Re: [Chicken-users] handling the undefined value

[Felix]

From: Jörg "F. Wittenberger" <address@hidden>
Subject: Re: [Chicken-users] handling the undefined value
Date: Mon, 29 Nov 2010 14:21:13 +0100

> Am Samstag, den 27.11.2010, 16:06 +0100 schrieb Felix:
>> From: Jörg "F. Wittenberger" <address@hidden>
>> Subject: Re: [Chicken-users] handling the undefined value
>> Date: Fri, 26 Nov 2010 22:10:46 +0100
>> 
>> > 
>> > Well, if it was a compiler switch, off by default, it should not do
>> > harm.  Would it?
>> > 
>> 
>> Yes, I'll do that. I could also least remove the warning for the
>> (common) case of a conditional performing a self-tailcall.
> 
> Do you mean that this pattern
> 
> (let ((x ...))
>   (let loop (...)
>     (if ...
>         (loop ...))))
> 
> in tail position of a procedure would still return one "undefined"
> value?

Yes, it would. This particular case is detected now, though.

> In tail position however its not quite be what I need.  -picky should
> force me to write:
> 
> (let ((x ...))
>   (let loop (...)
>     (if ...
>         (loop ...)))
>   x)

You get loops like the above in internal code, like the expansion
of the (internal) compiler macro for `for-each'.

>> How about "-picky" ?
> 
> -picky sounds good.

Ok. This will probably become a "change request".

> 
> Since -picky would enforce particular programming style(s), I'd suggest
> that it would take flags right from the start like -debug does.  So I
> could enforce "never return/reference undefined", later maybe complain
> about letrec binding to something not a procedure (zero reasons to do so
> from standard Scheme point of view, but sometimes likely to fix
> mistakes).  More ideas might come up (which systems do still have a
> command line length limit?).

Control over warnings has been requested already. I have to think more about
this, since it would be nice to keep things simple, yet provide some sort
of control over warnings (both statically and dynamically).


cheers,
felix