World: r3wp

For Post 1.3 discussion (moved here from the Debug 1.3 group) ...


What would your "perfect" FOR interface look like (anyone and everyone)? 
For me, I want it to hide the mechanical details more than CFOR does, 
and is nicer to read than the current FOR (for i 1 9 1 [...]). It 
might look like this:

for [i: from 1 to 9]	[...]
for [i: 1 .. 9 step 2]	[...]
for [i from 1..9 step 3]	[...]
for [i: 1 — 9 step 4]	[...]
for [i 0 to 1 step .1]	[...]

or maybe move the word outside the block:

for i [1 to 9]		[...]
for i [1 to 9 step 2]	[...]


Python has an Else clause, though it works backwards from what I 
expect it to; the idea is to have a clause that executes if 'break 
is used. It also has a Continue op that jumps to the head of the 
body for the next iteration.

I agree with Volker and Ladislav about the value of getting the stepping/increment 
code out of the body. I agree so much, I think it should be hidden 
entirely. :-)

The most common case, by far, is to step by one, so you should be 
able to omit that IMO.

cfor is not only about counting, but about anything like a "step". 
it may be counting, or stepping through a list by pos: next pos, 
or whatever. that can't be captured by better hiding.

Gregg: said "The most common case, by far, is to step by one, so 
you should be able to omit that IMO. " , for the most common situations, 
you would use REPEAT or LOOP. I virtually never use FOR, except if 
I'm thinking in some other language

for me, FOR is only there when one of the native looping structures 
doesn't suffice. 

But I do like then dialect options you present. My pref would be 
to keep the value outside (to keep closer to foreach syntax)., 
ie  for i [1 to 9 step 2][..]

I rarely use it myself Allen. The only reason I really think it's 
important at all is that new people coming to REBOL will look for 
FOR. Need to add a doc section on native verus mezz control functions.

Volker, it should operate on series values as well, like FOR does 
today. My examples are all numbers, because that's easier to do concisely. 
:-)

The only time I use FOR today is when I need to:
	a) start at a number other than 1
	b) step by a increment other than one.
	c) brevity and clarity is more important than performance.

If REPEAT had /start and /skip refinements...

assume: func [
    {If a value is not in a series, append it.}
    series [series! port!]
    value
][
    any [find series value insert tail series :value]
]

I propose this new function for Core. check the return values in 
both cases.

As imho is the case with ALTER, ASSUME in my ears sounds too general 
to give a hint to it's functionality just by it's name. 

A function like the one you suggest seems very useful to me, though. 
How's about APPEND-ONCE (could even be APPEND/ONCE on mezzanine level).

I'm losing my mind here ... my Linux rebol scripts just aren't working 
unless I invoke rebol directly myself.

debugging: IIRC I once asked this question already but can't remember 
the answer. Is it possible from inside a function to get the set-word 
this code is bound to? I would like to be able to print the set-word 
for debugging call-traces.

See the following mailing thread:

http://www.rebol.org/cgi-bin/cgiwrap/rebol/ml-display-thread.r?m=rmlGCNJ

You may use a redefined FUNC like

func: func [
    "Defines a user function with given spec and body."
    [catch]

    spec [block!] {Help string (opt) followed by arg words (and opt type 
    and string)}
    body [block!] "The body block of the function"
/name
    word [word!]
][
    either name [
        use [self] [
            self: word 
            throw-on-error [make function! spec bind body 'self]
        ]
    ][
        throw-on-error [make function! spec body]
    ]
]

>> a: b: c: func/name [x] [either none? x [print [self "called with 
X arg of value NONE"]][probe x]] 'my-func

>> a 0
0
== 0

>> a none
my-func called with X arg of value NONE

Might cause serious binding problems, though (I haven't tested it 
in any serious means). And it may probably not suit your needs exactly.

Thought about it twice. Doesn't make any sense, anyway: it has no 
benefits over printing "MYFUNC called with ..." directly.

