[REBOL] Re: object funnies
From: joel:neely:fedex at: 19-Oct-2001 7:36
Hi, Andrew,
Andrew Martin wrote:
> I wonder if it would be a smart idea to make a scanner function
> that looks inside nested blocks for a object! spec and adds any
> set-word! to the start of the object's block...
>
It sounds like an interesting group experiment, but see below...
I'd personally prefer an object checker that simply takes an
object and identifies those cases when out-of-context words
are set.
The proliferation of cases where we have multiple ways to do the
same (or worse, slightly different) things makes me just a bit
paranoid for a couple of reasons:
1) The more we hide (from newbies or ourselves) the areas in
which RDIOT (REBOL Does Its Own Thing), the more we increase
the likelihood that its unique behaviors will take even more time
to figure out when they do emerge.
2) The more we have proliferations of slight variations (if you'll
pardon the alliteration ;-) the more details and exceptions we
have to remember to get our code right.
I'm not sure that the marginal value added by
make function! [][][] vs.
function [][][] vs.
func [ ][] vs.
does []
(just to give one example, I'm sure everyone can immediately think
of other cases) is really all that much compared to the value of
tidying up some of the needless inconsistencies.
Frankly, this all reminds me of a statement I heard years ago:
"FORTH is not a programming language;
FORTH is a programming language construction kit!"
Perhaps there's a reason why radically extensible programming
languages have never made it to the mainstream (especially when
used in the most radically-extensible style) -- it becomes all too
easy for the code for each application to become in effect its own
private language. In the long term this poses risks to the safety
and stability of maintenance efforts.
To be fair, I recall the "shoebox" quotation in which Carl described
REBOL as a language for "programming in the small". If we expect to
use REBOL only for relatively short, throw-away scripts, then this
isn't an issue.
If that's NOT how someone wants to use REBOL, then there seem to be
a few choices that have to be made:
1) Excercise self-discipline and avoid radical use of extensibility,
so that any REBOL programmer (including the original author after
at least one nap or long coffee break ;-) can pick up a piece of code
and understand it quickly and maintain it safely.
2) Have some widely-used means of creating re-usable units of code
(components/modules/packages/...), whether delivered by RT or
created by the community, so that semi-standard extensions can be
widely (and consistently and conveniently) deployed and used.
3) Have more docs and examples of the cases where RDIOT and how the
conceptual model of REBOL differs from more "familiar" languages.
(I'm working on some ideas here and hope to provide some results in
the near future.)
I'm not suggesting that REBOL "should" look more like any particular
other language, but simply making a practical observation -- most of
the high-skill/high-performance programmers of my acquaintance are
quite comfortable with multiple languages and the use of multiple
languages as specialized tools within a single task. They tend not
to be willing to:
1) devote themselves to a single language for everything, nor to
2) spend lots of time experimenting and trying things out and running
into dead ends and solving syntactic or semantic puzzles just to
develop proficiency in a new language (at least not unless there's
some REALLY big payoff).
This brings me full-circle to your suggestion above and my first
response. The open-source community thrives on competitive evolution
of ideas. People try things out and, over time, the more useful,
reusable, learnable, (etc.) ideas tend to become the norm within that
subculture. In such a context, your suggestion regarding a variation
on the process of object-making would get debated, tried out, praised,
criticized, etc... until the community converged on a way to address
the issue in a consistent way.
We don't seem to have such a mechanism here.
For example, we've been intermittently talking about MAP for almost a
year and a half on this list (at least, that's how far back escribe
can find references). Several of us have (re)invented variations on
that theme, yet once the discussion dies down, it's forgotten. We
have no way to get this into any forum where we understand its progress
(or lack thereof) into the language or a "standard library". The same
could be said of many other specific proposals/discussions. Soooo...
In the absence of such mechanisms, I tend to be fairly conservative and
minimalist in terms of what I have time and attention span to think
about or experiment with. PLEASE let me emphasize that I'm not being
negative or critical of your suggestion, but just trying to raise what
I perceive as some practical limitations to our ability to experiment
with REBOL...
Thanks for listening to the rant! ;-)
-jn-
--
; sub REBOL {}; sub head ($) {@_[0]}
REBOL []
# despam: func [e] [replace replace/all e ":" "." "#" "@"]
; sub despam {my ($e) = @_; $e =~ tr/:#/.@/; return "\n$e"}
print head reverse despam "moc:xedef#yleen:leoj" ;
