# ------------------------------------
# Copyright (c) Microsoft Corporation.
# Licensed under the MIT License.
# ------------------------------------
import asyncio
import os
import shutil
import sys
from typing import Any, List, Optional
from azure.core.exceptions import ClientAuthenticationError
from azure.core.credentials import AccessToken, AccessTokenInfo, TokenRequestOptions
from .._internal import AsyncContextManager
from .._internal.decorators import log_get_token_async
from ... import CredentialUnavailableError
from ..._credentials.azure_cli import (
AzureCliCredential as _SyncAzureCliCredential,
CLI_NOT_FOUND,
COMMAND_LINE,
EXECUTABLE_NAME,
get_safe_working_dir,
NOT_LOGGED_IN,
parse_token,
sanitize_output,
)
from ..._internal import _scopes_to_resource, resolve_tenant, within_dac, validate_tenant_id, validate_scope
[docs]
class AzureCliCredential(AsyncContextManager):
"""Authenticates by requesting a token from the Azure CLI.
This requires previously logging in to Azure via "az login", and will use the CLI's currently logged in identity.
:keyword str tenant_id: Optional tenant to include in the token request.
:keyword List[str] additionally_allowed_tenants: Specifies tenants in addition to the specified "tenant_id"
for which the credential may acquire tokens. Add the wildcard value "*" to allow the credential to
acquire tokens for any tenant the application can access.
:keyword int process_timeout: Seconds to wait for the Azure CLI process to respond. Defaults to 10 seconds.
.. admonition:: Example:
.. literalinclude:: ../samples/credential_creation_code_snippets.py
:start-after: [START create_azure_cli_credential_async]
:end-before: [END create_azure_cli_credential_async]
:language: python
:dedent: 4
:caption: Create an AzureCliCredential.
"""
def __init__(
self,
*,
tenant_id: str = "",
additionally_allowed_tenants: Optional[List[str]] = None,
process_timeout: int = 10,
) -> None:
if tenant_id:
validate_tenant_id(tenant_id)
self.tenant_id = tenant_id
self._additionally_allowed_tenants = additionally_allowed_tenants or []
self._process_timeout = process_timeout
[docs]
@log_get_token_async
async def get_token(
self,
*scopes: str,
claims: Optional[str] = None, # pylint:disable=unused-argument
tenant_id: Optional[str] = None,
**kwargs: Any,
) -> AccessToken:
"""Request an access token for `scopes`.
This method is called automatically by Azure SDK clients. Applications calling this method directly must
also handle token caching because this credential doesn't cache the tokens it acquires.
:param str scopes: desired scope for the access token. This credential allows only one scope per request.
For more information about scopes, see
https://learn.microsoft.com/entra/identity-platform/scopes-oidc.
:keyword str claims: not used by this credential; any value provided will be ignored.
:keyword str tenant_id: optional tenant to include in the token request.
:return: An access token with the desired scopes.
:rtype: ~azure.core.credentials.AccessToken
:raises ~azure.identity.CredentialUnavailableError: the credential was unable to invoke the Azure CLI.
:raises ~azure.core.exceptions.ClientAuthenticationError: the credential invoked the Azure CLI but didn't
receive an access token.
"""
# only ProactorEventLoop supports subprocesses on Windows (and it isn't the default loop on Python < 3.8)
if sys.platform.startswith("win") and not isinstance(asyncio.get_event_loop(), asyncio.ProactorEventLoop):
return _SyncAzureCliCredential().get_token(*scopes, tenant_id=tenant_id, **kwargs)
options: TokenRequestOptions = {}
if tenant_id:
options["tenant_id"] = tenant_id
token_info = await self._get_token_base(*scopes, options=options, **kwargs)
return AccessToken(token_info.token, token_info.expires_on)
[docs]
@log_get_token_async
async def get_token_info(self, *scopes: str, options: Optional[TokenRequestOptions] = None) -> AccessTokenInfo:
"""Request an access token for `scopes`.
This is an alternative to `get_token` to enable certain scenarios that require additional properties
on the token. This method is called automatically by Azure SDK clients. Applications calling this method
directly must also handle token caching because this credential doesn't cache the tokens it acquires.
