Data corruption when copying an Oracle table with number fields

This has been a struggle for me for a few weeks too. I've created my own workaround for this problem. It took some time to finetune, but it works for me. Hope it helps you too.
First, I created a lookup that retrieves all the column names from the table in Oracle. The table and schema are pipeline parameters in my case, but can be hardcoded or retrieved in another way too.

The Query for the lookup is (I use XMLAGG instead of other options to mitigate the problem that the returned string was too long) :

@concat('SELECT RTRIM(XMLAGG(XMLELEMENT(E,CASE WHEN DATA_TYPE = ''NUMBER'' AND (CAST (DATA_SCALE AS NUMBER(5,0)) >= CAST (DATA_PRECISION AS NUMBER(5,0)) OR DATA_SCALE IS NULL OR DATA_PRECISION IS NULL) THEN ''CAST('' || COLUMN_NAME || '' AS NUMBER(38,18)) AS '' || COLUMN_NAME ELSE COLUMN_NAME END,'', '').EXTRACT(''//text()'') ORDER BY COLUMN_NAME).GetClobVal(),'', '') AS query_string
 FROM ALL_TAB_COLUMNS
 WHERE TABLE_NAME = ''', pipeline().parameters.table_name, '''
 AND OWNER = ''', pipeline().parameters.schema, '''')

 Then I use these retrieved columns in my copy data:

And the query for the source is:

@concat('SELECT ',activity('lookup_columns_from_table').output.firstRow.QUERY_STRING, ' FROM ', pipeline().parameters.schema, '.', pipeline().parameters.table_name)