World: r3wp

What if you want to generate code, eg:
>> reduce ['return: [integer!]]

== [return [integer!]]   ;  I wanted this --->  [return: [integer!]]

I know you can use to-set-word, but that's not very concise is it 
?

But as Ladislav pointed out above, it seems that  'return:  would 
have to be interpreted as a new datatype ( a lit-set-word! )

Maybe it's not worth it now to add such a new datatype.. or maybe 
it does not hurt at all, speedwise, because if the parser is going 
to catch the syntax error above then it may as well just assign the 
new datatype. The cost, I think, will be in maintaining a new datatype; 
bug-testing, docs etc.  So ok, I agree with Gabriele.

Ok, I posted RAMBO Ticket #-131   (waiting to be reviewed)

Anton: What if you want to generate code, eg:
>> reduce ['return: [integer!]]

try reduce [first [return:] [integer!]]

Yes, that's better, but still a workaround.. thanks :)

If you want something to be reduced and not others, you can use compose.

or use my http://www.fm.vslib.cz/~ladislav/rebol/build.r

Ok thanks, but it's no longer a problem.. :) I just noticed that 
it's a strange inconsistency, that's all.

just found a nasty bug!    if you call browse 5 times in any of the 
rebol view beta versions, rebol crashes.

on window 2k... not sure about other platforms

I posted it on RAMBO...

Same Happens here on rvdraw57e.exe & WinXp proffessional

I get the same as PhilB: "**Crash : expand series overflow"  rvdraw57e, 
 WinXP

Alan just found this issue:

>> site: http://www.dom.com/
== http://www.dom.com/
>> site/1.r
** Syntax Error: Invalid decimal -- 1.r
** Near: (line 1) site/1.r
>> site/1-r
** Syntax Error: Invalid date -- 1-r
** Near: (line 1) site/1-r
>> site/1-1-1r
== http://www.dom.com/1-Jan-2001
>> site/1_r
** Syntax Error: Invalid integer -- 1_r
** Near: (line 1) site/1_r

Actually, just looks like the rebol lexical parser doesn't like it. 
Something I can live with. Just work around using JOIN.

Anton, what 1.r should be, for you?

standard (= load known) rebol words cannot start with a number ('1a 
is not permitted).

Yes, that is what I realised.

the rebol parser could be more relaxed (what it is not an integer, 
an url, a file and so on) must be a word!, but so you will never 
see an error! for invalid integer! invalid url! and so on, and the 
parser would be also more slow

Ah yes, that makes sense.

RAMBO ticket #3244
http://www.rebol.net/cgi-bin/rambo.r?id=3244&

On WinXP: With Rebol/View 1.2.57.3.1c & e betas I can no longer demonstrate 
the bug. :-)  The older View 1.2.56.3.1agg9  still crashes at about 
9000 for me.

View 1.2.57.3.1c  completed successfully three times in different 
sessions, and three times in the same session.
View 1.2.58.3.1e  completed successfully twice.

The test code just scans HKLM, I wonder if I dare scan the entire 
registry ?

But I would like to stress-test by starting several rebol sessions 
to compete for the same keys at (hopefully) the same time.

In Firefox, the Example code section in the link above is shown with 
an extra newline for every line. Looking at the rambo-generated source:

<pre>rebol [
<br>	{
<br>		2.1.24 crashes windows XP at n = 214 in my system
<br>	}
<br>]
<br>n: 0
<br>scan: func [name blk /local y][
<br>	probe name
<br>	probe n: n + 1

<br>;	write clipboard:// name ;if this line is not commented, crash 
appears after more iterations
<br>	foreach x blk [
<br>		y: list-reg/HKLM join name [x "\"]
<br>		if y [scan join name [x "\"] y]
<br>	]
<br>]
<br>
<br>scan "" list-reg/HKLM {}
<br>halt
<br></pre>

You can see all the <br> tags, which are not needed when using the 
<pre> tag, as far as I know. (I thought html tags in the scope of 
a <pre> tag were supposed to be ignored).

If they are, then only some are ignored - I know <a> tags still work 
insire <pre> tags.

inside*

#3626 - path-thru - I'm using "_query_" instead of #"?" char in the 
path and "_atport_" to recognize different port of the url

so the file may look like %localhost_atport_88/test.php_query_id=100

anton: yes, seems that the crash is gone

#3623 the proposed split-path fails in some situations

for example here: %file or here: %e./

this is a different version which use parse and is > 4X  on my system

rebol[
	Author: "Romano Paolo Tenca"
	Date: 10/02/2005
]
split-path-3: func [

 "Splits a file or URL pos. Returns a block containing path and target."
    target [file! url!]
	/local dir pos
][
	parse/all target [
		[#"/" | 1 2 #"." opt #"/"] end (dir: dirize target) |
		pos: any [thru #"/" [end | pos:]] (

   all [empty? dir: copy/part target at target index? pos dir: %./]
			all [find [%. %..] pos: to file! pos insert tail pos #"/"]
		)
	]
	reduce [dir pos]
]

it also fix this split-path bug:
>> split-path %""
** Script Error: Out of range or past end
** Where: split-path
** Near: last target
>> split-path-3 %""
== [%./ %""]

Oldes, good ideas. Do you think the "?" -> "%3F" is a replacement 
which is more prone to collision (than a longer string such as "_query_" 
) ?

Or perhaps you think it could cause problems of automatic decoding 
when ( if it ever happens ), the public cache path is converted back 
to a URL! ?

Anyway, I think the "?" replacement is the one that fixes the most 
common problem. I think ports are not such a common problem. Obviously 
you needed it to work for something..., but I wonder what your opinion 
on it was.  We shall need to make a total solution at some stage...

Gosh! now I think it might be better to replace the "?" with a slash 
"_query/"  (<-- note the slash), so it becomes a new directory.

eg: path-thru http://www.dom.com/cgi.r?hello
;== view-root/public/www.dom.com/cgi.r/hello

Anton, if you do that, how do you convert it back?

I guess you can't, unless you know more. I just leapt at that idea 
because I thought it would reduce the amount of repetition in path 
and filenames.

(so all the cgi.r queries would be grouped as files in the cgi.r/ 
directory)

Mmm.. similar to Oldes' method, you could replace "?" -> "_query/", 
  then directories that end with "_query/" are converted back to 
"?"

Ahha! I just re-read my previous post... :) Duh!