World: r3wp

Jaime;  I get the impression you may really really like the challenge. 
 It goes way deep.

And check out the books.  The Implementation of Icon book goes nitty 
gritty into PLD.

All, I suggest you read this article from DrDobbs. It does a good 
job explaning syntax-case macros of Scheme.

Building Little Languages with Macros

 or "Picking up where language designers leave off" http://www.ddj.com/dept/architect/184405618

Excerpts: A syntax-case macro can perform arbitrary computation (using 
Scheme) on the results of a pattern match. A syntax-case macro does 
not manipulate plain text, or even plain symbols and lists (as does 
a Lisp macro). Instead, the macro manipulates syntax objects, which 
encode the lexical context and source locations of program fragments. 
Consequently, just like a syntax-rules macro, a syntax-case macro 
respects the lexical structure of the source program (by default) 
and it plays well with source-correlating tools.

Syntax objects enable the implementation of most any little language 
or language extension. Using syntax-case, we have implemented a Java-like 
class system for Scheme, lex- and yacc-like forms for building parsers, 
and constructs for defining and linking program components. Programmers 
using these constructs do not reason about them in terms of their 
expansion. Instead, syntax objects allow the expansion to be hidden 
behind abstract definitions of the constructs, just as the inner 
workings of any compiler are hidden behind a language definition.

A language's concrete syntax need not be parenthesized to make use 
of syntax objects. In particular, we are currently developing implementations 
of Java and ML for our programming environment, DrScheme (see "Fostering 
Little Languages," DDJ, March 2004). Since our parsing tools produce 
syntax objects, we can treat Java and ML like macro extensions of 
Scheme. These macro implementations resemble typical Java-to-Scheme 
and ML-to-Scheme compilers, but little additional work is needed 
to adapt our entire programming environment to new languages.

Jaime; I'm a little confused by your use of fringe.  I was not trying 
to say Icon was

fringe.  Far from it.  It is the most complete, complex programming 
language on the

planet IMHO.  I only mentioned that it was not popular, because it 
includes features

that are over the head of most programmers.  Unless you also include 
Scheme as

fringe, then I take back what I just said, and yes by number of adopters, 
Icon is
fringe as well.  :)

Icon is nice. I was using Icon before coming to REBOL. What it lacked 
for me was communication abilities.

Yeah, Dr. Ralph was writing the language before "icons"  :)  The 
internet was still in
the gopher phase.

For a POSIX layer of Icon with easier TCP, check Unicon
http://www.drones.com/unicon/

Brian, Do not worry. I put Scheme, ML, Rebol, SmallTalk and Icon 
in the fringe category. Java, Python, Perl, C, etc are not fringe 
for me.

Thought so.  :)  Just had to be sure.  For me fringe always means 
'requires an
uncommon form of thinking' leading to low adoption rates.

I used to use Icon as well, mostly for the things I use the REBOL 
parse dialect for now.

Where I see REBOL, scripting will always be the domain of the computer 
programmer,

but the data model is so close to being the domain of everybody, 
I think we are one
small (yet huge) step away.

Icon is an awesome learning step for anyone thinking of languare 
design.  But it does
require being 'smarter than the average bear'  :)

Brian;  Cool.  I'm still in BooBoo land when it comes to parse, but 
I can read and mod
dialects now.  Just not ready to write one.  :)

It was harder to debug Icon than any other language I've used. Not 
because the language required you to be smarter, but because error 
handling was badly designed. It always seemed ironic to me: In a 
language where "failure" was the name of a common control flow, real 
failure wasn't handled very well.

True...it could get pretty hairy, tracking the backtracking .

I never had any problem with the backtracking - that's where much 
of my understanding of PARSE control flow came from. The problem 
came when there was a real error, not an expected "failure", and 
it just backtracked like normal or worse, ignored it.

Brian;  That's because you seem to be accepting of 'uncommon forms 
of thinking'.  :)

Ain't that the truth :)

Like most rebols...

I'm still going to push for LOAD/RELAX.  It'll help make UIs that 
construction workers

can feel empowered working with.  I've seen it.  A boss typing his 
own data in 

notepad, they get a real sense of being in charge, and not lead by 
the nose...Boss

type personalities like to feel in charge of their surroundings. 
 No other language I've
seen even comes close.

