AltME groups: search

How do I get the window to resize withthe image face?

pic1: load %pic.jpg
pic_size: 200x200
view/new lay: layout [pic: image pic1 pic_size]
wait 2
unview lay 
pic/size: 50x50
view/new lay

you need to get a resize event from the window and apply that to 
the image. you can use INSERT-EVENT-FUNC for that.

img: load %your-image.png

view window: layout [pic: image img button "change size" [pic/size: 
300x200 window/size: pic/offset * 2 + pic/size show window]]

Never mind, I found a link Anton posted for image Magick image saver...

You don't need wrapper, but you cannot do: save-image %test.jpg to-image 
layout [button "hello world"]

Also IM can be used as a command line tool, which I was doing quite 
a long time. But finally I'm working on wrapper as it's faster than 
doing many calls to external app. Of course.. if you want to resize 
100 photos per year, you don't care if it's call or by using wrapper.

you can save temporarily as tmp, do whatever you want to image afterwards. 
I would not estimate overhead of calling an external app. Especially 
on unix systems, many tools are just command line tools anyway,no? 
:-)

For some reason I can't get  %CORE_RL_wand_.dll to load using Oldes' 
save-image script. I have it in the same directory as the script 
and I tried it in the systemn folders...
error:
** Access Error: Cannot open CORE_RL_wand_.dll as library
** Where: context

** Near: lib_ImageMagickWand: load/library either system/version/4 
=

Same if I have an ajax type of search ... I don't show an empty table 
until it fills .. I show a spinning image and then show the table 
when it is filled.

Is 'image-filter implemented in R2 ? Or is there another interpolation 
 means to improve the display of an image when changing it's size? 
Of course, I do not want to use external means like ImageMagick.

I am not looking for thumbnail generation, but for image display 
with rescaling with fractional zoom factor (let's say between 0.5 
and 2.0). When using effect resizing in a layout, the interpolation 
mode is nearest. Is it possible to set it differently (i.e. bilinear) 
?

Jocko, in R2 if you use effect dialect to resize images, yes it uses 
nearest interpolation.
But draw uses blinear by default.
see:

img: logo.gif
siz: img/size * 4
view layout [
	below
	image img
	image img siz effect [aspect]
	box siz effect [draw [image img 0x0 siz]]
]


In R3 it's annouced that  image-filter supports NEAREST,BILINEAR,BICUBIC 
and GAUSSIAN interpolation

I have a strange reaction in some AGG code...  calling show on the 
face is actually appending the new AGG drawing to previous such that 
the image is the result of all prior draws... but the draw stroke 
(verified) isn't growing!


anyone know what is causing this... AFAFIK I am not using the draw 
command, but setting the face/effect and calling to-image on the 
face.

DOH!  never mind... just found it... I was setting the face/image 
to that new image  ...  heheheh  obvious.

Giving 2 points as input , i can write it if you want, but give me 
the paremeters you want and the result expected (an image, a block 
of coordinates ?).
Give the specs

Oldes/graham


here is a complete application which shows you how to build and affect 
a face with a ghost effect.

----------------------------------
rebol []

ghost-blk: [
	ghost-data: none
	
	ghost: func [][
		self/ghost-data: make self []
		self/image: img: to-image self
		effect: [ merge grayscale contrast -50 ]
		;size: img/size
		edge: text: feel: pane: none
		show self
	]
	
	regenerate: func [/local attr][
		foreach attr [image effect feel edge text feel pane][
			set in self attr get in ghost-data attr
		]
		ghost-data: none
		show self
	]
]

stylize/master compose/only [
	field: field with ghost-blk
	button: button with ghost-blk
	btn: btn with ghost-blk
	scroller: scroller with  ghost-blk
]

view layout [
	across

 toggle "btn" [either face/data [my-btn/ghost][my-btn/regenerate]]	

 toggle "button" [either face/data [my-button/ghost][my-button/regenerate]]	

 toggle "field" [either face/data [my-field/ghost][my-field/regenerate]]

 toggle "scroller" [either face/data [my-scroller/ghost][my-scroller/regenerate]]
	return
	my-btn: btn "yippe"
	my-button: button "ka"
	my-field: field "yay"
	my-scroller: scroller 100x20
]

I made something quite similar recently; an expanding/collapsing 
dir-tree viewer, like the dir panel of a file browser. Each directory/file 
has to be an object, because I store some state along with it, like 
collapsed/expanded, and other interesting attributes can be stored 
in there in future, when I get around to collecting the info. The 
purpose of the app is basically to create an image of a directory 
structure, which can be saved to disk, viewed magnified etc. to give 
an overview of directory structure.

