[Shootout-list] Re: Total Characters vs Lines of Code

[Bengt Kleberg]

Peter Hinely wrote:
...deleted
> I don't think I like that idea. It would penalize languages that have
> more verbose (and in some people's opinion a more readable) style.

seconded.


> Maybe what we really want is some measure of the number of tokens of an
> entry, but it's not practical to implement.  I think lines of code, while
> a crude measure, may be hard to improve on.

tokes are better, but difficult. we could fake tokens by looking at 
identifiers (any string of only letters, numbers and underscore 
(probably '-' needs to be added since it is heavily used by scheme)) and 
  other characters (but not white space. ofcourse the language using 
white space as source would get good numbers :-). not as good as tokens. 
perhaps better than lines of code.


imho it is better to have a crude measurement, than none at all. there 
are many pitfalls in timing tests, but we still try with them.


we already have manual inspection to catch (unintentional) cheating for 
performance. we could do the same with loc.



bengt