Answer: Parquet

The correct answer is Parquet (D) because Snowflake documentation explicitly states that when unloading floating-point number columns, values are truncated to approximately (15,9) in CSV and JSON formats, but are NOT truncated when unloading to Parquet files. The community discussion strongly supports this with multiple references to the official documentation and high upvote counts for Parquet answers. Avro (A) is incorrect as it's not supported for unloading in Snowflake, and CSV (B) and JSON (C) both truncate floating-point values, making them unsuitable for preserving exact data values.

Author: LeetQuiz Editorial Team