CACHE SELECT column_name[, column_name, ...] FROM [db_name.]table_name [ WHERE boolean_expression ]
Cache the data accessed by the specified simple
SELECT query in the Optimizing Performance with Caching. You can choose a subset of columns to be cached by providing a list of column names and choose a subset of rows by providing a predicate. This enables subsequent queries to avoid scanning the original files as much as possible. This construct is applicable only to Parquet tables. Views are also supported, but the expanded queries are restricted to the simple queries, as described above.
See Delta and Apache Spark caching comparison for the differences between the RDD cache and the Databricks IO cache.
CACHE SELECT * FROM boxes CACHE SELECT width, length FROM boxes WHERE height=3