A python package providing an easy way to explicitly import Jupyter Notebooks files (.ipynb) the same way you would import regular .py files.

This is a fork of ipynb that adds support for importing ipynb file as a function, which enables you to start your next experiment as soon as you finish your current one.

You can install with:

pip install ipynb-function

You can use your ipython notebook directly as a function, just prefix it with ipynb.fs.function.

If you have a notebook file named transform.ipynb, and its content is:

import cv2

#param The input image <-- annotate your inputs (the following line) in comment.
img = cv2.imread(...)

#skip <-- annotate anything that prints the intermediate value so that the following line will be ignored.
print(img.shape)

#return The output image <-- annotate your return values (the folllowing line) in comment.
result = do.something.with(img)

You can do:

from ipynb.fs.function.transform import transform

The ipynb file will be turned into:

def transform(img=None):
    """
    Any markdown cell before the first line of code comes here.

    :param img: The input image
    :return: The output image
    """
    import cv2

    if img is none:
        img = cv2.imread(...)

    return do.something.with(img)

The function name is the same as function name.

You can do a 'full' import - this has the same semantics of importing a .py file. All the code in the .ipynb file is executed, and classes/functions/variables in the top level are available for use.

If you have a notebook file named server.ipynb, you can import it via:

import ipynb.fs.full.server

You can use the from ... import .. too.

from ipynb.fs.full.server import X, Y, X

Sometimes your notebook has been used as a way to run an analysis or other computation, and you only want to import the functions / classes defined in it - and not the extra statements you have in there. This can be accomplished via ipynb.fs.defs.

If you have a notebook file named server.ipynb, and do:

import ipynb.fs.defs.server

It'll only execute and make available the following parts of the code in server.ipynb:

• class definitions
• def function definitions
• import statements
• Assignment statements where the variables being assigned to are ALL_CAPS. These are assumed to be constants.

This skips most computational work and brings in your definitions only, making it easy to reuse functions / classes across similar analyses.

You can also easily do relative imports, both for full notebooks or for definitions only. This works inside notebooks too.

If you have a notebook called notebook1.ipynb in the same dir as your current notebook, you can import it with:

import ipynb.fs  # Boilerplate required

# Do a full import
from .full.notebook1 import foo

# Do a definitions-only import
from .defs.notebook1 import bar

This works transitively nicely - other code can import your notebook that's using relative imports and it'll all work well.