Have a look at QSV.

I would select the columns, pipe through either a search (if simple regex will do or luau filter if more complicated stuff is required and then redirect to the new file name for R to read.

Very fast. Has an index feature which could speed the search up even more.