On 31.03.2016, at 19:43, commits@source.squeak.org wrote:
A new version of Graphics was added to project The Inbox: http://source.squeak.org/inbox/Graphics-cmm.329.mcz
==================== Summary ====================
Name: Graphics-cmm.329 Author: cmm Time: 31 March 2016, 2:43:44.434481 pm UUID: a7a78301-477c-40ea-b010-c4e71c02a038 Ancestors: Graphics-tfel.328
Fix the green clone halo morph for SketchMorphs to not share its underlying Form.
Err. The clone handle is *supposed to* not duplicate the form. This is used a lot in Etoys. You can have many players sharing the same bitmap costume. We do *not* want to waste memory duplicating the form.
While editing the sketch, we work on a copy of the form (to support canceling). If accepting the painting, *then* we set it to the new form.
- Bert -
Okay, that's why this is in the Inbox. I started to formulate my question in long words in email, but decided this was the quickest and easiest way to ask it. :)
Okay, so this is definitely not the place to fix my problem. So I will instead try putting something into SketchMorph's veryDeepFixup: which will ensure its underlying form gets copied when I green-halo a SketchMorph. SketchMorph is an exception because he's a dumb data object, not a "player" who wants to share a costume.
-- I was working on an icon, writing scripts to change colors pixel-by-pixel, goofed up my "backup" SketchMorph I had torn off the green halo, only to find I corrupted my original, as well. Some work lost.
We should use real morphs as halo handles, not only forms. Then we can define/adjust that deep-copy operation there.
Nevertheless, Chris is right that the override of #veryDeepCopyWith: in Form outside any Morphic extension category is quite smelly. Deep copying is not Morphic-specific.
Best, Marcel
-- View this message in context: http://forum.world.st/The-Inbox-Graphics-cmm-329-mcz-tp4887677p4887736.html Sent from the Squeak - Dev mailing list archive at Nabble.com.
You can hold shift while dragging the green handle to change it's copying behavior. One makes a sibling, so they share the same form. One makes a copy of the morph and form.
Best, Karl
On Fri, Apr 1, 2016 at 9:43 AM, marcel.taeumel Marcel.Taeumel@hpi.de wrote:
We should use real morphs as halo handles, not only forms. Then we can define/adjust that deep-copy operation there.
Nevertheless, Chris is right that the override of #veryDeepCopyWith: in Form outside any Morphic extension category is quite smelly. Deep copying is not Morphic-specific.
Best, Marcel
-- View this message in context: http://forum.world.st/The-Inbox-Graphics-cmm-329-mcz-tp4887677p4887736.html Sent from the Squeak - Dev mailing list archive at Nabble.com.
On 01.04.2016, at 14:33, karl ramberg karlramberg@gmail.com wrote:
You can hold shift while dragging the green handle to change it's copying behavior. One makes a sibling, so they share the same form. One makes a copy of the morph and form.
Not true. Both make a copy of the morph, both do *not* copy the form.
The only difference is that the morph’s player’s class is the same for siblings, but unique for copies. Hence, scripts (methods) are shared between siblings, but independent for copies.
- Bert -
On Fri, Apr 1, 2016 at 9:43 AM, marcel.taeumel <Marcel.Taeumel@hpi.de mailto:Marcel.Taeumel@hpi.de> wrote: We should use real morphs as halo handles, not only forms. Then we can define/adjust that deep-copy operation there.
Nevertheless, Chris is right that the override of #veryDeepCopyWith: in Form outside any Morphic extension category is quite smelly. Deep copying is not Morphic-specific.
Best, Marcel
-- View this message in context: http://forum.world.st/The-Inbox-Graphics-cmm-329-mcz-tp4887677p4887736.html http://forum.world.st/The-Inbox-Graphics-cmm-329-mcz-tp4887677p4887736.html Sent from the Squeak - Dev mailing list archive at Nabble.com.
Ok, I miss understood the difference
Best, Karl
On Fri, Apr 1, 2016 at 3:29 PM, Bert Freudenberg bert@freudenbergs.de wrote:
On 01.04.2016, at 14:33, karl ramberg karlramberg@gmail.com wrote:
You can hold shift while dragging the green handle to change it's copying behavior. One makes a sibling, so they share the same form. One makes a copy of the morph and form.
Not true. Both make a copy of the morph, both do *not* copy the form.
The only difference is that the morph’s player’s class is the same for siblings, but unique for copies. Hence, scripts (methods) are shared between siblings, but independent for copies.
