The answer I gave was instead of loading it into memory to do your processing, you would instead preprocess the file you wanted using unix command line programs and then you create a stream to the DB file and read data in from it.
Basically the answer is you don't load it all into memory. You use input/output steams to process the data one line at a time.