SPS-C01問題集、Snowflake実際の試験問題

質問 1

You are tasked with creating a Snowpark stored procedure to perform complex data transformations using a Pandas DataFrame. You want to optimize the performance of the stored procedure by leveraging Snowpark's distributed execution capabilities. Consider the following code snippet:

Which of the following changes to the above code will significantly improve the performance by utilizing Snowpark's distributed execution?

A. Use instead of 'to_pandas()'. Then, after the transformation, use 'session.write_pandas(transformed_df, 'transformed_table')'.

B. Before calling , apply a 'limit()' function to the Snowflake Dataframe to reduce the size of the Pandas DataFrame to only a subset of the data. Then, apply the complex transformation.

C. Keep the code as is. Pandas DataFrames are always automatically distributed within Snowpark.

D. Split the Pandas DataFrame into smaller chunks and process each chunk in parallel using Python's multiprocessing library before converting back to a Snowpark DataFrame. Then save it to a table.

E. Replace with 'session.table('my_table')'. Perform the 'some_complex_transformation' directly using Snowpark DataFrame operations instead of converting to Pandas.

正解: E

解説: (PassTest メンバーにのみ表示されます)

質問 2

Consider the following Snowpark code snippet designed to create a temporary table:

A developer encounters an error when calling this function. The error message indicates that the table already exists. How should the developer modify the code to handle this scenario gracefully, preventing the error and ensuring the temporary table is either created or overwritten?

A. Add the 'mode='append" option to the function. This will append the data to the existing table.

B. Use to create the temporary table.

C. Add the 'mode='overwrite" option to the function. This will replace the existing table with the new data.

D. First drop the table using 'session.sql(fDROP TABLE IF EXISTS {table_name}')' before calling .

E. Add the 'mode='ignore" option to the function. This will silently skip the creation if the table already exists.

正解: C,D

解説: (PassTest メンバーにのみ表示されます)

質問 3

You have a complex data pipeline implemented using Snowpark Tasks in a Directed Acyclic Graph (DAG). One of the tasks, , depends on the successful completion of two parent tasks, and 'task B'. You need to implement error handling such that if 'task_R fails, 'task_C' should not be executed, but should still complete its execution regardless of status. If 'task B' fails, 'task_C' should not be executed. How do you configure the task dependencies and error handling in Snowflake to achieve this behavior?

A.

B.

C.

D.

E.

正解: C

解説: (PassTest メンバーにのみ表示されます)

質問 4

You are tasked with creating a Snowpark session that utilizes a specific Snowflake warehouse for all operations. Which of the following code snippets BEST demonstrates how to correctly specify the 'warehouse' parameter when creating a session using snowpark.Session.builder.configs'?

A.

B.

C.

D.

E.

正解: D

解説: (PassTest メンバーにのみ表示されます)

質問 5

You are developing a Snowpark application to process images stored in an internal stage. You have defined a Python UDF to detect objects in each image using a pre-trained model. The UDF takes the image file path as input and returns a JSON string containing the detected objects and their bounding boxes. However, you encounter "SerializationError' when running the UDF. Which of the following steps are MOST likely to resolve this issue effectively, assuming the model itself is correctly loaded and functions within the UDF environment?

A. Serialize the output of the UDF (the JSON string) using a custom serialization function that handles complex data types appropriately, and deserialize it in the Snowpark DataFrame.

B. Convert the image file path to the image file content using a Snowpark function such as 'snowpark.functions.read' before passing it to the UDF.

C. Reduce the size of the images before passing them to the UDF to reduce memory consumption and serialization overhead. Resize images before ingesting them.

D. Increase the value of the 'MAX MEMORY USAGE parameter for the warehouse to provide more memory for UDF execution. This will prevent running out of resources when processing large images.

E. Ensure that the Python environment used for UDF execution has the 'pillow' library installed by specifying it in the 'imports' parameter of the 'create_udf function with the corresponding packages for loading and preprocessing images.

正解: C,E

解説: (PassTest メンバーにのみ表示されます)

質問 6

You have a Snowpark DataFrame 'customer data df containing customer information, including 'customer id', 'email', and phone number'. You need to anonymize the 'email' and 'phone number" columns for customers residing in specific countries (e.g., 'USA', 'Canada') before persisting the changes back to the 'customers' table. Anonymization should replace sensitive data with 'XXXXX'. You want to leverage UDF for obfuscation. What is correct and optimal approach considering performance and security?

A. Option A

B. Option C

C. Option E

D. Option B

E. Option D

正解: C,D

解説: (PassTest メンバーにのみ表示されます)

質問 7

You have a Snowflake stage containing image files. You need to write a Snowpark Python application that extracts metadata (e.g., image resolution, format) from these images and stores the metadata in a Snowflake table. You want to leverage a Python library, such as Pillow (PIL), for image processing. Which of the following steps are necessary to correctly and efficiently implement this?

A. Use Snowpark's built-in image processing functions to extract metadata directly from the image files. This eliminates the need for external libraries like Pillow.

B. Create a Python UDF (User-Defined Function) that uses Pillow to extract metadata from the image files. Register the UDF with Snowflake. In a Snowpark DataFrame transformation, call the UDF for each image file to extract the metadata. Finally, write the resulting DataFrame to a Snowflake table.

C. Download all the image files to the Snowpark client, process them locally using Pillow, and then upload the extracted metadata to Snowflake using session

D. Upload the Pillow library as a zip file to a Snowflake internal stage. Create a Snowpark stored procedure. In the stored procedure code, import the Pillow library using 'import zipfile; sys.path.append('pillow.zip'); from PIL import Image'. Read the image files using , process them with Pillow to extract metadata, and then insert the metadata into the Snowflake table.

E. Create a Conda environment specification file ('environment.yml') that includes Pillow as a dependency. Upload the 'environment.ymr file to a Snowflake stage. Use 'session.add_packages' in the Snowpark session to load the Pillow library. Read the image files using , process them with Pillow, and then write the metadata to a Snowflake table using 'session.write_pandas()'.

正解: E

解説: (PassTest メンバーにのみ表示されます)

質問 8

Consider the following Snowpark Python code snippet that defines and registers a User-Defined Table Function (UDTF):

Which of the following statements is MOST accurate regarding the behavior and limitations of this UDTF when used in a Snowpark DataFrame transformation?

A. The UDTF can only be used with DataFrames that have been explicitly persisted as Snowflake tables.

B. If the input DataFrame column contains NULL values, the 'process' method will receive 'None' as the value for 'input_string'. The 'output_schema' correctly defines the structure of the output rows.

C. The UDTF will be executed within the same Python process as the Snowpark driver program, limiting its scalability for large datasets.

D. The UDTF will process each input string in parallel, with Snowflake automatically distributing the processing across multiple worker nodes.

E. The 'input_string' argument passed to the 'process' method will always be a single string value, even if the input DataFrame column contains NULL values.

正解: B

解説: (PassTest メンバーにのみ表示されます)

Snowflake Certified SnowPro Specialty - Snowpark - SPS-C01 模擬練習