[Shootout-list] Stuff (wide range of languages, external libraries)

[Bengt Kleberg]

Robert Seeger wrote:
...deleted
> My intention was just to state that I think having a wide variety of
> tests is a good thing. If some of them can't be implemented in a some
> languages, I don't see a problem in it. If that feature is important
> to you (Unicode, OOP, etc), then you'll pay attention to those tests.
> If not, you won't.

total agreement here.

...deleted
> There was a slight misinterpretation here. I didn't mean to imply that
> such unit tests needed to be written for every test. I just meant
> that, as the number and variety of unit tests needed for the code
> grows, so too does the likelihood that the Shootout test isn't
> suitable to be classified as a "micro-benchmark".

i agree, but still think that ''100 lines of code'' is simpler, and 
still sufficient, to define "micro-benchmark".


> 
> My stance is that either:
> 1) No external libraries should be allowed (my preference), or
> 3) External libraries should be allowed as they are commonly used for
> the language

1 is simpler than 3. ''commonly used for'' is vaguer than ''no''.
however, ''included with'' is not vague, and should reflect upon how the 
languiage is used. imho.


> For example, when working with Tcl, it very common to use external
> libraries written in C. In fact, most extentions for Tcl are written
> in C, since the language is designed to be easily extended with it. If
> you're not going to allow external libraries/resources as they are
> commonly used with the language, than they shouldn't be allowed at
> all.

very difficult question this. i see your point, but consider this:
in erlang the normal way is to do the system in pure erlang first. if it 
is too slow, then use the ffi to move the slowest parts to c. normally 
only very few parts of a system is implemented in c.
but if this is a benchmark in the shootout, then all erlang programs 
would be too slow and use the ffi to c.


bengt