I am a bit surprised that the author didn't try to implement a stream parser. This could avoid loading the entire file in memory or relying on OS features like memory-mapped files.