Is there an easy way of how to resize an image? I tried:

i1: load %my-img.png
i2: make image! 82x63 i1

but the image is black :-)

Untested, but should work :

i1: load %my-img.png
i2: make image! 82x63
draw i2 [image i1 0x0 82x63]

i2: to-image layout [image i1 82x63 effect [fit]]

What I need is smaller PNG, because SharePoint just sends original 
image to browser, instructing it to resize. IE is pretty much ugly, 
FF is OK. I tried in xnView, my friend in IrfanView, no matter what 
we do, we can't save so that it preserves transparent color ...

I think, you can preseve transparent by making a separate image from 
the alpha channel, resizing rgb and alpha separately and combine 
them in the end to a final image.

compound rasterizer: well, this is another type of rasterizer(which 
has been already implemented into AGG2.4 codebase). I haven't made 
any perofrmance tests comparing the scanline AA rasterizer we are 
using but it could be a bit faster because this method shouldn't 
redraw objects one over another.(ie it renders the image 'flattened') 
I can't tell you how much faster it could be but the more layers 
you are using the bigger difference could be seen.

I'm trying to print a vid face face via conversion to image...
what's the best way to get good text quality...
ie...is the a best font to use...
my out put is a little jaggy

either you use AGG for aliased fonts, or you increase image size 
so its equal to the dpi you want to print.

How far did Henrik get with his vid to pdf dialect?

My layout is all text except for one small b/w image...

Maxim, so to display a proper page I would need to increase the sacel 
of my vid image quite a bit...larger than screen?

when I was doing an art project, I had to create a printout for a 
36 inch wide image... the file was an 508 mb, 300 dpi image.

I was using a print method calling MS image viewer transparently 
to print without opening up but I as adivised to keep my image at 
612x792 to keep it on one page...how do increase 5x  now?

you need to set the dpi of the printout to a higher value.  I don't 
know about that solution, but either it supports the dpi value of 
the input image, when it has one, or you should be able to set it 
directly.

612 / 72 = 8.5  ;-)  so your values are standard video dpis

photoshop allows you to change the meta information about dpi without 
resizing the image... I don't know if there are little tools which 
you can run command-line that will do it too.

doesn't support bitmaps yet (have only used it for printing text), 
but you use it like to-image.

window: layout [i: image rate 5 load %gtfo.jpg key escape [halt]]

zoom-in: does [i/size: i/size * 5 / 4]
zoom-out: does [i/size: i/size * 4 / 5]
down: false
alt-down: false

insert-event-func [
	probe event/type
	switch/all event/type [
		down [down: true]
		up [down: false]
		time [if down [zoom-in]]		
		alt-down [alt-down: true]
		alt-up [alt-down: false]
		time [if down [zoom-in] if alt-down [zoom-out]]
	]
	center-face i
	show window
	event
]

view/options window [resize]

REBOL []


window: layout [i: image rate 5 load http://rebol.com/graphics/kits.jpg
key escape [halt]]

zoom-in: does [i/size: i/size * 5 / 4]
zoom-out: does [i/size: i/size * 4 / 5]
down: false
alt-down: false

insert-event-func [
	probe event/type
	switch/all event/type [
		down [down: true]
		up [down: false]
		time [if down [zoom-in]]		
		alt-down [alt-down: true]
		alt-up [alt-down: false]
		time [if down [zoom-in] if alt-down [zoom-out]]
	]
	center-face i
	show window
	event
]

view/options window [resize]

Does anyone know if IMAGE-FILTER BILINEAR is working in Draw? I see 
no differences with NEAREST mode (which looks horrible when scaling 
down a big image, even with preserved aspect ratio).

there are wrappers to image-magick, no? Or - you could use nconvert 
from xnview package from command line too ...

Just a test:

set 'resize func [img /local sz c1 c3 i2] [
	if file? img [img: load-image img]

	scale: img/size/x / 160
	sz: as-pair 160 round img/size/y / scale
	c1: 1x0 * sz
	c3: 0x1 * sz


 i2: to-image layout [origin 0 box sz effect [draw [image img 0x0 
 c1 sz c3]]]
	save/png %img.png i2
]

Call RESIZE with a loaded image. It'll save the img.png file, that'll 
have x-size of 160. It looks ok here.

you can blur the image before downscaling, but it can smear it more 
and the image isn't as clear.

