DAA-C01 Question 12
Single answerUnstructuredA retail analytics team stores monthly product catalog PDFs and product images in an internal Snowflake stage. Analysts need to query this unstructured content in SQL to identify files, inspect metadata such as size and last modified timestamp, and then process selected files with document AI pipelines later. They want the simplest approach that does not require loading the files into relational tables first. Which solution best meets these requirements?
- A
Create a directory table on the stage and query the directory table metadata for the staged files.
- B
Use COPY INTO to load the PDFs and images into a VARIANT column, then query file metadata from the target table.
- C
Create an external table over the internal stage to expose the PDFs and images as rows with metadata columns.
- D
Use LIST in a SQL query as a table function and join it directly to downstream SQL transformations.
Show answer and explanation
Correct answer: A
Explanation
The best answer is to create and query a directory table on the stage. In Snowflake, directory tables are the primary mechanism for exposing metadata about files in a stage, which is especially useful for unstructured data workflows. This lets analysts query file-level information in SQL without loading file contents into database tables first. That aligns with the requirement to inspect metadata such as file identity and timestamps, then process selected files later using downstream unstructured-data capabilities. By contrast, COPY INTO is for ingestion, not simple metadata discovery; external tables are associated with external stages and structured/semi-structured file-query patterns; and LIST is an administrative command rather than a durable SQL-queryable metadata layer. This approach follows Snowflake best practices for staged unstructured data and staged file metadata management.
- A. Correct.
Correct. For unstructured data in stages, a directory table provides queryable metadata about staged files, including file names and related attributes used to work with unstructured content. This is the standard Snowflake approach when you want SQL access to file-level metadata without first ingesting file contents into relational columns. It works well with internal stages and supports downstream processing patterns such as document AI or scoped file URL generation.
- B. Incorrect.
Incorrect. COPY INTO is used to load data from files into tables, but that is not the simplest or most appropriate way to work with unstructured files such as PDFs and images when the requirement is to inspect staged file metadata without loading the files first. Also, unstructured binaries are not typically loaded into VARIANT for this purpose.
- C. Incorrect.
Incorrect. External tables are designed for files in external stages, primarily for semi-structured data querying patterns. They are not the preferred mechanism for querying metadata of unstructured files in an internal stage. This option is plausible because external tables do expose metadata, but it does not fit the stated storage location or recommended design.
- D. Incorrect.
Incorrect. LIST is a command for viewing files in a stage, but it is not used as a relational table source in SQL transformations in the way described. Teams sometimes confuse operational stage commands with queryable metadata objects. For repeatable SQL-based analytics workflows, a directory table is the appropriate feature.