SnowPro Associate: Platform Question 290
Single answer3.4 Explain how to work with unstructured data.A media company stores product manuals and warranty PDFs in an Amazon S3 bucket and wants to make them available through Snowflake for centralized governance and programmatic access. The company does not need to parse the file contents into relational columns yet, but it must be able to reference the files securely from SQL. Which approach should the Snowflake administrator use?
- A
Create an external stage that points to the S3 bucket and use directory tables and scoped file URLs to reference the unstructured files.
- B
Load each PDF into a VARIANT column with COPY INTO so Snowflake can manage the files as semi-structured data.
- C
Create an external function to call Amazon S3 directly, because unstructured data can only be accessed in Snowflake through external functions.
- D
Convert the PDFs to CSV before loading them into an internal stage, because Snowflake stages do not support unstructured file types.
Show answer and explanation
Correct answer: A
Explanation
The best answer is to use an external stage on the S3 bucket and work with Snowflake features designed for unstructured data. In Snowflake, unstructured data files can be stored in internal or external stages. Directory tables expose file-level metadata for staged files, and scoped file URLs provide secure, time-limited access to specific files. This allows organizations to govern access through Snowflake without immediately extracting file contents into relational tables. By contrast, VARIANT is intended for semi-structured formats, not arbitrary binary documents like PDFs. External functions are optional for downstream processing, not the primary access method. This aligns with Snowflake documentation on unstructured data, stages, directory tables, and file URL access patterns.
- A. Correct.
Correct. Snowflake supports unstructured data in stages, including external stages that point to cloud storage such as Amazon S3. Directory tables can provide metadata about staged files, and scoped file URLs can be generated for secure access to individual files. This is the appropriate approach when the goal is governed access to files without first transforming their contents into table columns.
- B. Incorrect.
Incorrect. VARIANT is designed for semi-structured data such as JSON, Avro, ORC, Parquet, and XML. A PDF is unstructured binary content and is not loaded into a VARIANT column using COPY INTO in the way described. This option reflects the common misconception that all non-tabular data belongs in VARIANT.
- C. Incorrect.
Incorrect. External functions are used to invoke remote services from Snowflake, not as the required mechanism for storing or referencing unstructured files. Snowflake can manage references to unstructured data directly through stages, directory tables, and file URLs. An external function might be used later for file processing, but it is not required for basic governed access.
- D. Incorrect.
Incorrect. Snowflake stages can store and reference unstructured file types, including documents, images, and other binaries. There is no requirement to convert PDFs to CSV. This option is based on the misconception that stages are only for structured or semi-structured ingestion workflows.