You need to migrate a SQL Server code that uses STRING_AGG function. Unfortunately, it's a SQL Server-specific feature. To migrate it, you will have to use other transformations available in Apache Spark SQL. Below is one of the examples that supports sorting on a column which is not part of the aggregation (called here sort_column):
CONCAT_WS('; ', FILTER(AGGREGATE( SORT_ARRAY(COLLECT_LIST(CONCAT_WS(' :: ', sort_column, output_column))), ARRAY(''), (string_so_far, new_entry) -> ARRAY_APPEND(string_so_far, SPLIT_PART(new_entry, ' :: ', 2))), entry -> entry != '')