Snowflake SPS-C01 : Snowflake Certified SnowPro Specialty - Snowpark

You are using Snowpark Python to analyze sales data stored in a Snowflake table named 'SALES DATA. The table has columns PRODUCT ICY, 'REGION', and 'SALE DATE. You need to calculate the total sale amount for each product in each region. You intend to use the 'group_by' and 'agg' functions. Which of the following Snowpark Python code snippets correctly performs this aggregation and renames the aggregated column to 'TOTAL SALES'? (Assume 'session' is a valid Snowpark session object.)

• A.
  
• B.
  
• C.
  
• D.
  
• E.
  

You have a Snowpark DataFrame 'df_orders' containing order data'. You want to delete all records from the underlying Snowflake table 'ORDERS TABLE' where the 'order_date' is older than '2023-01-01' using a Snowpark DataFrame operation. Which of the following code snippets is the MOST efficient and recommended way to achieve this, assuming 'spark' is your Snowpark Session object?

• A. Option D
• B. Option E
• C. Option A
• D. Option B
• E. Option C

You are developing a Snowpark application to process sales data. The application uses a UDF that calls an external Python library with a large memory footprint. After deploying the application, you observe that the Snowflake warehouse frequently runs out of memory, causing the application to fail. Which of the following strategies would be MOST effective in mitigating this issue, while minimizing cost and maintaining performance? Assume the data volume is relatively large and the UDF is computationally intensive.

• A. Implement the Python UDF as a Snowpark Stored Procedure. Deploy the UDF with the same warehouse size.
• B. Increase the warehouse size to the largest available option. This will provide more memory to the UDE
• C. Rewrite the UDF in Java using Snowpark API, which generally has a smaller memory footprint than Python. Deploy the UDF with the same warehouse size.
• D. Implement a caching mechanism within the UDF to store intermediate results and reduce the number of calls to the external library. Deploy the UDF with the same warehouse size.
• E. Modify the UDF to process data in smaller batches using a generator pattern, reducing memory consumption at any given time. Deploy the UDF with the same warehouse size.

You are using Snowpark Python to build a machine learning pipeline. One step in the pipeline involves feature engineering using a large dataset. This feature engineering step is computationally expensive and involves several transformations. You want to optimize the performance of this step by caching intermediate results. Given the following code snippet, which of the following strategies would be MOST effective for optimizing the performance, considering the use of

• A. Avoid using altogether because it can introduce overhead and is not always beneficial.
• B. Cache each intermediate DataFrame after each individual transformation step, even if the DataFrame is only used once.
• C. Cache the initial raw data DataFrame before applying any transformations.
• D. Identify DataFrames that are reused multiple times and cache them using after the transformations that generate them.
• E. Cache the final DataFrame only after all feature engineering steps are completed.

You have a Snowpark DataFrame containing sensor data'. You need to write this data to a Snowflake stage 'sensor_stage' , creating a new set of files every hour based on the 'timestamp' column (data type: Timestamp). You also want to ensure that the file names include the hour of the timestamp and are written in Avro format with Zstandard compression. The directory structure on the stage should reflect the hourly partitioning. Which of the following approaches offers the most efficient and scalable way to achieve this, while minimizing the number of files written per hour?

• A. Using scala user defined function (UDF) for write dataframe into stage in avro file format partitioned by Hour and calling it in snowpark dataframe.
• B. Create a view on top of the data and schedule a task which create file in avro with zstd compression by running the select statment with group by hour.
• C. Define a stored procedure that iterates through hourly intervals, filters the DataFrame based on the current hour, and writes the filtered DataFrame to the stage using 'df.write.format('avro').option('compression', 'zstd').mode('append').save(f'@sensor_stage/hour={current_hour}/')'
• D. Create a new DataFrame by adding an 'hour' column extracted from the 'timestamp' column. Then use 'df.write.partitionBy('hour').format('avro').option('compression', 'zstd').mode('append').save('@sensor_stage/')'.
• E. Write a Python script that connects to Snowflake, retrieves the entire DataFrame, iterates through each row, determines the hour from the 'timestamp', and writes each row to a separate Avro file named after the hour in the 'sensor_stage'