Tutorial: Create a Jupyter Notebook in Azure Cosmos DB for NoSQL to analyze and visualize data (preview)

APPLIES TO: NoSQL

Warning

The Jupyter Notebooks feature of Azure Cosmos DB will be retired March 30, 2024; you will not be able to use built-in Jupyter notebooks from the Azure Cosmos DB account. We recommend using Visual Studio Code's support for Jupyter notebooks or your preferred notebooks client.

This tutorial walks through how to use the Jupyter Notebooks feature of Azure Cosmos DB to import sample retail data to an Azure Cosmos DB for NoSQL account. You'll see how to use the Azure Cosmos DB magic commands to run queries, analyze the data, and visualize the results.

Prerequisites

• An existing Azure Cosmos DB for NoSQL account.
  • If you have an existing Azure subscription, create a new account.
  • No Azure subscription? You can try Azure Cosmos DB free with no credit card required.

Create a new notebook

In this section, you'll create the Azure Cosmos database, container, and import the retail data to the container.

1. Navigate to your Azure Cosmos DB account and open the Data Explorer.

2. Select New Notebook.

   

3. In the confirmation dialog that appears, select Create.

   Note

   A temporary workspace will be created to enable you to work with Jupyter Notebooks. When the session expires, any notebooks in the workspace will be removed.

4. Select the kernel you wish to use for the notebook.

Python

C#

Tip

Now that the new notebook has been created, you can rename it to something like VisualizeRetailData.ipynb.

Create a database and container using the SDK

Python

1. Start in the default code cell.

2. Import any packages you require for this tutorial.

   import azure.cosmos
   from azure.cosmos.partition_key import PartitionKey
   

3. Create a database named RetailIngest using the built-in SDK.

   database = cosmos_client.create_database_if_not_exists('RetailIngest')
   

4. Create a container named WebsiteMetrics with a partition key of /CartID.

   container = database.create_container_if_not_exists(id='WebsiteMetrics', partition_key=PartitionKey(path='/CartID'))
   

5. Select Run to create the database and container resource.

   

C#

1. Start in the default code cell.

2. Import any packages you require for this tutorial.

   using Microsoft.Azure.Cosmos;
   

3. Create a new instance of the client type using the built-in SDK.

   var cosmosClient = new CosmosClient(Cosmos.Endpoint, Cosmos.Key);
   

4. Create a database named RetailIngest.

   Database database = await cosmosClient.CreateDatabaseIfNotExistsAsync("RetailIngest");
   

5. Create a container named WebsiteMetrics with a partition key of /CartID.

   Container container = await database.CreateContainerIfNotExistsAsync("WebsiteMetrics", "/CartID");
   

6. Select Run to create the database and container resource.

   

Import data using magic commands

1. Add a new code cell.

2. Within the code cell, add the following magic command to upload, to your existing container, the JSON data from this url: https://cosmosnotebooksdata.blob.core.windows.net/notebookdata/websiteData.json

   %%upload --databaseName RetailIngest --containerName WebsiteMetrics --url https://cosmosnotebooksdata.blob.core.windows.net/notebookdata/websiteData.json
   

3. Select Run Active Cell to only run the command in this specific cell.

   

   Note

   The import command should take 5-10 seconds to complete.

4. Observe the output from the run command. Ensure that 2,654 documents were imported.

   Documents successfully uploaded to WebsiteMetrics
   Total number of documents imported:
     Success: 2654
     Failure: 0
   Total time taken : 00:00:04 hours
   Total RUs consumed : 27309.660000001593
   

Visualize your data

Python

1. Create another new code cell.

2. In the code cell, use a SQL query to populate a Pandas DataFrame.

   %%sql --database RetailIngest --container WebsiteMetrics --output df_cosmos
   SELECT c.Action, c.Price as ItemRevenue, c.Country, c.Item FROM c
   

3. Select Run Active Cell to only run the command in this specific cell.

4. Create another new code cell.

5. In the code cell, output the top 10 items from the dataframe.

   df_cosmos.head(10)
   

6. Select Run Active Cell to only run the command in this specific cell.

7. Observe the output of running the command.

   	Action	ItemRevenue	Country	Item
   0	Purchased	19.99	Macedonia	Button-Up Shirt
   1	Viewed	12.00	Papua New Guinea	Necklace
   2	Viewed	25.00	Slovakia (Slovak Republic)	Cardigan Sweater
   3	Purchased	14.00	Senegal	Flip Flop Shoes
   4	Viewed	50.00	Panama	Denim Shorts
   5	Viewed	14.00	Senegal	Flip Flop Shoes
   6	Added	14.00	Senegal	Flip Flop Shoes
   7	Added	50.00	Panama	Denim Shorts
   8	Purchased	33.00	Palestinian Territory	Red Top
   9	Viewed	30.00	Malta	Green Sweater

8. Create another new code cell.

9. In the code cell, import the pandas package to customize the output of the dataframe.

   import pandas as pd
   pd.options.display.html.table_schema = True
   pd.options.display.max_rows = None
   
   df_cosmos.groupby("Item").size()
   

10. Select Run Active Cell to only run the command in this specific cell.

11. In the output, select the Line Chart option to view a different visualization of the data.

    

C#

1. Create a new code cell.

2. In the code cell, create a new C# class to represent an item in the container.

   public class Record
   {
     public string Action { get; set; }
     public decimal Price { get; set; }
     public string Country { get; set; }
     public string Item { get; set; }
   }
   

3. Create another new code cell.

4. In the code cell, add code to execute a SQL query using the SDK storing the output of the query in a variable of type List<T> named results.

   using System.Collections.Generic;
   
   var query = new QueryDefinition(
       query: "SELECT c.Action, c.Price, c.Country, c.Item FROM c"
   );
   
   FeedIterator<Record> feed = container.GetItemQueryIterator<Record>(
       queryDefinition: query
   );
   
   var results = new List<Record>();
   while (feed.HasMoreResults)
   {
       FeedResponse<Record> response = await feed.ReadNextAsync();
       foreach (Record result in response)
       {
           results.Add(result);
       }
   }
   

5. Create another new code cell.

6. In the code cell, create a dictionary by adding unique permutations of the Item field as the key and the data in the Price field as the value.

   var dictionary = new Dictionary<string, decimal>();
   
   foreach(var result in results)
   {
       dictionary.TryAdd (result.Item, result.Price); 
   }
   
   dictionary
   

7. Select Run Active Cell to only run the command in this specific cell.

8. Observe the output with unique combinations of the Item and Price fields.

   ...
   Denim Jacket:31.99
   Fleece Jacket:65
   Sandals:12
   Socks:3.75
   Sandal:35.5
   Light Jeans:80
   ...
   

9. Create another new code cell.

10. In the code cell, output the results variable.

    results
    

11. Select Run Active Cell to only run the command in this specific cell.

12. In the output, select the Box Plot option to view a different visualization of the data.

    

Persist your notebook

1. In the Notebooks section, open the context menu for the notebook you created for this tutorial and select Download.

   

   Tip

   To save your work permanently, save your notebooks to a GitHub repository or download the notebooks to your local machine before the session ends.

Next steps

• Learn about the Jupyter Notebooks feature in Azure Cosmos DB
• Import notebooks from GitHub into an Azure Cosmos DB for NoSQL account
• Review the FAQ on Jupyter Notebook support