- Bert -
On Fri, Apr 1, 2016 at 9:43 AM, marcel.taeumel Marcel.Taeumel@hpi.de wrote:
We should use real morphs as halo handles, not only forms. Then we can define/adjust that deep-copy operation there.
Nevertheless, Chris is right that the override of #veryDeepCopyWith: in Form outside any Morphic extension category is quite smelly. Deep copying is not Morphic-specific.
Best, Marcel
-- View this message in context: http://forum.world.st/The-Inbox-Graphics-cmm-329-mcz-tp4887677p4887736.html Sent from the Squeak - Dev mailing list archive at Nabble.com http://nabble.com.
On 01.04.2016, at 01:41, Chris Muller asqueaker@gmail.com wrote:
Okay, that's why this is in the Inbox. I started to formulate my question in long words in email, but decided this was the quickest and easiest way to ask it. :)
Okay, so this is definitely not the place to fix my problem. So I will instead try putting something into SketchMorph's veryDeepFixup: which will ensure its underlying form gets copied when I green-halo a SketchMorph. SketchMorph is an exception because he's a dumb data object, not a "player" who wants to share a costume.
SketchMorph is the single most-used user object in Etoys :)
-- I was working on an icon, writing scripts to change colors pixel-by-pixel, goofed up my "backup" SketchMorph I had torn off the green halo, only to find I corrupted my original, as well. Some work lost.
I’m sorry you lost time. Nevertheless, the originalForm in SketchMorph is designed to be shared.
- Bert -
Really? That's really surprising; SketchMorph is just a dumb data object, a painting, making a copy of it via the UI is a valid use-case. For every other kind of Morph, the green halo makes a deep copy, SketchMorph is inconsistent, shouldn't Etoys simply override that halo to call its own #semiDeepCopy to suit its own purposes?
Is Etoys running in trunk? It is not targeted for trunk though, is it?
I want to be able to make easy copies of SketchMorph's. Surely we can come up with some solution; what about Karl's suggestion? I think Shift should be the sharing, because the "modifer" keys are used to "modify" the standard behavior, which is a true veryDeepCopy..
On Fri, Apr 1, 2016 at 8:45 AM, Bert Freudenberg bert@freudenbergs.de wrote:
On 01.04.2016, at 01:41, Chris Muller asqueaker@gmail.com wrote:
Okay, that's why this is in the Inbox. I started to formulate my question in long words in email, but decided this was the quickest and easiest way to ask it. :)
Okay, so this is definitely not the place to fix my problem. So I will instead try putting something into SketchMorph's veryDeepFixup: which will ensure its underlying form gets copied when I green-halo a SketchMorph. SketchMorph is an exception because he's a dumb data object, not a "player" who wants to share a costume.
SketchMorph is the single most-used user object in Etoys :)
-- I was working on an icon, writing scripts to change colors pixel-by-pixel, goofed up my "backup" SketchMorph I had torn off the green halo, only to find I corrupted my original, as well. Some work lost.
I’m sorry you lost time. Nevertheless, the originalForm in SketchMorph is designed to be shared.
- Bert -
Yes of course Etoys should work in trunk!
It would be really good if some members of the squeak-dev community could take an interest in this, and maybe make a project out of doing whatever updates may be needed to clean up trunk in this regard. The Etoys image itself provides a clear reference of how things should work, and I'm sure that Bert and others would be happy to offer some guidance and oversight to anyone who might want to contribute.
Personally I think it would be very cool if you could open a new trunk image, start a new project, and be able to select Morphic, MVC, or Etoys as types of project. Then entering a new Etoys project could take you directly into an environment just like that of the reference Etoys image.
Dave
Really? That's really surprising; SketchMorph is just a dumb data object, a painting, making a copy of it via the UI is a valid use-case. For every other kind of Morph, the green halo makes a deep copy, SketchMorph is inconsistent, shouldn't Etoys simply override that halo to call its own #semiDeepCopy to suit its own purposes?
Is Etoys running in trunk? It is not targeted for trunk though, is it?
I want to be able to make easy copies of SketchMorph's. Surely we can come up with some solution; what about Karl's suggestion? I think Shift should be the sharing, because the "modifer" keys are used to "modify" the standard behavior, which is a true veryDeepCopy..
On Fri, Apr 1, 2016 at 8:45 AM, Bert Freudenberg bert@freudenbergs.de wrote:
On 01.04.2016, at 01:41, Chris Muller asqueaker@gmail.com wrote:
Okay, that's why this is in the Inbox. I started to formulate my question in long words in email, but decided this was the quickest and easiest way to ask it. :)
Okay, so this is definitely not the place to fix my problem. So I will instead try putting something into SketchMorph's veryDeepFixup: which will ensure its underlying form gets copied when I green-halo a SketchMorph. SketchMorph is an exception because he's a dumb data object, not a "player" who wants to share a costume.