:param str scopes: desired scopes for the access token. This credential allows only one scope per request.
For more information about scopes, see https://learn.microsoft.com/entra/identity-platform/scopes-oidc.
:keyword options: A dictionary of options for the token request. Unknown options will be ignored. Optional.
:paramtype options: ~azure.core.credentials.TokenRequestOptions
:rtype: AccessTokenInfo
:return: An AccessTokenInfo instance containing information about the token.
:raises ~azure.identity.CredentialUnavailableError: the credential was unable to invoke the Azure CLI.
:raises ~azure.core.exceptions.ClientAuthenticationError: the credential invoked the Azure CLI but didn't
receive an access token.
"""
# only ProactorEventLoop supports subprocesses on Windows (and it isn't the default loop on Python < 3.8)
if sys.platform.startswith("win") and not isinstance(asyncio.get_event_loop(), asyncio.ProactorEventLoop):
return _SyncAzureCliCredential().get_token_info(*scopes, options=options)
return await self._get_token_base(*scopes, options=options)
async def _get_token_base(
self, *scopes: str, options: Optional[TokenRequestOptions] = None, **kwargs: Any
) -> AccessTokenInfo:
tenant_id = options.get("tenant_id") if options else None
if tenant_id:
validate_tenant_id(tenant_id)
for scope in scopes:
validate_scope(scope)
resource = _scopes_to_resource(*scopes)
command = COMMAND_LINE.format(resource)
tenant = resolve_tenant(
default_tenant=self.tenant_id,
tenant_id=tenant_id,
additionally_allowed_tenants=self._additionally_allowed_tenants,
**kwargs,
)
if tenant:
command += " --tenant " + tenant
output = await _run_command(command, self._process_timeout)
token = parse_token(output)
if not token:
sanitized_output = sanitize_output(output)
message = (
f"Unexpected output from Azure CLI: '{sanitized_output}'. \n"
f"To mitigate this issue, please refer to the troubleshooting guidelines here at "
f"https://aka.ms/azsdk/python/identity/azclicredential/troubleshoot."
)
if within_dac.get():
raise CredentialUnavailableError(message=message)
raise ClientAuthenticationError(message=message)
return token
[docs]
async def close(self) -> None:
"""Calling this method is unnecessary"""
async def _run_command(command: str, timeout: int) -> str:
# Ensure executable exists in PATH first. This avoids a subprocess call that would fail anyway.
if shutil.which(EXECUTABLE_NAME) is None:
raise CredentialUnavailableError(message=CLI_NOT_FOUND)
if sys.platform.startswith("win"):
args = ("cmd", "/c " + command)
else:
args = ("/bin/sh", "-c", command)
working_directory = get_safe_working_dir()
try:
proc = await asyncio.create_subprocess_exec(
*args,
stdout=asyncio.subprocess.PIPE,
stderr=asyncio.subprocess.PIPE,
stdin=asyncio.subprocess.DEVNULL,
cwd=working_directory,
env=dict(os.environ, AZURE_CORE_NO_COLOR="true"),
)
stdout_b, stderr_b = await asyncio.wait_for(proc.communicate(), timeout)
output = stdout_b.decode()
stderr = stderr_b.decode()
except asyncio.TimeoutError as ex:
proc.kill()
raise CredentialUnavailableError(message="Timed out waiting for Azure CLI") from ex
except OSError as ex:
# failed to execute 'cmd' or '/bin/sh'
error = CredentialUnavailableError(message="Failed to execute '{}'".format(args[0]))
raise error from ex
if proc.returncode == 0:
return output
# Fallback check in case the executable is not found while executing subprocess.
if proc.returncode == 127 or stderr.startswith("'az' is not recognized"):
raise CredentialUnavailableError(CLI_NOT_FOUND)
if ("az login" in stderr or "az account set" in stderr) and "AADSTS" not in stderr:
raise CredentialUnavailableError(message=NOT_LOGGED_IN)
message = sanitize_output(stderr) if stderr else "Failed to invoke Azure CLI"
if within_dac.get():
raise CredentialUnavailableError(message=message)
raise ClientAuthenticationError(message=message)