ably_rest.auth.request_token capabilities call - unnecessary json.dumps() ?

[NgyAnthony]

Following the tutorial at https://ably.com/docs/auth/capabilities:

# Token request that specifies capabilities:
capabilities = {
    "chat:bob": ["subscribe"],  # only "subscribe" intersects
    "status": ["*"],  # "*" intersects with "subscribe"
    "secret": ["publish", "subscribe"]  # key does not have access to "secret" channel
}


token_details = await ably_rest.auth.request_token({
    'capability': json.dumps(capabilities)
})

Will result in the following error:

  File "/usr/local/lib/python3.12/site-packages/ably/rest/auth.py", line 272, in create_token_request
    token_request['capability'] = str(Capability(capability))
                                      ^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/site-packages/ably/types/capability.py", line 13, in __init__
    self.__dict = dict(obj)
                  ^^^^^^^^^
ValueError: dictionary update sequence element #0 has length 1; 2 is required

This is due to the following init form Capability:

class Capability(MutableMapping):
    def __init__(self, obj=None):
        if obj is None:
            obj = {}
        self.__dict = dict(obj)
        for k, v in obj.items():
            self[k] = v

Unwrapping capabilities from the suggested json.dumps() solves the issue

┆Issue is synchronized with this Jira Task by Unito

[VeskeR]

Hi @NgyAnthony !

Thank you for reporting this issue.

The SDK is intended to be able to accept a JSON string for the capability parameter (as well as a dictionary) when creating token details. The fact that it cannot process json.dumps(capabilities) is a bug. We will fix this in the library.

In the meantime, you can pass a dictionary directly when creating a token, just as you suggested:

token_details = await ably_rest.auth.request_token({
    'capability': capabilities
})