World: r3wp

Ah, cool. Didn't know this. Always something new after so many years...

Just tried it with R3-alpha ..... It does  not work.
I'll report it in Curecode.

Good, at least it helped to catch a bug.

:-)
  to-decimal $3.4

works in R3 but not R2.....So it may be a change of behavior, not 
a bug. Look out for reponse to curecode #807

But does it give the trailing 0 in R3?

Does not look like it:
    >> to-decimal $3.4
    == 3.4

Looks like to-decimal will be the R3 way:
http://curecode.org/rebol3/ticket.rsp?id=807

How can I track the number of times a word/phrase occurs in a string!
I tried this: 
	if find str "a phrase" [print "found phrase"]
but it seems to only find the first occurance.

is that a joke ?

use parse

There are examples on Rebol.com on how to use parse to count occurences 
ot tags etc.

of course it would be nice to have something like find/all which 
returns a block of all the occurences

Steeve... this is not the humour group!

hmm...
 i: 0 while [str: find/tail str "a phrase"][i: i + 1]

parse str [any [thru "a phrase" (print "found phrase")]]

but it would be more relevant to post such request in "i'm new"

I'm using to highlight parts of strings in a face during a text search. 
so I need to find caret offsets for each and append to the face/effect/draw 
block

findall: func [ st occurrence /local result ][
	result: copy []
	while [ st: find st occurrence ][ append result st st: next st]
	result	
]

>> findall "testing" "t"
== ["testing" "ting"]

in that case, would you want the indexes?

graham's solution does exactly that

something like that has also been proposed for R3.

findall: func [ st occurrence /local result ][
	result: copy []

 while [ st: find st occurrence ][ append result index? st st: next 
 st]
	result	
]

>> findall "testing" "t"
== [1 4]

would it be too clever to return the block at the positions instead 
of the index?

he just needs the offsets

a reference or an offset is the same thing to my mind

whereas a reference fullfills more use cases

but can't be molded

and loaded back

Thanks Graham, I had something almost like that func but not quite...

This is an example of the code I used but it only hi-lites first 
occurance: 

if find face/text butt [
						hs: caret-to-offset face find face/text butt
						he: caret-to-offset face find/tail face/text butt
						fsize: face/size/x
						draw_contents: copy face/effect/draw 
						get_xys ;(Find the shape of the drawn hi-lite)
						append face/effect/draw compose/deep [(shape)]
					]

If you know the length of text "butt" you don't need to find the 
end .. just calculate it, as it should be faster.

amacleod, I wrote some efficient string search functions. The file 
has simple, first implementations alongside some efficient implementations 
which use PARSE.

http://anton.wildit.net.au/rebol/library/demo-string-search-functions.r

I'll check it out.  thanks, Anton.

I got it working...

Sort of inserted my hi-lite mapping code into Graham's function...
Thanks again, Graham!

It allowed me to step through the face one find after another. Exactly 
what I wanted to do.

I'm having a problem with call.


when I use the /output/error and supply  strings , I get a strange 
error in the stdout holding string:

Unable to read from standard input: The handle is invalid.


the really strange one is that if I use /console, the expeted out 
is effectively printed out to the rebol console.... so the cmd does 
return data and rebol is able to handle it, but not within the /output/ 
refinement!


also, another call, using another command, works as expected... so 
I don't think I have a error in my code.

not that I have tried using /shell and it didn't help

has anyone seen /output and /console react this way?

not=note

maybe you also need to provide /input ? (it's hard to guess what 
the problem is)

oddly this group was private but everyone was in it .. made it public.

I will try with/input.  providing a few "^/"  but It definitely doesn't 
need input on the command-line.

but sometimes with progamming... logic doesn't prevail, hehehe

gabriele, the /input  works !!!  damn why didn't I think of that.. 
I lost several hours trying to find other command-line ssh tools 
which would work... none really do... the putty tools really are 
the best ones out there.

what is the fastest way to count the occurence of each numbers
within a block containing 5000+ numbers?  

e.g [1 3 6 2 4 9 33 6 67 2 12 23 34 12 2 4 56 2 ...]

calculating the max and min with a simple loop and counter is 
very time consuming for thousends of these blocks. 
inserting into mysql and doing a 'group by' also. 
any ideas for a faster solution?

what if you sort the block first?

then FIND on the first occurrence of 1, of 2, of 3, of 4, etc. and 
store the indexes.

yes henrik, nice idea in this moment I was thinking similar but with 
copy/part but storing the index is much faster.

the numbers for the  FIND can be supplied from a copy of the block 
with UNIQE

yep, will give it a try, thx henrik...

I usually use Joel's tally function -- it saves me the time needed 
to write a faster function:
http://www.rebol.org/ml-display-message.r?m=rmlKDDS

I think the fastest would be a variation of radix sort, if you know 
the max/min..

Setup for example:
>> Data: array/initial 5000 does [random 200]

== [36 119 148 112 87 59 176 21 95 138 183 28 1 119 14 74 39 78 141 
99 56 71 47 174 92 81 157 182 122 123 98 73 35 170 150 11 183 8...

>> MaxVal: fold Data 0 :max ;Just use a simple loop, I defined fold 
long ago for simplicity
== 200

What I mean:
>> Vals: array/initial MaxVal 0

== [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0...
>> foreach Val Data [Vals/:Val: Vals/:Val + 1]
== 27

>> repeat I MaxVal [if 0 <> Vals/:I [print rejoin [I {, } Vals/:I 
{ times}]]]
1, 24 times
2, 20 times
..........
199, 20 times
200, 16 times