Quickstart: Use Azure Cosmos DB for NoSQL with Azure SDK for Python

Article
12/06/2024
Applies to:

✅ NoSQL

In this quickstart, you deploy a basic Azure Cosmos DB for Table application using the Azure SDK for Python. Azure Cosmos DB for Table is a schemaless data store allowing applications to store structured table data in the cloud. You learn how to create tables, rows, and perform basic tasks within your Azure Cosmos DB resource using the Azure SDK for Python.

API reference documentation | Library source code | Package (PyPI) | Azure Developer CLI

Prerequisites

Azure Developer CLI
Docker Desktop
Python 3.12

If you don't have an Azure account, create a Trial before you begin.

Initialize the project

Use the Azure Developer CLI (azd) to create an Azure Cosmos DB for Table account and deploy a containerized sample application. The sample application uses the client library to manage, create, read, and query sample data.

Open a terminal in an empty directory.
If you're not already authenticated, authenticate to the Azure Developer CLI using azd auth login. Follow the steps specified by the tool to authenticate to the CLI using your preferred Azure credentials.
```
azd auth login
```

Use azd init to initialize the project.

azd init --template cosmos-db-nosql-python-quickstart

During initialization, configure a unique environment name.
Deploy the Azure Cosmos DB account using azd up. The Bicep templates also deploy a sample web application.
```
azd up
```
During the provisioning process, select your subscription, desired location, and target resource group. Wait for the provisioning process to complete. The process can take approximately five minutes.

Once the provisioning of your Azure resources is done, a URL to the running web application is included in the output.

Deploying services (azd deploy)

  (✓) Done: Deploying service web
- Endpoint: <https://[container-app-sub-domain].azurecontainerapps.io>

SUCCESS: Your application was provisioned and deployed to Azure in 5 minutes 0 seconds.

Use the URL in the console to navigate to your web application in the browser. Observe the output of the running app.

Screenshot of the running web application.

Install the client library

The client library is available through the Python Package Index, as the azure-cosmos library.

Open a terminal and navigate to the /src folder.
```
cd ./src
```
If not already installed, install the azure-cosmos package using pip install.
```
pip install azure-cosmos
```
Also, install the azure-identity package if not already installed.
```
pip install azure-identity
```
Open and review the src/requirements.txt file to validate that the azure-cosmos and azure-identity entries both exist.

Object model

Name	Description
`CosmosClient`	This class is the primary client class and is used to manage account-wide metadata or databases.
`DatabaseProxy`	This class represents a database within the account.
`ContainerProxy`	This class is primarily used to perform read, update, and delete operations on either the container or the items stored within the container.
`PartitionKey`	This class represents a logical partition key. This class is required for many common operations and queries.

The sample code in the template uses a database named cosmicworks and container named products. The products container contains details such as name, category, quantity, a unique identifier, and a sale flag for each product. The container uses the /category property as a logical partition key.

Authenticate the client

This sample creates a new instance of the CosmosClient type and authenticates using a DefaultAzureCredential instance.

credential = DefaultAzureCredential()

client = CosmosClient(url="<azure-cosmos-db-nosql-account-endpoint>", credential=credential)

Get a database

Use client.get_database_client to retrieve the existing database named cosmicworks.

database = client.get_database_client("cosmicworks")

Get a container

Retrieve the existing products container using database.get_container_client.

container = database.get_container_client("products")

Create an item

Build a new object with all of the members you want to serialize into JSON. In this example, the type has a unique identifier, and fields for category, name, quantity, price, and sale. Create an item in the container using container.upsert_item. This method "upserts" the item effectively replacing the item if it already exists.

new_item = {
    "id": "aaaaaaaa-0000-1111-2222-bbbbbbbbbbbb",
    "category": "gear-surf-surfboards",
    "name": "Yamba Surfboard",
    "quantity": 12,
    "sale": False,
}

created_item = container.upsert_item(new_item)

Read an item

Perform a point read operation by using both the unique identifier (id) and partition key fields. Use container.read_item to efficiently retrieve the specific item.

existing_item = container.read_item(
    item="aaaaaaaa-0000-1111-2222-bbbbbbbbbbbb",
    partition_key="gear-surf-surfboards",
)

Query items

Perform a query over multiple items in a container using container.GetItemQueryIterator. Find all items within a specified category using this parameterized query:

SELECT * FROM products p WHERE p.category = @category

queryText = "SELECT * FROM products p WHERE p.category = @category"

results = container.query_items(
    query=queryText,
    parameters=[
        dict(
            name="@category",
            value="gear-surf-surfboards",
        )
    ],
    enable_cross_partition_query=False,
)

Loop through the results of the query.

items = [item for item in results]

output = json.dumps(items, indent=True)

Clean up resources

When you no longer need the sample application or resources, remove the corresponding deployment and all resources.

azd down