I still prefer the opposite. Did you see the discussions we had on 
the blog? If you modify my proposal for the additional parse operation 
in the way Volker suggested, by getting rid of the LOAD keyword, 
you can unify the block and string parse dialects. That way you can 
leverage the REBOL loader right in your parse rules but still fall 
back on string parsing if that doesn't work.

And by fall back, I mean backtrack in the Icon sense.

:)  It's why I mentioned it...  I like that approach...but I'd still 
like to be able to explain

the scripts I write (in front of bosses) without going too techie. 
 Maybe we'll get both :)

Do you have to do a line-by-line explanation, or can you give an 
overview?

Both.  Line by line for the simpler, then I'll code a few lines and 
wait to see if they ask
what the sequence is doing.

It puts them in a very comfortable zone and in a head space where 
they can think
about and request very specific options and outcomes.

And I don't have to write UIs that don't attack the problem at hand 
(their problem at
hand, not the one I the coder nerd might be thinking).

Btiffin, from previous discussions, I'm probably with BrianH on this 
one. But I'd like it if you could come up with some scenarios from 
real life. Then each of us could try our approach to solving the 
problems and compare.

load/relax can be written based on this: http://www.rebol.net/cookbook/recipes/0042.html
;)

Volker; Not sure.  This type of thing happens when bosses type their 
own data...They

don't really really need to type their own data, but it empowers 
them.

10-Mar-2007 $12,002.34 "Home Hardware" "Tile Saw"
1--Mar-2007 $12002.34 "Home Hardware" "Tile Saw"


It's the fact that REBOL "knows" it's a syntax error, that got me 
to thinking about 

gibberish! or the invalid? test in the first place.  I'd like to 
be able to show the user

where the data failed, call notepad and let them try again.  Now 
I just say
try again
 and call notepad,  can't help them much.

text: {
10-Mar-2007 $12,002.34 "Home Hardware" "Tile Saw"
1--Mar-2007 $12002.34 "Home Hardware" "Tile Saw"
}
parse text blk-rule: [
    some [
        str:
        newline |

        #";" [thru newline | to end] new: (probe copy/part str new) |
        [#"[" | #"("] blk-rule |
        [#"]" | #")"] break |
        skip (
            either attempt [
                set [value new] load/next str
            ] [
                probe :value
            ] [
                new: find str " "
                print ["GIBBERISH" copy/part str new]
            ]
        ) :new
    ]
]

based on that you can build your own load (add code to collect stuff 
in blocks) and add "spellchecking" in the gibberish-part.

ml-contest, smartest loader? :)

Cool...

Jaime, (f a b) *must* be quoted or be in a quoted list (maybe using 
the "funny" way of quoting available for macros ;) for it not being 
a function call.

Jaime: you can write a REBOL interpreter in C but you can't compile 
REBOL to C. You can compile Scheme to REBOL but you can't compile 
REBOL to Scheme. The conclusion is that there is something in REBOL 
that Scheme lacks. If you make a language X that has the same features 
as REBOL, then it becomes easy to compile REBOL to X (X will need 
to be an interpreter - well, always assuming you don't have a special 
CPU that can make REBOL compilable).

Gabriele - isn't it in fact big defficiency of Rebol, that it can't 
be compiled? That way we are one level slower than others, forever, 
no? :-) well, I probably oversimplified it few bits :-) I suspect 
that is the price for the ability to have self modifying code, but 
if I understand it correctly from what Jaime says, Scheme or Lisp 
are self-modifying too, yet those can be compiled .... Where is the 
difference? Any simple clarification for non low-level language person 
like me?

rebol is not compilable because it can express things that a compilable 
language cannot express.

if you are willing to give up the advantage in expressivity, you 
can just use a compilable dialect and compile to rebcode or even 
C etc.

also keep in mind that 90% (or maybe even 99%) or rebol scripts that 
people write could be compiled, as they don't take advantage of 100% 
of rebol

(but I guess that 80% of my scripts are not compilable ;)

hotspot can compile code which changes later, so i am not conviced 
:)

i think self and its "childs" could do it.

Err... most of my code surely cannot compile either :)

Remember, too, that REBOL is not first and foremost a programming 
language. If that had been the main goal for REBOL, I have no doubt 
that Carl would have designed it to be compilable, and probably provided 
a compiler from the beginning.

I still think that most of my code is compilable - you just need 
to think about the semantics in the right way.

Its compilable if some of the info is gathered on runtime IMHO. Maybe 
with the option to switch back to the intepreted code, to solve gabireles 
changed 'if.