An opportunity to put custom SQL functions on physical datasets

[NeilPerry1]

Hello there,

Are there any plans to introduce a new feature in Superset that allows users to incorporate custom SQL functions into physical datasets? Currently, Superset only determines the column's name, but it would be beneficial to be able to use functions such as some_function('column_name'), for example, for Clickhouse's dictGet "dictGet('schema.table_name', 'dict_user_name', column_name) as column_name_1, dictGet('schema.table_name', 'dict_office_name', column_name) as column_name_2 etc".

The current approach of using virtual datasets can be suboptimal for columnar databases because it requires reading the full dataset in a subquery, resulting in the retrieval of more columns than necessary.

By incorporating the proposed feature to include custom SQL functions in physical datasets, we can minimize the usage of virtual datasets in these cases. This would optimize performance by ensuring that only the required columns are retrieved from the database, resulting in more efficient and consistent fact tables.