Nbdev_ignore

Add a #|ignore directive to hide the cell and avoid execution during testing and docs generation when using nbdev for development.

What this is

When using the nbdev library, I often find myself having to write

#| hide
#| eval: false
...

in order for that cell in Jupyter Notebook to be ignored by both the tests and docs. It would great to have a simple directive to do this, and this is the ignore directive:

#| ignore
...

It’s equivalent to the above directives.

Installation

pip install nbdev_ignore

How to use

To enable the #| ignore directive, in your settings.ini make sure to:

• Add nbdev_ignore as a requirement
• Add procs = nbdev_igore.core:ignore_, where it should point to the exact function being called
• Add it to test flags: tst_flags = ignore, so that cells with this directive also avoids testing. If you already have other test flags, separate them with space. For example, if you already have notest as your test flag, then tst_flags = notest ignore.

How it works

There’s only one function ignore_ in this module. It’s exactly the same as the hide_ processor in nbdev, so that “#| ignore” serves the same purpose as “#| hide”. For details on how it works, see here.

In order for it to also serve as a test flag, we will need to manually add it to tst_flags (see below) so tests will ignore cells with this directive. Currently there’s no simple way to hack the nbdev_test process, so this has to be done manually.

Result

We can now write #|ignore for the cell to be ignored by both tests and docs, but still kept in the notebook:

#|ignore

from pyspark.sql import SparkSession
from mock import patch

spark = (SparkSession.builder
        .config("fs.s3a.aws.credentials.provider",
                "com.amazonaws.auth.profile.ProfileCredentialsProvider")
        .appName("app_name")
        .getOrCreate()
        )

columns = ["address", "date"]
data = [("address1", "20211001")]
df = spark.createDataFrame(data).toDF(*columns)

@patch("pyspark.sql.session.DataFrameReader.load", return_value = df)
def test(spark_load_mock):
    df = fake_func(spark)
    return df

dfpd = test()