aio-cosmos

Asyncio SDK for Azure Cosmos DB. This library is intended to be a very thin asyncio wrapper around the Azure Comsos DB Rest API. It is not intended to have feature parity with the Microsoft Azure SDKs but to provide async versions of the most commonly used interfaces.

Feature Support

Databases

✅ List
✅ Create
✅ Delete

Containers

✅ Create
✅ Delete

Documents

✅ Create Single
✅ Create Concurrent Multiple
✅ Delete
✅ Get
✅ Query

Limitations

The library currently only supports Session level consistency, this may change in the future. For concurrent writes the maximum concurrency level is based on a maximum of 100 concurrent connections from the underlying aiohttp library. This may be exposed to the as a client setting in a future version.

Sessions are managed automatically for document operations. The session token is returned in the result so it is possible to manage sessions manually by providing this value in session_token to the appropriate methods. This facilitates sending the token value back to an end client in a session cookie so that writes and reads can maintain consistency across multiple instances of Cosmos.

There is currently no retry policy on failed connections/broken connections and this must be entirely managed by the end user code. This may be implemented in the future

Installation

pip install aio-cosmos

Usage

Client Setup and Basic Usage

The client can be instantiated using either the context manager as below or directly using the CosmosClient class. If using the CosmosClient class directly the user is responsible for calling the .connect() and .close() methods to ensure the client is boot-strapped and resources released at the appropriate times.

from aio_cosmos.client import get_client

async with get_client(endpoint, key) as client:
    await client.create_database('database-name')
    await client.create_container('database-name', 'container-name', '/partition_key_document_path')
    doc_id = str(uuid4())
    res = await client.create_document(f'database-name', 'container-name',
                                       {'id': doc_id, 'partition_key_document_path': 'Account-1', 'description': 'tax surcharge'}, partition_key="Account-1")

Querying Documents

Documents can be queried using the query_documents method on the client. This method returns an AsyncGenerator and should be used in an async for statement as below. The generator automatically handles paging for large datasets. If you don't wish to iterate through the results use a list comprehension to collate all of them.

async for doc in client.query_documents(f'database-name', 'container-name',
                                        query="select * from r where r.account = 'Account-1'",
                                        partition_key="Account-1"):
    print(f'doc returned by query: {doc}')

Concurrent Writes / Multiple Documents

The client provides the ability to issue concurrent document writes using asyncio/aiohttp. Each document is represented by a tuple of (document, partition key value) as below.

docs = [
    ({'id': str(uuid4()), 'account': 'Account-1', 'description': 'invoice paid'}, 'Account-1'),
    ({'id': str(uuid4()), 'account': 'Account-1', 'description': 'VAT remitted'}, 'Account-1'),
    ({'id': str(uuid4()), 'account': 'Account-1', 'description': 'interest paid'}, 'Account-1'),
    ({'id': str(uuid4()), 'account': 'Account-2', 'description': 'annual fees'}, 'Account-2'),
    ({'id': str(uuid4()), 'account': 'Account-2', 'description': 'commission'}, 'Account-2'),
]

res = await client.create_documents(f'database-name', 'container-name', docs)

Results

Results are returned in a dictionary with the following format:

{
    'status': str,
    'code': int,
    'session_token': Optional[str],
    'error': Optional[str],
    'data': Union[dict,list]
}

• status will be either 'ok' or 'failed'
• code is the integer HTTP response code
• session_token is the string session code vector returned by Cosmos
• error is a string error message to provide context to a failed status
• data is the direct JSON response from Cosmos and will contain any error information in the case of failed operations

Note, to see an error return in the above format you must pass raise_on_failure=False to the client constructor.