Great presentation. One question: since this is a performance-oriented library, did you consider the possibility of letting the user specify a "chunk" function

Great presentation. One question: since this is a performance-oriented library, did you consider the possibility of letting the user specify a "chunk" function that receives an array of lines, instead of a "step" function that only receives one? That way, if you are parsing a file with one million lines, you can invoke the callback ten thousand times (with a hundred lines each time) instead of a million times.