Numpy Arrays Being Converted to Strings in Testbook Tests

[AnsonSavage]

Problem Description:

When passing numpy arrays as arguments to a function in the notebook, they seem to be converted to strings, leading to TypeError exceptions.

The function I’m testing is supposed to adjust the brightness of a grayscale image (represented as a 2D numpy array).

However, during testing, it seems like the numpy array is converted to a string before being passed to the function, which results in the following error:

TypeError: unsupported operand type(s) for -: 'str' and 'int'

Minimum Breaking Example:

1. Jupyter Notebook Code (brightAdjust function):

import numpy as np

def brightAdjust(image, c):
    return np.clip(image + c, 0, 255)

2. Test Code Using Testbook:

I’m using the testbook library to test the function from the notebook. Here’s the test script I’m running:

# test_lab_2.py

from testbook import testbook
import numpy as np
import pytest

NOTEBOOK_PATH = '/path_to_notebook/notebook.ipynb'

@testbook(NOTEBOOK_PATH, execute=True)
def test_bright_adjust(tb):
    # Get reference to the brightAdjust function in the notebook
    bright_adjust_func = tb.ref('brightAdjust')

    # Create a very small grayscale image (1x1 numpy array)
    image_small = np.array([[100]], dtype=np.uint8)

    # Apply brightness adjustment with c = 50 on the small array
    adjusted_image_small = bright_adjust_func(image_small, 50)

    # Expected output for the small image: brightness increased by 50
    expected_image_small = np.array([[150]], dtype=np.uint8)

    # Assert that the adjusted small image matches the expected output
    np.testing.assert_array_equal(adjusted_image_small, expected_image_small)

3. Error Output:

Here’s the error traceback I’m getting when I run the test:

TypeError: unsupported operand type(s) for -: 'str' and 'int'

During handling of the above exception, another exception occurred:

testbook.exceptions.TestbookRuntimeError: An error occurred while executing the following cell:
------------------
brightAdjust(*("[[100]]", 50, ), **{})
------------------

It seems like the numpy array image_small is being passed as the string "[[100]]", rather than as a numpy array, causing the function to throw an error. The 50 seems to be properly being passed in.

What I’ve Tried:

• I’ve confirmed that the notebook runs as expected when I manually execute the cell with the brightAdjust function in Jupyter, so the issue seems to be specifically related to testbook.
• I’ve checked that the data I’m passing from the test script is indeed a numpy array, but somehow it’s converted to a string when sent to the notebook function.

Expected Behavior:

The numpy array should be passed as-is to the function in the notebook, not as a string.

Request:

Could anyone provide guidance on how to prevent numpy arrays from being converted to strings when passing them from the test code to the notebook function? If this is a limitation of testbook, is there a recommended workaround?

Thank you in advance for any help!

[avneetsunner]

+1: Also looks related to #147