My example would be where I would have a statement selecting products, with various information and url paths. The url path I would pass through the google analytics parser, and when defining the parser I would provide the column as the first parameter, which I will handle as a filter in my extension - something like the below - based on dbr.ga triggering the column extension
select 'dbr.ga', 'page_views','<ga:id>,'<ga:metric>','<start-date>','<end-date>';
select sku, product_name, sum(row_total), url_path as pageviews, count(qty) as sold_items, sold_items / pageviews as conversion_rate from order_items
where ...
In the above example the intention would be to replace the url_path with the pageviews from google analytics for that url_path.
The initial idea of preprocessing the data before it's passed to the table parser might be more flexible though - but I can see various use cases for both