The original image is the CureCode homepage viewed inside Chrome 
(http://curecode.org)

try image magic.  its fast, easy to use on the command-line, it can 
process in up to 640 bits images and there is limit on capacity as 
it can actually use disk as ram while its processing!  I rendered 
20GB worth of images into one composite ... using 500MB of ram.

and there are no higher quality engines out there... but that all 
depends if it does what you need in terms of image and graphics.

Max, thanks for the additional info. Actually, NConvert is great 
but only as a command-line tool, it's not available in library form, 
so I'll probably end up wrapping Image Magick.

there is already an image magic R2 wrapper on rebol.org you can use 
to start, I've used it and it works... but you need to find the proper 
version of image magick (an older version).

If yes, then you could make a thumbnail algorithm in plain C and 
build your own dll. Calculating each pixel in the thumbnail as a 
mean value of a number of pixels in the original image should give 
a nice result, I think.

I think the correct term is 'filter with resampling'. If you add 
the resampling step then the image looks much beter. Then different 
filter algorithms can even improve the quality (usually the beter 
filter the slower the calcualtion is)

but Image Magick is probably more popular, dunno ..

the issue being that image magic does things incredibly fast.  its 
a library with its origins in very high-end visual effects, and is 
used massively by that industry.

and a few other image formats too.  (image magic loads/saves more 
than 100)

This is not obvious with VID, because you might build such a layout 
in a special setup, but in the VID Extension Kit, you have face constructs. 
A face construct is simply a style that accepts a particular dialect 
as input. The dialect is parsed into VID which then in turn generates 
an internal face tree. This means you can build styles that do something 
like:

view layout [
	paper [
		size a4
		dpi 300
		at 20x20 image bay.jpg
	]
]


The point is that every time you feed the paper new data, using SET-FACE, 
it's re-rendered from scratch using VID.


You can also build very complicated list views that dramatically 
change appearance and number of faces, every time you feed it new 
data.

it's something you wrote that allows you to pan an image

ok, the image-pan style?

pan-image style

but it happens after I move the image around a bit

image processing requires a lot of CPU juice.


 we have to render the AGG, use the bitmap in a face, apply an effect 
 on it, and then re-create a new bitmap out of it.  we aren't just 
 drawing/effecting over and over the same image memory area but creating 
 a new image at every refresh.


it would be nice if there was a complement to the 'draw function 
called 'effect.  maybe its in R3, or maybe it should be.

updated the animated vector projection tests script:

-uses the chrono library for more precise time management.

-now has option to prevent clearing the image at each refresh (feedback), 
press enter

-you can also manually increase and decrease max frame-rate on mouse 
drag using arrows.


http://www.pointillistic.com/open-REBOL/moa/files/ptest-preboled.r

on my system, I can easily go up to a 100fps before feeling lag.

Hi!


Is there a good way to draw an image with an alpha channel? In the 
example below, I would expect a green outlined circle with a transparent 
background so I could composite it over the red box in the layout 
(for instance). However, I get solid (and very un-transparent) black.


    x: draw 512x512 [pen green line-width 3 fill-pen blue circle 256x256 
    200]
    view layout/tight [box 512x512 red  at 0x0 image x]

Any ideas? Thanks very much!

so there are a few options.


I'd build a transparent image to start and graph stuff over it using 
draw.

ex:

x: make image! 512x512
x/alpha: 255
draw x [fill-pen red circle  256x256 200]

If you cut part of an image out, the rest of the image collapses 
in.

R2's chroma key effect  uses a single color as the transparency. 
 so you'd need to build an alpha image using a range based on luma 
or some other more complex algorithm... which you can easily find 
on the net.

game engines use this simple system.


-create very transparent images with a gaussian fall off.  actually 
give the prefered shape to your image... so if you want a triangle-like 
flame, generate a smooth triangle with alpha/color falloff.

