Hi, Sorry, I didn't read all of your mails and just noticed you had problem regarding layer ordering. Don't know if it helps but it is worth a note: If you want to deactivate dynamic layer assignment for certain data nodes you can add a bool property called "fixedLayer" and set it to true. Those data nodes keep their layer. See QmitkDataStorageTreeModel.cpp:670.
Cheers, Stefan Von: Rostislav Khlebnikov [mailto:rostislav.khlebni...@kcl.ac.uk] Gesendet: Donnerstag, 3. April 2014 14:46 An: mitk-users@lists.sourceforge.net Betreff: Re: [mitk-users] Persistent helper objects, recursive deletion, and incremental scene saving Hi guys, I just tried it with the QSortFilterProxyModel and it works absolutely perfectly (node insertions/deletions as well as context menus). The nodes that need to be hidden are hidden, yet the layers are applied correctly and thus the scene saves and loads smoothly. So ping me if you feel you would like to integrate these changes (probably after 2014.03 release) and I will create the regular bug-branch-pull_request chain. Rostislav. On 03/04/2014 12:43, Rostislav Khlebnikov wrote: An alternative approach would be to add all the nodes to the QmitkDataStorageTreeModel and thus have the layers adjusted for all the nodes and delegate which nodes are actually shown to a QSortFilterProxyModel<http://qt-project.org/doc/qt-4.8/qsortfilterproxymodel.html>. Then the loader could be left as is (although I feel it would make more sence with the changes I described before). This would also nicely separate the core responsibilities of a QmitkDataStorageTreeModel (DataStorage sync) from the actual display of items which will be delegated to QSortFilterProxyModel. Again, please tell me if these changes seem reasonable to you as I am quite eager to implement them and have time for it in the coming days. Rostislav. On 03/04/2014 12:28, Rostislav Khlebnikov wrote: Hi, To Ingmar: thanks for the tip! I can avoid the MITK core modifications at least for hiding the nodes. The layer problem is however still there. To all: I am wondering why does the layer property have such a strong influence on creating the parent-child relationships. Isn't the information about the parent-child relationship in conjunction with a simple topological sort enough? I feel it would be better to take the UID-based information to create a graph, then make a topo-sort on it and only then perhaps to use the layer property to sort the sibling items? Would such a contribution to SceneReaderV1 be welcome? Rostislav. On 01/04/2014 20:01, Wegner wrote: Hi Rostislav, did you see, that you also can use nodes without a dedicated BaseData (empty node) to hide child-nodes? There is an option in the Workbench preferences (section DataManager) where you can toggle visibility of these nodes. Default behavior is (i think) that empty nodes are not shown in the DataManager. All child-nodes are also invisible. For debug purposes you could activate the preference and for regular users you can deactivate it again. Maybe that solves one of your problems. ;) Ingmnar Am 01.04.2014 19:52, schrieb Rostislav Khlebnikov: Hi guys, it seems to me that there is no way to save the objects marked as "helper object" when the scene is being saved (it is kinda hard-coded in the scene writer). While I feel that this is absolutely justified (these are "helper objects" after all), I would like to have some nodes hidden from the user, but which would be saved with a project. My use case is: I have the PlanarFigure data node which represents a segmentation of a vessel path cross section. It could be a simple circle created manually, but it can also be a result of smoothing the boundary of a 2D segmentation. The segmentation is performed on a small (say, 50x50mm) slice extracted from a reference image and I want to save this slice along with the segmented slice image to the scene. On the other hand I don't want to clutter the Data Manager with these nodes. After looking at the QmitkDataStorageTreeModel and QmitkExtFileSaveProjectAction it seems to me that this behavior is impossible without modifications to the MITK code. What do you think would be a good approach to this problem? For now I have modified the QmitkDataStorageTreeModel to not show the nodes marked as ("hidden object" || "helper object"). It works fine except that the layers are not adjusted for helper objects and I have to set them manually for the scene to load correctly. Overall, I feel that this is quite a dirty hack and I would like to hear your opinion on this. There is still one more problem, however. As the nodes are not visible in the Data Manager, they cannot be directly deleted when the parent node is. I can obviously handle this in my view which creates these objects, but what if the view is closed? It would be nice if I could mark a node as "Requires hierarchical deletion", which would be processed by the data storage. What would be the best approach to this? Perhaps having a data storage listener which would be registered when my plugin is loaded? Or having my own datastorage subclass which would be set as the active data storage? Or do you think this might be useful for a wider range of applications and could be implemented in the StandaloneDataStorage? Perhaps, a ContourModelSet might be just a node with metainformation about its hidden children - then the separate mappers, interactors and IO classes /for the set /wouldn't be necessary anymore and it might rely on the ContourModel subclasses for that. I feel that this is a topic for a separate discussion though and I will write another e-mail about that as it is unclear how to implement some of the necessary functionalities, but I do believe this is a more flexible approach. One unrelated thing is the incremental scene saving. What I mean by that is while a user works on a project it would be nice to save the progress from time to time. However, the fact that saving an image takes a very significant amount of time and that it is being repacked evey time the scene is saved, I'm afraid that the users will be discouraged to do so. So it would be nice to speed up the saving process - perhaps by detecting that the image hasn't changed and it doesn't need to be repacked or by having the save process in a separate thread. I'm pretty sure such things already came up and perhaps there is already some solution to this. Thank you, Rostislav. ------------------------------------------------------------------------------ _______________________________________________ mitk-users mailing list mitk-users@lists.sourceforge.net<mailto:mitk-users@lists.sourceforge.net> https://lists.sourceforge.net/lists/listinfo/mitk-users ------------------------------------------------------------------------------ _______________________________________________ mitk-users mailing list mitk-users@lists.sourceforge.net<mailto:mitk-users@lists.sourceforge.net> https://lists.sourceforge.net/lists/listinfo/mitk-users ------------------------------------------------------------------------------ _______________________________________________ mitk-users mailing list mitk-users@lists.sourceforge.net<mailto:mitk-users@lists.sourceforge.net> https://lists.sourceforge.net/lists/listinfo/mitk-users ------------------------------------------------------------------------------ _______________________________________________ mitk-users mailing list mitk-users@lists.sourceforge.net<mailto:mitk-users@lists.sourceforge.net> https://lists.sourceforge.net/lists/listinfo/mitk-users
------------------------------------------------------------------------------
_______________________________________________ mitk-users mailing list mitk-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/mitk-users
