Re: empty-line-p

[Stefan Monnier <monnier@iro.umontreal.ca>]

>> >  (if (looking-at "^[ \t\f\r]*$")
>> >      t
>> >    nil)))
>> 
>> (if <foo> t nil)
>> 
>> is an eta-regexp, which can be advantageously reduced to just <foo>.
>> 
>> 
>> Stefan
>> 
>> 
>> PS: The classical eta-regexp is (lambda (x) (<foo> x)), which is equivalent to
>> just <foo>.  Another variant is (cons (car <foo>) (cdr <foo>)), which
>> eta-reduces to just <foo>.

> That last one is different, because the new cons is not eq to the 
> original <foo>.

Indeed, the eta rules are generally not supported 100% by most languages:
there are always some "minor" semantic differences.  Aside from the problem
you mentioned for the `cons' case, the `if' case changes the result when the
tested value can be non-t and non-nil, and the `lambda' case can change the
behavior when the function is called with more than 1 argument.

So most compilers don't apply such "optimizations", which makes it that much
more valuable to apply them by hand.


        Stefan