SketchMorph is the single most-used user object in Etoys :)
-- I was working on an icon, writing scripts to change colors pixel-by-pixel, goofed up my "backup" SketchMorph I had torn off the green halo, only to find I corrupted my original, as well. Some work lost.
Iâm sorry you lost time. Nevertheless, the originalForm in SketchMorph is designed to be shared.
- Bert -
On 01.04.2016, at 17:18, David T. Lewis lewis@mail.msen.com wrote:
Yes of course Etoys should work in trunk!
It would be really good if some members of the squeak-dev community could take an interest in this, and maybe make a project out of doing whatever updates may be needed to clean up trunk in this regard. The Etoys image itself provides a clear reference of how things should work, and I'm sure that Bert and others would be happy to offer some guidance and oversight to anyone who might want to contribute.
Personally I think it would be very cool if you could open a new trunk image, start a new project, and be able to select Morphic, MVC, or Etoys as types of project. Then entering a new Etoys project could take you directly into an environment just like that of the reference Etoys image.
Noted.
Dave
Really? That's really surprising; SketchMorph is just a dumb data object, a painting, making a copy of it via the UI is a valid use-case. For every other kind of Morph, the green halo makes a deep copy, SketchMorph is inconsistent, shouldn't Etoys simply override that halo to call its own #semiDeepCopy to suit its own purposes?
Is Etoys running in trunk? It is not targeted for trunk though, is it?
I want to be able to make easy copies of SketchMorph's. Surely we can come up with some solution; what about Karl's suggestion? I think Shift should be the sharing, because the "modifer" keys are used to "modify" the standard behavior, which is a true veryDeepCopy..
On Fri, Apr 1, 2016 at 8:45 AM, Bert Freudenberg bert@freudenbergs.de wrote:
On 01.04.2016, at 01:41, Chris Muller asqueaker@gmail.com wrote:
Okay, that's why this is in the Inbox. I started to formulate my question in long words in email, but decided this was the quickest and easiest way to ask it. :)
Okay, so this is definitely not the place to fix my problem. So I will instead try putting something into SketchMorph's veryDeepFixup: which will ensure its underlying form gets copied when I green-halo a SketchMorph. SketchMorph is an exception because he's a dumb data object, not a "player" who wants to share a costume.
SketchMorph is the single most-used user object in Etoys :)
-- I was working on an icon, writing scripts to change colors pixel-by-pixel, goofed up my "backup" SketchMorph I had torn off the green halo, only to find I corrupted my original, as well. Some work lost.
I’m sorry you lost time. Nevertheless, the originalForm in SketchMorph is designed to be shared.
- Bert -
On 01.04.2016, at 16:53, Chris Muller asqueaker@gmail.com wrote:
Really? That's really surprising; SketchMorph is just a dumb data object, a painting, making a copy of it via the UI is a valid use-case. For every other kind of Morph, the green halo makes a deep copy.
Every other morph is cheap to copy, compared to deep-copying bitmaps.
SketchMorph is inconsistent, shouldn't Etoys simply override that halo to call its own #semiDeepCopy to suit its own purposes?
Is Etoys running in trunk?
Yes it is. There are even a few people using it (e.g. https://realm.io/news/altconf-daniel-steinberg-swift-playgrounds-education/ ).
It is not targeted for trunk though, is it?
It’s still desirable to retire the Etoys fork of Squeak. Having a unified codebase (and faster VM) would be very nice. We may finally have found students to work on that, but even if it does not happen soon it’s still the plan.
I want to be able to make easy copies of SketchMorph's. Surely we can come up with some solution; what about Karl's suggestion? I think Shift should be the sharing, because the "modifer" keys are used to "modify" the standard behavior, which is a true veryDeepCopy..
Why not adhere to the copy-on-write scheme for forms in Morphic? That’s how it was designed.
E.g. when you make a green-handle copy of a SystemWindow, why would you want to deep-copy all the icons?
Instead, before you modify a form (which is an extremely rare operation compared to all uses of forms), make a copy of it. That is how all the Sketch editing functions work (e.g. in the halo’s “painting…” submenu).
Btw you may consider using an ImageMorph if you don’t actually need all the specialized SketchMorph functionality. It does share the form when deep-copying too, though.