-create a block which will store a list of pairs, each one holds 
the position of a single "particle"

if you want smoke, just add the "dying" particles to a second list, 
using the same process, to animate them but with a bigger/softer 
image.  

the dying fire becomes the birth of the smoke.

for the particles, either pregenerate (and simply offset) small faces 
with an transparent image or build an AGG block at each refresh.

the plasma effect you saw used repetitive blur to an image, with 
new particles added and some offset added to the previous image.

the problem with this is that you cannot change the color and the 
fact that the whole image gets faded.

it could work if R2's effect system also managed the alpha channel 
but it doesn't AFAIK.

Very cool, Maxim...thanks. 

is this done as an interactive process?

Not sure what you are asking but stage 1 would be to allow these 
fire "objects" to be placed on top of an image via drag and drop 
and allow for some editing such as sizing. (Later versions would 
allow adjusting those other variables you mentioned above). 


Stage 2 would allow for some transitions from one set of settings 
to another so for example the fire can become more intense over time 
or when clicked on or a button is pressed. 


I just had a week of training on a flash based system...its cute 
and does teh job for what we are going to use these "simulations" 
for but I thought it was a little too complicated for your average 
fireman to use if he wanted to create his own sims for drill purposes.


The flash extensions used all those variables you mentioned above 
to allow for a great degree of controll of the effects.

by interactive I mean will this be rendered or running live?


with a rendered system you can crank up the particle count and just 
pregenerate the whole data set as a series of images.  then when 
you need to see it, just cycle the image of a face.


usually, people use a few particles for preview... then generate 
"flipbooks" which are just rendered images and include thousands 
of particles with their alpha channel density reduced to practically 
nothing (like 2% visibility)  this generates very pretty effects, 
but at a cost of rendering time.

you could render it as a single plane mask and use it only as transparency 
(alpha) for another image.

better off having an image which simulates the cluster to begin with.

the script will end-up being about 800kb with nothing but image data 
and 9k of text at the end...

How to clear the image cache? I always forget that:/

load-image/clear

Yes, but what if you don't want to load image yourself.. for example 
in this simplifed example:

write/binary %tmp.jpg read/binary http://www.rebol.com/graphics/carl1208-120.jpg
view layout [ image %tmp.jpg button "close" [unview/all]]

write/binary %tmp.jpg read/binary http://www.rebol.net/photos/carl3.jpg
;now there should be new image used, but it's not:
view layout [ image %tmp.jpg button "close" [unview/all]]
delete %tmp.jpg


I think that the solution is to clear the image cache on unview/all. 
What do you think?

The only way how to clear the cache without loading new image is 
using:

write/binary %tmp.jpg read/binary http://www.rebol.com/graphics/carl1208-120.jpg
view layout [ image %tmp.jpg button "close" [unview/all]]
clear second second :load-image

write/binary %tmp.jpg read/binary http://www.rebol.net/photos/carl3.jpg
view layout [ image %tmp.jpg button "close" [unview/all]]
delete %tmp.jpg

Oldes shows that you can't just "load-image/clear", you must "load-image/clear 
imgfile", which loads a new image after clearing the cache. It offers 
no way to separate the two functionalities. Annoying, isn't it?


O  Use  clear second second :load-image , as you are, but I suggest 
doing this straight after the loading, eg.

 layout [ image %tmp.jpg  do [clear second second :load-image] ... 
 ]

O  Patch or replace LOAD-IMAGE  eg. so its IMAGE-FILE argument can 
also be NONE!

O  Patch the IMAGE style by changing the FILE function in the MULTI 
object:
	print mold get in svv/vid-styles/image/multi 'file
    so it doesn't use LOAD-IMAGE.
[ ]  Submit this deficiency to curecode.

Or  
>> view layout [image (load %bay.jpg)]
>> mold second second :load-image
== "[]"

am I stating something obvious when I say check the source for load-image?

>> source load-image
load-image: func [
    "Load an image through an in-memory image cache."
    image-file [file! url!] "Local file or remote URL"
    /update "Force update from source site"
    /clear "Purge the entire cache"
    /local image image-cache
][
    image-cache: []
    if clear [system/words/clear image-cache recycle]
    if any [update not image: select image-cache image-file] [

        if all [update image] [remove/part find image-cache image-file 2]
        repend image-cache [
            image-file
            image: either file? image-file [load image-file] [

                either update [load-thru/binary/update image-file] [load-thru/binary 
                image-file]
            ]
        ]
    ]
    image
]

I was checking the source of course, without it I would not come 
with code like:
	clear second second :load-image


But it was not for the first time when I had the situation with the 
unexpected image. Of course, to use :
	layout compose [image (load %image.jpg)]

is a solution as well. I just have to remember it not be confused 
again with code:
	layout [image %image.jpg]


The problem is, that normal user who want's to just display simple 
view layout does not know that there is the function load-image.

another rokaround is:
load-image/clear http://
or load-image/clear to-url none
etc.

