Your changeset is much too large as you replaced all the indentation of four whitespace characters with tabulators. Additionally, having a quick glance at the code, tests are still missing and pyId is no valid PEP8 variable/parameter name.

Is the wheel package installed inside your virtual environment? Otherwise, pip will try to install numpy from source, while you can use the existing wheels if you install the wheel package beforehand.

Stefan is correct on all of those issues. Please run black . and isort . --profile black from the command line on the code for proper formatting. You can install them with pip install black and pip install isort. We also need tests for this, so make some example data and tests, and then run the tests with pytest . in the command line of the repo's home directory.

Thank you for your feedback. I will try to run the tests (I am not very experienced yet).
In the meanwhile, I confirm that the proposed implementation is compatible with Java Jackson library (https://github.com/FasterXML/jackson) - we have been able to fully code and decode a JSON file with object references exchanging it back and forth between Java and Python :)