- Bert -
On Fri, Apr 1, 2016 at 8:45 AM, Bert Freudenberg bert@freudenbergs.de wrote:
On 01.04.2016, at 01:41, Chris Muller asqueaker@gmail.com wrote:
Okay, that's why this is in the Inbox. I started to formulate my question in long words in email, but decided this was the quickest and easiest way to ask it. :)
Okay, so this is definitely not the place to fix my problem. So I will instead try putting something into SketchMorph's veryDeepFixup: which will ensure its underlying form gets copied when I green-halo a SketchMorph. SketchMorph is an exception because he's a dumb data object, not a "player" who wants to share a costume.
SketchMorph is the single most-used user object in Etoys :)
-- I was working on an icon, writing scripts to change colors pixel-by-pixel, goofed up my "backup" SketchMorph I had torn off the green halo, only to find I corrupted my original, as well. Some work lost.
I’m sorry you lost time. Nevertheless, the originalForm in SketchMorph is designed to be shared.
- Bert -
It’s still desirable to retire the Etoys fork of Squeak. Having a unified codebase (and faster VM) would be very nice. We may finally have found students to work on that, but even if it does not happen soon it’s still the plan.
+1
Stef
Hi All,
On Apr 1, 2016, at 9:45 AM, Bert Freudenberg bert@freudenbergs.de wrote:
On 01.04.2016, at 16:53, Chris Muller asqueaker@gmail.com wrote:
Really? That's really surprising; SketchMorph is just a dumb data object, a painting, making a copy of it via the UI is a valid use-case. For every other kind of Morph, the green halo makes a deep copy.
Every other morph is cheap to copy, compared to deep-copying bitmaps.
That makes sense.
SketchMorph is inconsistent, shouldn't Etoys simply override that halo to call its own #semiDeepCopy to suit its own purposes?
Is Etoys running in trunk?
Yes it is. There are even a few people using it (e.g. https://realm.io/news/altconf-daniel-steinberg-swift-playgrounds-education/ ).
It is not targeted for trunk though, is it?
It’s still desirable to retire the Etoys fork of Squeak. Having a unified codebase (and faster VM) would be very nice. We may finally have found students to work on that, but even if it does not happen soon it’s still the plan.
+1
I want to be able to make easy copies of SketchMorph's. Surely we can come up with some solution; what about Karl's suggestion? I think Shift should be the sharing, because the "modifer" keys are used to "modify" the standard behavior, which is a true veryDeepCopy..
Why not adhere to the copy-on-write scheme for forms in Morphic? That’s how it was designed.
E.g. when you make a green-handle copy of a SystemWindow, why would you want to deep-copy all the icons?
Instead, before you modify a form (which is an extremely rare operation compared to all uses of forms), make a copy of it. That is how all the Sketch editing functions work (e.g. in the halo’s “painting…” submenu).
+1. But it needs to be documented in all the relevant Morph veryDeepCopy methods, not just SketchMorph but Morph too (& possibly in the class comments?).
I suspect that too often we write simple schemes like this that over time get obscured by layers of new code (I see uniclass support in morph VDC methods) and even if the original intent was clear in the first version, in these later versions it is difficult to see. Conventions like this /must/ be well-documented to thrive.
Btw you may consider using an ImageMorph if you don’t actually need all the specialized SketchMorph functionality. It does share the form when deep-copying too, though.
- Bert -
On Fri, Apr 1, 2016 at 8:45 AM, Bert Freudenberg bert@freudenbergs.de wrote:
On 01.04.2016, at 01:41, Chris Muller asqueaker@gmail.com wrote:
Okay, that's why this is in the Inbox. I started to formulate my question in long words in email, but decided this was the quickest and easiest way to ask it. :)
Okay, so this is definitely not the place to fix my problem. So I will instead try putting something into SketchMorph's veryDeepFixup: which will ensure its underlying form gets copied when I green-halo a SketchMorph. SketchMorph is an exception because he's a dumb data object, not a "player" who wants to share a costume.
SketchMorph is the single most-used user object in Etoys :)
-- I was working on an icon, writing scripts to change colors pixel-by-pixel, goofed up my "backup" SketchMorph I had torn off the green halo, only to find I corrupted my original, as well. Some work lost.
I’m sorry you lost time. Nevertheless, the originalForm in SketchMorph is designed to be shared.
- Bert -
Why not adhere to the copy-on-write scheme for forms in Morphic? That’s how it was designed.
E.g. when you make a green-handle copy of a SystemWindow, why would you want to deep-copy all the icons?
