World: r3wp

ask Carl why he added remove-each as a native? ;-) you could use 
mezz too, no? ;-)

The FIND/MATCH behaviour seems correct to me, Pekr.

correct yes - usefull - no :-)

In your particular case, perhaps.

we started with remove-each, we could continue with change-each, 
find-each, no?

dunno - I would find them being consistent additions and they would 
be fast ...

do http://hmkdesign.dk/rebol/rch4.r
requires View 1.2

code is messy, many one-letter variables :-)

find-each: func [dataset [series!] value /local result][
    result: copy []
    parse dataset [

        any [set word string! (if find word value [append result word])]
    ]
    result
]

>> find-each ["Jaime" "Carl" "Cyphre"] "a"
== ["Jaime" "Carl"]
>> find-each ["Jaime" "Carl" "Cyphre"] "y"
== ["Cyphre"]

Small modification it takes care of non-string values in the block.

Ha! That is so cool. Never thought function which serves different 
purpose can be used in such case ;-)

find-each: func [dataset [series!] value /local result][
    result: copy [] 
    parse dataset [

        some [set word string! (if find word value [append result word]) 
        | skip]
    ] 
    result
]


>> find-each ["Jaime" 1 "Carl" 2 "Cyphre" 3 http://google.com"Ladislav"] 
"a"   
== ["Jaime" "Carl" "Ladislav"]

I wonder if that one will be faster than loop?

that should be easy to test, will do so tomorrow ...

I remember someone used 'parse in the past as a trick to get pointer 
to binary data in rebol ...

it was somethin with images IIRC ...

Anton, I got access to your include.r now. Interesting option to 
only include certain functions or words from a script!

oh.. ? I think it's essential.

.. to avoid inadvertent pollution of your target context (usually 
the global context). This should avoid many bugs. It always annoyed 
me when C coding that when I included a library for a particular 
function, I had to check that library to see what else I was including. 
Then of course some libraries include things from other libraries...

This reminds me of Tao Elate. In that OS, all library functions are 
small VP asm files on disk. So if you use e.g. printf, only that 
function and not the whole stdlib is loaded in memory. The same function 
is also shared among all running programs minimizing memory overhead. 
Genius, as I see it!


Something like that can be implemented in REBOL with the use of objects 
in objects (that are not multiplied in mem). It's the way, e.g. the 
feel object is implemented in View. To be really efficient, only 
the functions (in the object) needed should be included into mem 
from disk.

note: that feature can be imitated using

    make object! [#include %somedefinition.r]


when using my INCLUDE. The only trouble is, when the author of the 
script does some "nonstandard" things. Then it may not be protective 
enough, which is the case of Anton's include too, where you have 
to rely on the discipline of the original author.

when you need a reduce/deep and there isn't one, what do you use 
instead?

I want to reduce something inside a nested block      reduce [ 'blah 
[  to-word "desired-literal-word" ] ]  ;sorta thing

reduce [ 'blah reduce [  to-word "desired-literal-word" reduce [to-word 
"deep-literal
-word"]] ]   ;ss this works...just talking to myself...nevermind

hmmn, reduce/deep or reduce/nested would be more elegant nonetheless.

what are you doing?

is compose/deep/only an option? Also a reduce/deep would be short, 
if you need it.

trying to reduce a set of nested blocks ... compose/deep/only would 
not reduce the inner blocks, but leave them as they are...

compose/deep [ (a) [ (b) ] ] ; would work

I guess reduce/deep would not be very hard to implement...was just 
surprised there isn't one already...;-)

depends if ou code some themplate, then maybe compose/deep. if its 
data, maybe better reduce/deep.

compose makes strings...I am trying to get it down to literal words.

No, compose makes blocks. If  you mean "flattens blocks", use /only.

going from VID to literal-words...to do comms syncing...so say a: 
"cat" and b: "dog"  I want [cat [dog]] NOT ["cat"["dog"]]

compose [ blah [  (to-word "desired-literal-word") ] ]

[blah [(to-word "desired-literal-word")]]   ;is what returns, instead 
of  [blah [ desired-literal-word]]

sorry,
  compose/deep [ blah [  (to-word "desired-literal-word") ] ]

and to be defensive, compose/deep/only.

A first version:
reduce-deep: func[blk /local][
 blk: reduce blk
 forall blk[
  if block? blk/1[
   blk/1: reduce-deep blk/1
  ]
 ]
 blk
]

probe reduce-deep [ 1 + 2 [ 3 + 4 ] ]
probe reduce-deep [ a: [1 + 2] a/1] ; limitation, does not work

reduce-deep: func[blk /local][
 forall blk[
  if block? blk/1[
   blk/1: reduce-deep blk/1
  ]
 ]
 reduce blk
]
probe reduce-deep [ 1 + 2 [ 3 + 4 ] ]
probe reduce-deep [ a: [1 + 2] a/1] ; limitation, does not work
[3 [7]]
[[3] 3]

better. perfect one is hard.

(drop the "limitation, does not work" in last post)

this breaks then, because inner blocks are evaluated first:
  a: 2 probe reduce-deep [ a: 1 [a]]

that's interesting, I figured the order of evaluation might be why 
there isn't on...why I instinctively thought of /nested as a refinement...thanks 
for the insights.

problem here is, we need do/next to know how long one experession 
is. but before do/next, we can not reduce subblocks. that should 
be done only for one expression. but we do not which blocks before 
do/next.. maybe it should be really inbuild?

but if you don't do such tricky assigining in reduce, this should 
work

What did i write? Trying explanation again.. We need do/next to know 
which blocks are in the next expression. Then expand only these. 
But by using do/next, we have already used the old, unreduced blocks..

So the workarounds is to expand all blocks at once, either before 
reducing that level or after. breaks a bit semantic, means do not 
depend on evaluation-order.

thanks...sorry, got pulled off by my daughter to google when her 
power would be restored...your thoughts here are greatly appreciated, 
as always, Volker.

I went down this road once. Using do/next terribly slows things down. 
I think that you need parse for this.