Selectively import from another Jupyter Notebook

Question

I arranged my Jupyter notebooks into: data.ipynb, methods.ipynb and results.ipynb. How can I selectively import cells from data and methods notebooks for use in the results notebook?

I know of nbimporter and ipynb but neither of those offers selective import of variables. There is an option to import definitions - including variables that are uppercase - but this does not work for me as I would have to convert most of the variables in my notebooks to uppercase.

I would rather import everything except for two or three cells that take a long time to evaluate. Ideally, I would like to defer the execution of some assignments to the very moment I access them (lazy evaluation) - but I understand that it might be difficult to implement.

Here is the overview, in pseudocode (each line repesents a cell):

data.ipynb:

raw_data = load_data()
dataset = munge(raw_data)
describe(dataset)             # I want this line to be skipped at import

methods.ipynb:

import data
method = lambda x: x * x
# showcase how the method works on a subset of the dataset
method(data.dataset[:5])      # I want this line to be skipped at import

results.ipynb:

import data
import methods
result = methods.method(data.dataset)
describe(result)

The motivation is that my real data and methods notebooks:

• are way much longer and complicated, hence I want to use an import system
• there are only a couple of cells that take more than seconds to evaluate

also, the methods notebook cannot be replaced with methods.py file. In fact, I have such a file which contains the implementation details of my method. The notebook is more of a place to specify default parameters, showcase how my method works and explain example results.

This question is essentially a combination of:

• How can I import from another ipython-notebook?, and
• Simple way to choose which cells to run in ipython notebook during run all

I read through answers to both and none satisfied my requirements.

In my answer below I present my solution that uses custom cell magics and monkey-patching. However, I would prefer a solution which allows specifying which cells/expressions to exclude/include not in the notebook of origin (e.g. data.ipynb) but in the target one (e.g. in methods.ipynb).

For example, it could use regular expressions:

# all variables starting with 'result' would be ignored
nbimporter.options['exclude'] = '^result.*'

or (even better) lazy evaluation:

# only `a` and `b` would be evaluated and imported
from data import a, b

All ideas will be appreciated!

Answer 1

So far I've been monkey-patching nbimporter and selecting cells to exclude using cell magic:

from IPython.core import magic

@magic.register_cell_magic
def skip_on_import(args, cell):
    get_ipython().ex(cell)

The code used to monkey-patch of cell remover:

import ast

class SkippingTransformer(ast.NodeTransformer):
    # usage:
    # import nbimporter 
    # nbimporter.CellDeleter = SkippingTransformer

    def visit(self, node):
        if (
            isinstance(node, ast.Expr)
            and isinstance(node.value, ast.Call)
            and isinstance(node.value.func, ast.Attribute)
            and node.value.func.attr == 'run_cell_magic'
            and node.value.args[0].s == 'skip_on_import'
        ):
            return
        return node

And an actual example, data.ipynb:

And methods.ipynb (the exception at the end is intended - it means success!):