>> view layout [
[     box white 327x327 draw [
[          pen logo.gif fill-pen off box 100x100 300x300
[         ]
[    ]
** Script Error: draw expected image argument of type: image pair
** Where: do-facets
** Near: draw [
    pen logo.gif fill-pen off box 100x100 300x300
]

Is there a way to scale an image using effect [draw [image pic]] 
without giving coordinates as in effect [ draw [image pic 0x0 200x200]]


I believe the coordinates are for location on in the frame and I 
just want it rendered inline with the rest of the vid objects

Got it...I had to set the vid image face to teh size I wanted first 
and then match it in the draw coordinates

You can also use scale:

view layout [box effect [draw [scale 2.0 2.0 image logo.gif]]]

I'm trying to launch an "alert" window with an animated image while 
downloading a file. I've tried using "View/new" but the animation 
does not animate. What's a good way to get this effect?

There could be some number of image binaries so it takes as long 
as 10's of seconds to a minute....long enough to make user think 
program as frozen!

I'm using Christopher Ross-Gill's animation style. I wrote a little 
function to convert an animated gif to an image file that the style 
can use....

Kaj, I thought about doing something like that but how do you kill 
the process from rebol?

>> layout [X: button]
>> ? X/doc
X/DOC is an object of value: 
   info            string!   "rectangular, rendered buttons" 
   string          block!    length: 2 
   image           block!    length: 2 
   logic           none!     none 
   integer         string!   "width of button" 
   pair            string!   "width and height of button" 
   tuple           block!    length: 2 
   file            none!     none 
   url             none!     none 
   decimal         none!     none 
   time            none!     none 
   block           block!    length: 2 
   keywords        none!     none 

>> ? X/doc/block

X/DOC/BLOCK is a block of value: ["execute when clicked" "execute 
when alt-clicked"]
>> view layout [button [print "clicked"][print "alt-clicked"]]
clicked
alt-clicked

More poking around:

>> foreach [name style] system/view/VID/vid-styles [? name]
NAME is a word of value: face
NAME is a word of value: blank-face
NAME is a word of value: IMAGE
NAME is a word of value: BACKDROP
NAME is a word of value: BACKTILE
NAME is a word of value: BOX
.....more...

It helps to understand, what LAYOUT is doing: As you may know, VID 
is a dialect, which is a parsed collection of words, numbers, strings, 
etc. in a specific format, essentially a sub-domain language, within 
REBOL. The job of LAYOUT is to parse the dialect block and convert 
that into a face tree (a face is just an object, try typing FACE 
in the console). The face tree can then be fed to the VIEW function, 
so the layout can be displayed, so:

VID -> layout -> view -> window displays with content


You can also create a face tree manually, but that's far more laborious, 
which is why VID is there.


When LAYOUT creates a face, it does so in accordance with a style 
from Izkata's shown style list, such as IMAGE or FIELD. A style is 
essentially a prototype face.

Nice !

p: 0x0 img: make image! [200x200]

loop 100000 [poke img p red p: (pick [100x0 0x200 200x200] random 
3) + p / 2]
view layout [image img]

drop-file.r show how get draggin from windows enviroment. Is possible 
 to initiate a Drag 'n Drop operation from REBOL3? (ie for to send 
an image to another application).

that should renderize 1 image then the second on the lower then block 
of 4 images in sqaure  the below an image then a secon column same 
way

REBOL [
	Title: "REBOL 3 GUI - Development Module Loader"
]

files: [
	%x-funcs.r
	%g-defs.r
	%g-debug.r
	%g-funcs.r
	%g-styles.r
	%g-faces.r
	%g-panels.r
	%g-events.r
	%g-text.r
	%g-effects.r

	%s-fonts.r

	%s-panel.r
	%s-button.r
	%s-bars.r
	%s-text.r
	%s-image.r
	%s-lists.r
]

code: collect [foreach file files [keep load file]]

len-kb?: func [s] [round/to divide length? s 1024 .1]

src: mold/flat code
gui-cmp: compress src
gui-size: len-kb? gui-cmp 

print ["GUI Size:" gui-size "KB /" len-kb? src "KB"]

do code

Today I was supposed to try to edit one corporate .swf file. Oldes 
suggested me some decompiler. There are items like - shape, image, 
font, text, button, frame, action, misc tags ( http://www.xidys.com/pekr/flash-structure.jpg
). It seems like flash (dunno if decompiled view is identikal to 
their proper IDE) is using kind of page-layout engine. Our corporate 
flash presentation is like browsing dynamic web. So far, with VID3, 
I can't see enough of  "rebol browser" in there, but more an app 
dev. platform. What is our equivalent of "page" - is it a 'panel?

