AzureBlobFileSystem returns UnicodeError: Label too long

When instantiating an AzureBlobFileSystem only with account, container and sas_token. I receive an codec error when trying to basically run ls in the container.

What happened:

Traceback (most recent call last):
  File "C:\Users\bbk63\Anaconda3\lib\encodings\idna.py", line 167, in encode
    raise UnicodeError("label too long")
UnicodeError: label too long
...
UnicodeError: encoding with 'idna' codec failed (UnicodeError: label too long)

What you expected to happen: Returns a list of path inside the container

Minimal Complete Verifiable Example:

abfs = AzureBlobFileSystem(account_name=ACCOUNT_NAME, sas_token=sas_token, container_name=CONTAINER_NAME)
abfs.ls('*')

Environment:

• Dask version: 2021.2.0
• Python version: 3.6
• Operating System: Windows 10 Enterprise
• Install method (conda, pip, source): pip

[hayesgb]

Thanks.

adfs completes the URI by appending the sas token to the account url, which highlights a shortcoming in the approach.

Currently, the user passes ACCOUNT_NAME, and the token would be created as: https://{self.account_name}.blob.core.windows.net + sas_token

But it needs to conform to: https://{self.account_name}.blob.core.windows.net + "/" + {container_name} + "/" + {blob} + "?" + sas_token

The most straight solution would be to allow the end-user to create and pass the full URI (with SAS) and passed to it to adlfs, requiring account_name to be None.

Thoughts?

[hayesgb]

I’ve added a check for the leading “?” in the sas_token, as mentioned above.