Let me check this… Indeed, it would be great to have a deep discussion about composite shapes. Will you be at ESUG? Cheers, Alexandre -- _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: Alexandre Bergel http://www.bergel.eu ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;.

I'm not sure yet, it depends more on my university than me... but hopefully I will know soon. I have produced version 1.52 of Roassal. ConfigurationOfRoassal2 in the Thanks, but would it be possible to also include Roassal2Spec-johanfabry.6? This would probably be needed in all versions, because otherwise one cannot use Roassal2Spec in latest Pharo. But best is to solve the Trachel issue with the callback. I know, but I need more stable environment (I can easily deal with it, but when other people are working with it and are not necessarily devs, I need a more stable environment for them). I have some ideas about having more free composite shape, where I could specify (either with blocks, or having the option to subclass CompositeShape) what kind of behavior I want, but it definitely needs more refining. Because recently I ran into issue that when resizing a composite shape I wanted to resize only two of them (borders) but not another two (icon and label), so I had to abandon CompositeShape for that problem and do it manually. Thanks, Peter

Hi I'd like to break into your discussion, so we won't talk about the same in two threads. Composite shapes should, to outer world, act line normal shapes, which could be moved, resized, used callbacks on and they should act like we expect usual shapes to. Thing is it is not quite clear what is this "expected" behavior and in my opinion it depends on actual usage of composite shapes, especially with resizing and callbacks. One might want to resize only one border-like shape and keep rest as it is, or to resize all subshapes to new size, or resize all proportionally, like increase width by 50 %. The same is there for callbacks, when we might want to have callback for every change of extent of any inner shape, or we might want to have callback only when one of them changes extent, or only when whole shape changes extent (because when we change extent of shape in the center of larger one, it does not change extent of whole shape). I thought about using replacable blocks for composite shapes actions like adding callbacks, resizing or moving composite shape, reaction to moving or resizing sub-shapes etc. Jan abergel wrote -- View this message in context: http://forum.world.st/Roassal-TRCompositeShape-regression-tp4816726p4818430… Sent from the Moose mailing list archive at Nabble.com.

Excellent!!! Alexandre -- _,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;: Alexandre Bergel http://www.bergel.eu ^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;._,.;:~^~:;.

We already do this. :) (and of course it is draggable and resizable, both the outer and inner elements :)) But since we have an extra layer on top of Roassal we can somewhat offset this problem — each shape has a controller, where we define the necessary behavior; however this is really an abuse of the controller as it has different responsibilities. And if there is no best way or systematic structure than it is always reimplemented manually (of course everywhere differently), so it is a constantly increasing mess... but this will definitely require more thinking and designing to get this whole problem right. Peter

Well yesterday I had to add new method to TRMorph because I was changing dragged shape on drag start... so I would say we are well beyond the limits. But at least there will be more things to discuss at ESUG. :) (And more work for me to clean it all up so it can be integrated into Roassal.) Peter

Hi once again I'd like to bring here together all current problems of composite shapes, add one more and propose one fix. Problem 1) This topic was started because remove callbacks got triggered from same shape multiple times by single event. I think I fixed this problem. And I did few more things to TRCompositeShape while I was at it. I fixed exactly same problem with translation callbacks, I modified composite shape callback test to test this case with removing, and I added and used shapesDo: and firstShape methods to simplify code and make it faster - calling Collection>>#collect:thenDo: is faster than (collect:)do:, because is does not create temporary collection in the middle of it. mcz: http://www.mediafire.com/download/lh37m9n0rq21e7n/Trachel-bliznjan.275.zip Problem 2) Unfortunately, in the meantime, I found much nastier bug, which is not so common to trigger, but it causes infinite loop and it happens in our DynaCASE app. It happens only if: We have a composite shape, canvas of this composite shape is not set (is nil), and then we add TRTranslationCallback to that shape - callback which gets encompassingRectangle of this shape in its block. If canvas is nil, method encompassingRectangle sends (for some reason unclear to me) positionShapesAfterBeingAdded, which translates subshapes, which triggers translation callback, which sends encompassingRectangle, which sends positionShapesAfterBeingAdded etc... Same problem happens by calling many methods which then call encompassingRectanhle, like center, position, extent, height and width. Example code: | shape process | shape := (RTBox new + RTBox new) element. shape addCallback: (TRTranslationCallback block: [shape encompassingRectangle]). process := [shape translateTo: 100@100] fork. (Delay forMilliseconds: 300) wait. process suspend; debug. Problem 3) As we discussed in topic "[Roassal] TRCompositeShape not sending callbacks" ( http://forum.world.st/Roassal-TRCompositeShape-not-sending-callbacks-td4814… ), there is a problem if I use simple addCallback: for translation or extent callbacks, because this way it triggers only when first subshape is changed, but after that, the rest of subshapes get changed, but no callback is triggered anymore. You proposed method addCallbackToAllShapes:, but it would require to make nasty exceptions and type checking for composite shapes in multiple places in our code, because TRRemoveCallbacks have to be triggered only once per composite shape (which seems to be opposite of translation and extent callbacks). Only possible solution that occurs to me is to edit addCallback: just for our need in subclass of RTCompositeShape, or use mentioned replacable blocks for that in TRCompositeShape. Question 1 (maybe Problem 4?)) And to finish this, I have one question. What is and/or should be difference between TRShape>>#position and TRShape>>#center? In case of RTElement, it is explicitely the same, but what about here, where it sometimes is and sometimes isn't the same, especially with composite shapes (but not only them)? Jan Blizničenko -- View this message in context: http://forum.world.st/Roassal-TRCompositeShape-regression-tp4816726p4820464… Sent from the Moose mailing list archive at Nabble.com.

I do not think that using anything else besides center is a good idea. There are many places (e.g. TRConstraint) that counts on the fact that position is a center. Other option would be to always return topLeft corner, however the most important part that it is always the same. If you wouldn't know what the position refers to than you can hardly use it. Peter