Henrik, BrianH: could you please ask Carl to add one rather simple 
style, called Icon? It should consist of Image, text description, 
and it should be selectable (blue translucent color upon selection). 
I would like to see, how such selection is being drawn, if not part 
of draw block. If you want small but real-life app to test, please 
port effect-lab2.r to VID 3.4. We can then compare. Gab's version 
handles dragging and mouse-pointer change too.

We are not working with a single-complex-gob system only, we also 
allow compound gobs, rich-text-only gobs, image gobs, whichever is 
the best for the situation. We even have modes and custom draw handlers 
for more complicated situations. The new GUI's approach is both simpler 
and more flexible than the frames system you describe, even flexible 
enough that you can implement that frames system in your style if 
you like.

Henrik - when I compare your panel skin designs so far (e.g. 40 vs 
48), dunno why, but I like of liked the panel look in image number 
40. It remings me like you mentioned those "natural materials". It 
almost looks like it is carved into the surface. Not sure it would 
work if multiple such panels would be on one page ....

Reichard...it's by design, that you must download the image to see 
it, or it is bug in Qtask? :)

and sunanda... I've also discovered that it doesn't always recycle 
the memory its got !  image manipulation is an example of that.

just a joke to try to generate a coverflow animation.
REBOL []
angle: 10
angle2: 0
decal: 0x0
rot: 0
scene: [
	transform angle 50x0 1 1 0x0
	pen red fill-pen red circle 70x0 2
	pen green fill-pen green circle 85x0 2
	pen yellow fill-pen yellow circle 100x0 2
]
scene2: [
	transform angle2 decal 1 1 0x0
	pen red fill-pen red circle rpair 2 
	pen green fill-pen green circle gpair 2
	pen yellow fill-pen yellow circle ypair 2
]
img: copy empty: make image! 100x50
logo: copy logo-empty: make image! 200x100
locate: func [color /local idx][
	idx: (index? find/skip rgb to binary! color 3) / 3
	as-pair idx  // img/size/x to integer! idx / img/size/x 
]
view/new l: layout [image img image logo]
rev: false
for i 10 170 5 [
	angle: i
	change img empty
	draw img scene	
	rgb: img/rgb
	gpair: locate green
	rpair: locate red 
	ypair: locate yellow 

	either all [gpair/x > 35 gpair/x < 65][
		rev: true
		angle2: angle2 - 15
		decal: gpair
		change img empty
		draw img scene2	
		rgb: img/rgb
		rpair: locate red 
		ypair: locate yellow 
	][if rev [pair: ypair ypair: rpair rpair: pair]]
	change logo logo-empty
	b1: as-pair rpair/x 25 - rpair/y
	b3: as-pair rpair/x 25 + rpair/y 
	b2: as-pair ypair/x 25 - ypair/y
	b4: as-pair ypair/x 25 + ypair/y 
	draw logo [pen red scale 2 2 image logo.gif b1 b3 b4 b2 border]
	show l
	;*** PUSH ENTER TO ADVANCE
	input
]
do-events

I always tought that the load/save refinements for library and image 
formats was pretty strange.

also, for lib usage a GC-independant alloc() dealloc() equivalent 
would be VERY usefull.  something we can manually free, in order 
to work with large dataset we really don't want the GC to grab.  
for example, using image magic, I can work with 10GB image files. 
 composing 100 of those images into another.  I can't let rebol's 
GC play around with such specific needs... this is just one example, 
but there are others.


sometimes handling stuff manually is easier than hoping an automatic 
system will do so properly.  in R2, images are never freed, for example, 
so for a task which rebol would be the ultimate image manipulator, 
I can't use it cause the third image I load from disk craps out rebol, 
since it grows beyond 1.5GB of RAM.