A schema specifies a dataset’s fields and data types before structured data is loaded. In Snowpark, StructType and StructField define that contract for CSV input. section 6.6
A practical schema records:
- Field names such as Customer, Order_date, Delivery_date, Baked_good_type, and Quantity. section 6.6
- Field types including StringType, DateType, and DecimalType. section 6.6
- The mapping used when reading staged data into a data frame and copying it into a staging table. section 6.6
Defining and applying a CSV schema
Create a StructType containing one StructField per CSV column, then pass it to session.read.schema(schema_for_csv).csv("@orders_stage") before copying the data into ORDERS_STG.
Schemas are explicit for structured CSV data but are not required in the same way for semistructured JSON. Snowflake can store changing or nested JSON in VARIANT before flattening it into relational columns. section 4.1.4
