Adjusting token cache format

[rayluo]

• Change scope's internal representation from a list to a string
• Remove client_info from RefreshToken because it was optional and now obsolete
• Add test case to ensure the token cache representation complies with the Unified Schema
• The existing token cache implementation is already extendable, we only need to add a test case to assert that.

You can install this branch for your testing:

pip install git+https://github.com/AzureAD/microsoft-authentication-library-for-python.git@target-in-string

[abhidnya13]

[henrik-me]

int [](start = 34, length = 3)

it seems like expires_in is already an int? Did -now change it to something else? Should that then have been converted to an int?

[rayluo]

Good question! It is true that the expires_in in the original server-side response is an integer. But here in this code path we hit a token in the cache, and then re-calculate a new expires_in based on its original value and the current time(), which happens to be a float in Python. So we do a Forceful Unobvious Conversion to Keep the Integer Time, at the exact place where it is needed.

[henrik-me]

        logger.debug(

should this be info or error instead of debug? (not sure how the python logger works though)

---

Refers to: msal/application.py:322 in 9110eca. [](commit_id = 9110eca, deletion_comment = False)

[henrik-me]

is target scopes? if yes, do you have a place where you remove dupes?

[rayluo]

Yes target is exactly scope in Unified Schema's dialect.

The set(...) method creates a native set object in Python. It automatically deduplicates the scopes.

[henrik-me]