Trying this on my own lexer, increasing the channel buffer size from 10 (which is what it is now) to 512 almost doubles the runtime speed(!) I tried some other

Trying this on my own lexer, increasing the channel buffer size from 10 (which is what it is now) to 512 almost doubles the runtime speed(!) I tried some other values, and it seems that about 5 to 15 is the "optimal size". The input file the author uses is about 1M in size, and most of my inputs are much smaller (which is realistic for my use case); my feeling is that it's due to this, but I didn't verify; I'll have to check later and maybe set a more dynamic size based on the size of the input. The lesson here is what we should (hopefully!) all know already: benchmark these changes because what works brilliant in situation A may be detrimental in situation B.