[wglb]

Actually it is a good metric, certainly to the first order. Yes, you can have a line or three that is more complex than the rest but practically it isn't going to reduce the line count that much.

And token counts don't help as code that insists that each brace must be on its own line detracts from readability. For one thing it pushes the last bit of the function off the bottom of the screen meaning you have to scroll.

A line that is overly complex is eventually get rewritten.

I say this as someone who has written large bodies of code in sigma 5 assembly, Fortran II and IV bliss 36, C, C++, and Lisp. Perhaps more to the point, these days I read large bodies of code measured in millions. Lines of code dictates how long it will take to understand it.

Peter Norvig in paip gives some examples of small code and how it can be exceedingly clear.