Re: [Kawa-commonlisp-dev] [PATCH] DECLARE processing.

[Jamison Hope]

On Aug 14, 2012, at 3:49 PM, Charles Turner wrote:

> On 14 August 2012 19:30, Per Bothner <address@hidden> wrote:
> > A minor style issue:  If local variable has a single assignment,
> > you should normally combine the declaration and initialization.
> > For example aliasedDecl - and others.
>
> OK. I always thought it was more efficient to instead of
>
> while (,,)
>  Object o = ...
>
> Have:
>
> Object o;
> while (,,,)
>  o = ...
>
> From an experiment, I now think it makes no difference, apart from
> being slightly harder to read!

I think that used to be true, or maybe it's true (or was long ago) only
for certain languages/compilers.  I have a feeling a lot of those rules
of thumb are just hand-me-downs from the C compilers of 30 years ago.

> > I think this could be avoided by making rewriteBody protected, and
> > changing process processDeclare to something like:
> >
> >  @Override
> >  protected void makeBody (LList forms)
> >  {
> >     ...
> >     letexp = letDone(super.rewriteBbody(body));
> >     formStack.add(letExp);
> >  }
> >
> > Not sure if some variant will work, but it seems it might be
> > worth a try.
>
> I'm not sure if you made a typo above, I found it difficult to
> understand what you meant. I've updated the patch to do something
> similar, apologies in advance if I've botched it.
>
> In a void-returning method like rewriteBody, what's the best way to
> signal a syntax error? The methods I know about from Compilation push
> the handling of error messages for later, are there any that just bail
> out immediately? There's a FIXME: in the attached patch signaling an
> area where I'd like this behavior.

I wasn't sure what he meant, either, but I think you got it right.
I don't see the FIXME you mentioned, though (at least not in so many
words).  Is it this comment?

+        // if declIterator.next() is not an instance of Pair, fail.



--
Jamison Hope
The PTR Group
www.theptrgroup.com