Yeah, that's going to have some binding issues.

But you have the right idea

Ok, than we should add better trace functionallity to Rebol :-))

I agree with Allen, but I would expect FOR to be able to handle variables, 
eg:
	for i [1 to 5][
		for j [i * 2 to i * 3 step 2][...]
	]

What's the trick to getting the @ symbol into your username when 
checking POP mail in REBOL?

If you check the ML topic index  for pop, you'll find 


http://www.rebol.org/cgi-bin/cgiwrap/rebol/ml-display-thread.r?m=rmlFSJQ

http://www.rebol.org/cgi-bin/cgiwrap/rebol/ml-display-thread.r?m=rmlQGBQ

http://www.rebol.org/cgi-bin/cgiwrap/rebol/ml-display-thread.r?m=rmlQGBQ

Each of which contains the answer.

Which is :


 open [ scheme: 'pop user: "[[username-:-coxinet]]" pass: "password" host: 
 "pop.coxmail.com" ]

Yup, found an email you sent to the mailing list  referencing this 
just before you posted here... ;~>

So you answered my question before I asked it.  That's pretty good. 
;~>

I'm curios as to why do people use @ in a username.? I thought the 
rfc said not to.

In this case, ask Google why...

It's often because the ISP passes the user onto another pop server 
ie. the first proxies for the final pop server

Tiscali has bought many Provider in France. They handle email in 
@tiscali.fr, @freesbee.fr, infonie.fr ... with the same POP server 
(proxy?). So they need the full email as login.

Does anyone know if REBOL runs on Windows CE.NET 4.2 ? I'd need to 
use REBOL on that platform but I don't have access to a CE.NET terminal 
and can't use an emulator because there's no REBOL WinCE x86 binary.

This probably is a silly question, but ... What is the benefit of 
DO GET IN OBJECT WORD over just writing OBJECT/WORD ? 

I see this a lot in VIEW source (e.g. in DO-FACE) and related guru 
sources. I really don't see the point of doing so, be there *must* 
be a reason. Anyone willing to educate me? Is this a path evaluation 
issue?

Maybe the word is dynamic?

Oh, sorry, typo here: should have read DO GET IN OBJECT 'WORD - it's 
LIT-WORD!, actually.

Of course, there *are* differences, e.g. OBJECT/WORD doesn't throw 
an error if WORD is unset. But I'm uncertain on the exact reason 
why that idiom is used.

Dynamic dispatch of code, if the slot exist.

Still don't get it. Both 


 do-face: func [face value][do get in face 'action face either value 
 [value] [face/data]]

 do-face: func [face value][face/action face either value [value] 
 [face/data]]

seem to work.

o1: context [
    init: [print "Init code for o1"]
    eval: [print "Eval code for o1"]
    end: [print "End code for o1"]
]

o2: context [
    init: [print "Init code for o2"]
    eval: [print "Eval code for o2"]
]

eval-obj: func [obj][
    do get in obj 'init
    do get in obj 'eval
    do get in obj 'end
]

eval-obj o1
eval-obj o2

In your example the face may not have an action slot.

Therefor throwing an error instead of graceful fail.

Oh, thanks, now it's all clear. I tried with setting WORD to any 
value I could image, but not with object without WORD. As always, 
so simple, so obvious! Thanks again.

Kind of obvious after someone makes it obvious to me, though ;-)

That seems to be the norm with Rebol. But we are so trained to look 
for complex explanations that we skip the path of less surprise.

Plus, the behavior of object/word depends upon what datatype the 
word is set to and (for some datatypes) which version of REBOL you 
are running. If WORD is set to a block, OBJECT/WORD is the same as 
GET IN OBJECT 'WORD, while if WORD is set to a function, or sometimes 
a path or paren, it means DO GET IN OBJECT 'WORD. By spelling out 
what you want to do, you won't be as tied to variances in REBOL path 
evaluation and your code can be more general.