Lazy evaluation postpones Snowpark data-frame operations until an action requires results. Batching transformations lets the server execute them together, reducing data transferred to the client and generally improving performance. section 6.5
Trigger execution with an action such as:
- collect(), which materializes rows for inspection. section 6.5
- count() or show(), which request computed results. section 6.5
- save_as_table(), which persists the data frame to Snowflake. section 6.5
Inspecting and saving a data frame
Construct a data frame from holiday_flags with an explicit day and holiday_flg schema. print(df) alone produces no data; print(df.collect()) executes the pipeline for inspection. Then use df.write.mode("overwrite").save_as_table("DIM_DATE") to persist it.
