Analyze data in a storage account

In this tutorial, you'll learn how to analyze data located in a storage account.

Overview

So far, we've covered scenarios where data resides in databases in the workspace. Now we'll show you how to work with files in storage accounts. In this scenario, we'll use the primary storage account of the workspace and container that we specified when creating the workspace.

• The name of the storage account: contosolake
• The name of the container in the storage account: users

Create CSV and Parquet files in your storage account

Run the following code in a notebook in a new code cell. It creates a CSV file and a parquet file in the storage account.

%%pyspark
df = spark.sql("SELECT * FROM nyctaxi.passengercountstats")
df = df.repartition(1) # This ensures we'll get a single file during write()
df.write.mode("overwrite").csv("/NYCTaxi/PassengerCountStats_csvformat")
df.write.mode("overwrite").parquet("/NYCTaxi/PassengerCountStats_parquetformat")

Analyze data in a storage account

You can analyze the data in your workspace default ADLS Gen2 account or you can link an ADLS Gen2 or Blob storage account to your workspace through "Manage" > "Linked Services" > "New" (The steps below will refer to the primary ADLS Gen2 account).

1. In Synapse Studio, go to the Data hub, and then select Linked.

2. Go to Azure Data Lake Storage Gen2 > myworkspace (Primary - contosolake).

3. Select users (Primary). You should see the NYCTaxi folder. Inside you should see two folders called PassengerCountStats_csvformat and PassengerCountStats_parquetformat.

4. Open the PassengerCountStats_parquetformat folder. Inside, you'll see a parquet file with a name like part-00000-2638e00c-0790-496b-a523-578da9a15019-c000.snappy.parquet.

5. Right-click .parquet, then select New notebook, then select Load to DataFrame. A new notebook is created with a cell like this:

   %%pyspark
   abspath = 'abfss://users@contosolake.dfs.core.windows.net/NYCTaxi/PassengerCountStats_parquetformat/part-00000-1f251a58-d8ac-4972-9215-8d528d490690-c000.snappy.parquet'
   df = spark.read.load(abspath, format='parquet')
   display(df.limit(10))
   

6. Attach to the Spark pool named Spark1. Run the cell. If you run into an error related to lack of cores, this spark pool may be used by another session. Cancel all the existing sessions and retry.

7. Select back to the users folder. Right-click the .parquet file again, and then select New SQL script > SELECT TOP 100 rows. It creates a SQL script like this:

   SELECT 
       TOP 100 *
   FROM OPENROWSET(
       BULK 'https://contosolake.dfs.core.windows.net/users/NYCTaxi/PassengerCountStats_parquetformat/part-00000-1f251a58-d8ac-4972-9215-8d528d490690-c000.snappy.parquet',
       FORMAT='PARQUET'
   ) AS [result]
   

   In the script window, make sure the Connect to field is set to the Built-in serverless SQL pool.

8. Run the script.

Next steps