Community
Participate
Working Groups
Created attachment 223807 [details] Reproducible test case. There is no point in opening all diagrams in a model at once, and especially not when a model has been split into parts. What about "lazy loading" ? How is it affected when all diagrams are opened? This should be investigated. Attached is a ws with proj backup and ppt slide.
Diagrams that are opened when you open a model are the one that was still open when you closed the model. Papyrus keeps the history of its configuration including which views (both diagrams and tables) was open when a model is closed. Some users of Papyrus do want this feature. I will asked other Ericsson users to know their views on your proposal. One possiblity is to have preference dedicated to either save or not the configuration when closing a model.
Some additional comments on this one: It is OK to leave the tabs for the diagrams that were "open" when you last closed the model, but it is not OK to open all diagrams again the next time you open the model, that is, to load all diagram contents again. (related to bugzilla 394911) Then one can wonder where the "configuration history" should be stored, and why a model is/were marked as "touched" as soon as you switch diagram. (related to bugzilla 394775) You then get/got this asterisk saying that the model was modified and should be saved, and if you do save the model, all the files are saved. (?) (that is, the .di, the .notation and the .uml file) This "configuration history" should perhaps instead be stored in the .project file or some other non-model-file.
By "OK to leave the tabs" above I mean that it's OK to show the tabs but not OK to load the model content / content of "hidden" diagrams which are not currently shown, until the user choose to display the content. This is against your presented Model loading policy saying "Avoid unneeded model part loading". For that reason this bugzilla can not be regarded as an "enhancement", since the selected model loading policy for "needed" resources is buggy.
I'm still unsure this is even possible, unless the diagrams themselves are Controlled. In EMF, the Resource is a loading unit. But even if we put each diagram in a single resource, we will need to load that resource for retrieving the diagram name and type (Displayed in the Tab and in the ModelExplorer). It might be interesting to duplicate this information, and store it in the *.di file, in the "available pages", so that we don't need to load the whole notation resource to get the diagram name and type.
There is no "Related to", so I added to "Depends on".
Because the main performance issue is related to the "nested editors" being loaded (And not the Model itself), I rename this task
Lazy diagram opening has been implemented in the context of bug 519107
