[plt-scheme] scribblings on PLT Scheme 4.0

On Sun, 2007-05-27 at 22:27 -0400, Shriram Krishnamurthi wrote:
> On 5/27/07, Bill Wood <william.wood3 at comcast.net> wrote:
> >
> > But there's space efficiency to be considered as well as time, isn't
> > there?  There's a lot of consing going on with the pure variants.
> 
> If you have no other references to the same cons cells, then the GC
> will (eventually) make sure you use no additional space.  If you do
> have other references to the same cons cells, then you've perhaps
> inadvertently destroyed invariants that those other references depend
> upon; as Matthew said,
> 
>   I used `append!' and `reverse!' a lot in my early Scheming, too,
>   because they were faster. It bit me often enough (no surprise) that
>   I decided to stay away
> 
> So you should really only use the ! versions if you can either be
> *sure* there are no other references (and, more importantly, that
> there will not be other references in future versions!), or you *want*
> those other references to see the changes.  If you cannot be certain
> that you firmly land, now and for the foreseeable future, in one of
> these two categories, why take the risk?

I plead noli contendere as far as the general safety consideration goes,
and I've written a lot of CL with that in mind.  The fact is, I rarely
do list surgery.  However, it is a good thing that the capability is
there when/if careful, knowledgeable consideration deems it's use
appropriate.  It seems to me that space efficiency can lead in that
direction, although somewhat less frequently now than in the days of
kiloword memory :-).  Speaking of which, in 1985 I found a really nice
lisp implementation, Franz-like, including reader macros with splicing,
for an 8080/Z80-based computer (64K bytes!).  Well-documented too.  I
was impressed.

 -- Bill Wood