AzureBlobFileSystem to support CertificateCredential about adlfs HOT 2 OPEN

from adlfs.

Does a CertificateCredential work with azure.storage.blob.ContainerClient?

No, as far as I can see, the ContainerClient can only be created using an "AzureNamedKeyCredential", "AzureSasCredential" or "TokenCredential".

But I actually already found a solution to read partitioned parquet with a CertificateCredential, using azure.storage.filedatalake.DataLakeServiceClient and another library called 'pyarrowfs_adlgen2' from Robin Kåveland.

What I think is interesting here, is that the constructor of the DataLakeServiceClient class also does not mention the usage of an CertificateCredential explicitly - just the three Credential options, I also mentioned above.
So the CertificateCredential might also work somehow with the adlfs lib, but since I have a working solution now, I probaly won't test this.

However, in case anyone else is facing the same issue, here the code I used:

from azure.storage.filedatalake import DataLakeServiceClient
from azure.identity import CertificateCredential
import pyarrowfs_adlgen2
import pyarrow
import pandas

credential = CertificateCredential(tenant_id, client_id, certificate_data=certificate)
datalake_service_client= DataLakeServiceClient(account_url=account_url, credential=credential)

handler = pyarrowfs_adlgen2.AccountHandler(datalake_service=datalake_service_client)
filesystem = pyarrow.fs.PyFileSystem(handler)

df = pandas.read_parquet(path=path, filesystem=filesystem)

@TomAugspurger feel free to close this issue, if no further action required from your side.

from adlfs.