Modify the REFRESH TABLE METADATA command to specify selected columns which are deemed interesting in some form - either sorted/partitioned/clustered by and column metadata will be stored only for those columns. The proposed syntax is
REFRESH TABLE METADATA [ COLUMNS (list of columns) / NONE ] <table path>
For example, REFRESH TABLE METADATA COLUMNS (age, salary) `/tmp/employee` stores column metadata only for the age and salary columns. REFRESH TABLE METADATA COLUMNS NONE `/tmp/employee` will not store column metadata for any column.
By default, if the optional 'COLUMNS' clause is omitted, column metadata is collected for all the columns.