[jupyter-gcs-contents-manager] Reading JSON files from GCS fails when running the contents manager under Python 3.

[ojarjur]

Reproduction steps:

1. Open a terminal in Jupyter(Lab)
2. echo "{}" > test.json
3. gsutil cp test.json gs://<BUCKET>/notebooks/jupyter

Try to open file in either Jupyter or JupyterLab

This looks like a bug in the way we handle files in GCS with a content type other than test/* when the contents manager is running in Python3.

As a work-around, you can specify a content type of text/plain when uploading the file to GCS, e.g.:

gsutil -h "Content-Type:text/plain" cp <FILE>.json gs://<BUCKET>/notebooks/jupyter/

Then the file will open in Jupyter and JupyterLab.

The root cause is this line

On that line we base64 encode the content before sticking it in an object that we later serialize to JSON. However, with Python3 that method returns a type of bytes instead of string, and Python JSON serialization does not support the type bytes.

The fix is to convert the response to a string before sticking it in the object. In short, we need to change that line to say

      blob_obj['content'] = base64.b64encode(content_str).decode(utf8_encoding)