[blixt]

I guess you could take all lines in the file except the one you're testing and measure the filesize, then add the line and measure again. The delta should then be more fair. You could even do this by concatenating all code files and then testing line by line across the entire repo, but that would probably be too slow.