type', 'metadata', 'source']
filename = ...  # specify filename here

with open(filename) as f:
    data = json.load(f)

for index, cell in enumerate(data['cells'], 1):
    if cell['cell_type'] == 'markdown':
        extra_keys = [key for key in cell.keys() if key not in valid_keys]
        if extra_keys:    
            print(f'Cell {index} has the following keys which are invalid for a markdown cell: {extra_keys}')

Answer 2

In addition to the previous reply, if the error verifiably stems from markdown cells with empty outputs properties, you can use a regex to strip those. It won't affect any other legit outputs.
Here's a python script to do just that:

"""
This script fixes the auto-corrupted jupyter notebooks, specifically removes empty outputs properties (which are not legal for markdown cells).
"""

import re

with open('MyNotebook.ipynb', 'r') as file:
    # Read original notebook file to string
    jupyter = file.read()

    # Run a regex based search and replace, wipe all empty outputs properties.
    outputs_tag_removed = re.sub(',\n\s+\"outputs\":\ \[\]', '', jupyter)

    # Overwrite original notebook file
    print(outputs_tag_removed, file=open('MyNotebook.ipynb', 'w'))

There seem to be similar issues, related to markdown cells with illegal execution count and code cells without the mandatory execution count property. See discussion on GitHub.