[Haskell-cafe] Moving the messages for compiler to another file?

[Alexander Dunlap]

On Tue, Dec 2, 2008 at 2:56 AM, nml <arumakanil at gmail.com> wrote:
> How about moving the messages for compiler to an additional file?
>
> My motivation is that we often face a trade-off between
> aesthetical(elegant code) and practical(efficient code).
> Like pragmas and strictness annotations, I often feel they make the
> source code ugly.
> They don't affect the semantics of our programs, do they?
>
> Some people would say this beautifying should be accomplished by an
> editor, like hiding/showing option for those information.
>
> But such separation has additional benefits.
> For instance, making the source code more compiler-independent.
> (yeah, this is not the case with language-extensions)
> And we avoid dispersing those information among our lines. (or even files)
> In some cases, it would be convenient.
>
> I'm not sure if this idea is reasonable, reachable or just naive.
> Suggestions?
>
> Best regards
> -nml
> _______________________________________________
> Haskell-Cafe mailing list
> Haskell-Cafe at haskell.org
> http://www.haskell.org/mailman/listinfo/haskell-cafe
>

One problem is that in many cases, the things that you're talking
about DO affect semantics. Consider strictness annotations:

> f (!x) = Just x
> g x = Just x
> h (Just _) = True
> h Nothing = False

Now h (f _|_) == _|_, but h (g _|_) == True, even though a strictness
annotation was the only difference between them. (For those not
already familiar with it, _|_ ("bottom") is a semantically undefined
value, equivalent to an infinite loop.)

I agree that it's good to reduce semantically-irrelevant code, but I'm
not sure how feasible the proposal would be. I think GHC's rewrite
rules may be of interest in this problem (the debate between elegant
and fast code). Don Stewart has also written about this issue and has
a couple of really good posts on his blog
(http://cgi.cse.unsw.edu.au/~dons/blog/2008/05/16#fast).

Warm regards,
Alex