I don't. I want to deepCopy the form bits when I invoke the green-halo it on a SketchMorph, not a SystemWindow. That's what Morph-cmm.1096 does, without affecting it when copying SystemWindows or anything else besides a SketchMorph.
But you said the problem with that is that Etoys students are putting their behaviors directly on SketchMorphs instead of something more abstract like a PlayerMorph. With a PlayerMorph they could set his #fillStyle to some Form which would still share on veryDeepCopy, and yet with my #veryDeepFixup: in SketchMorph, the green-halo would ALSO work properly when invoked directly on a SketchMorph -- the best of both worlds.
But we have to break this common use-case so that this design choice to use SketchMorphs everywhere, will be more memory-efficient.. Sigh..
Instead, before you modify a form (which is an extremely rare operation compared to all uses of forms), make a copy of it.
As I said, that's exactly what I did -- or thought I was doing. I just did it via the green halo because I expected it to behave as it does on all other Morphs -- to make a meaningfully separate copy.
On 02.04.2016, at 20:39, Chris Muller asqueaker@gmail.com wrote:
Why not adhere to the copy-on-write scheme for forms in Morphic? That’s how it was designed.
E.g. when you make a green-handle copy of a SystemWindow, why would you want to deep-copy all the icons?
I don't. I want to deepCopy the form bits when I invoke the green-halo it on a SketchMorph, not a SystemWindow. That's what Morph-cmm.1096 does, without affecting it when copying SystemWindows or anything else besides a SketchMorph.
But you said the problem with that is that Etoys students are putting their behaviors directly on SketchMorphs instead of something more abstract like a PlayerMorph. With a PlayerMorph they could set his #fillStyle to some Form which would still share on veryDeepCopy, and yet with my #veryDeepFixup: in SketchMorph, the green-halo would ALSO work properly when invoked directly on a SketchMorph -- the best of both worlds.
That’s not at all how Etoys works. Yes it’s a prototype-based fat-object OO system on top of Smalltalk, but that’s not how students think about it. Students do not deal with morphs or fill styles at all.
As in many prototype-based systems, copying is a fundamental operation, and it needs to be efficient. Duplicating the bits of a form is not efficient, especially not when they get stored in projects.
But we have to break this common use-case so that this design choice to use SketchMorphs everywhere, will be more memory-efficient.. Sigh..
I’d say that modifying the bits of a SketchMorph using low-level methods while expecting a copy to not be affected is very far from a common use case. It’s not at all unreasonable, but it is not what it’s designed for, or commonly used for.
Instead, before you modify a form (which is an extremely rare operation compared to all uses of forms), make a copy of it.
As I said, that's exactly what I did -- or thought I was doing. I just did it via the green halo because I expected it to behave as it does on all other Morphs -- to make a meaningfully separate copy.
I can see what you’re getting at, but we can’t put the burden on the students. But let’s think about what it would take to have our cake and eat it, too.
What if we could automate the copy-on-write behavior? If the form used in a SketchMorph could be “locked down” so that any attempt at modifying it could be caught. Then it could either just warn the user, or it could be duplicated automatically.
I can’t easily think of a solution, but maybe it’s possible.
- Bert -
What if we could automate the copy-on-write behavior? If the form used in a SketchMorph could be “locked down” so that any attempt at modifying it could be caught. Then it could either just warn the user, or it could be duplicated automatically.
I do like the sound of such an idea..
I can’t easily think of a solution, but maybe it’s possible.
Okay, lets let it simmer, I'm satisfied we at least explored this idea for now, thanks. I could also consider extending SketchMorph>>#veryDeepFixup: into my own package..
On 03-04-2016, at 8:22 AM, Chris Muller asqueaker@gmail.com wrote:
What if we could automate the copy-on-write behavior? If the form used in a SketchMorph could be “locked down” so that any attempt at modifying it could be caught. Then it could either just warn the user, or it could be duplicated automatically.
I do like the sound of such an idea..
I can’t easily think of a solution, but maybe it’s possible.
Isn’t that what the immutable thingy is for? You try to write, get a trap, work out what happened, make a mutable copy, retry the write...
tim -- tim Rowledge; tim@rowledge.org; http://www.rowledge.org/tim Useful random insult:- S p a c e d o u t .
squeak-dev@lists.squeakfoundation.org
-
Bert Freudenberg -
Chris Muller -
David T. Lewis -
Eliot Miranda -
karl ramberg -
marcel.taeumel -
Stéphane Rollandin -
tim Rowledge -
Tobias Pape