[CS3 Win] Doc-Observer, new architecture

Question

Hi all,<br /><br />so I finally get around to port stuff to CS3. While things have gone smoothly until now, I'm obviously getting to the tougher part now: the new command-architecture.<br />I've written a command to import text from a file. I guess I don't really need my own command here, but in CS2 I could do all the stuff needed at undo in the Undo()-Method of my command. Needless to say that this is obsolete now. So I figured I'd just broadcast changes on my own PMIID in the commands DoNotify() and use a doc-observer with LazyUpdate(), since that'll tell me of changes not only at Do() but also at Undo()/Redo() - I hope so at least. Unfortunately attaching the observer according to the porting-guide/API-reference crashes InDesign.<br />Now, here's some code, I've stripped it of nil-pointer checks for breviety:<br /><br />from the DoNotify():<br /><br />IDocument* doc = Utils<ILayoutUIUtils>()->GetFrontDocument();<br />InterfacePtr<IMCMImportTextData> importData(this, UseDefaultIID());<br />InterfacePtr<ISubject> docSubject(doc, UseDefaultIID());<br />ListLazyNotificationData<PMString>* lnData = new ListLazyNotificationData<PMString>;<br />lnData->ItemChanged(importData->GetPath());<br />docSubject->ModelChange(kMCMImportTextCmdBoss, IID_IMCMIMPORTTEXTDATA, this, lnData);<br /><br />IMCMImportTextData ist an interface aggregated on my CmdBoss to allow for Data-Exchange, in this case it's a path to file, still need to change that to an IDFile due to the changes to PMString. Seems to work ...<br /><br />Now the attach-method on my doc-observer - at which point ID crashes when opening the doc:<br /><br />InterfacePtr<ISubject> iDocSubject(iDocument, UseDefaultIID());<br />if (!iDocSubject->IsAttached(this, IID_IMCMIMPORTTEXTDATA, ISubject::kRegularAttachment)){<br />iDocSubject->AttachObserver(this, IID_IMCMIMPORTTEXTDATA, ISubject::kRegularAttachment);<br />}<br />if (!iDocSubject->IsAttached(this, IID_IHIERARCHY, ISubject::kRegularAttachment)){<br />iDocSubject->AttachObserver(this, IID_IHIERARCHY, ISubject::kRegularAttachment);<br />}<br /><br />iDocument is an IDocument* passed as a parameter, I've also tried the various AttachmentTypes. If I do it the "old-fashioned-way" (iDocSubject->AttachObserver(this, IID_IMCMIMPORTTEXTDATA, IID_MCMDOCOBSERVER)) it works, but obviously only Update() will be called, not UpdateLazy() which I'll prefer and need. I used IID_IHIERARCHY for test-purposes, commenting out my own IID - unfortunately with the same result.<br />Now I'm stuck - I'm not sure whether I'm missing something basic here, if I just didn't get the documentation or I don't know what.<br /><br />Thanks in advance for any help,<br /><br />Bernt

Answer 1

The selection filter was mentioned for a very specific use case - Jelle had to notify an external application. For details see that other thread, but for you that approach is probably wrong anyway.
Typically you follow the selection in order to update a widget - if not, please explain.
For that case you would aggregate a selection observer (e.g. derived from ActiveSelectionObserver* ) on your widget or its parent panel. In the observer, you would _not_ override Update(), instead use the various Handle() methods.
That notification should be sufficient for simple selection changes on page items, the selection should not need to follow multiple protocols. If there is a deeper sense behind your quoted enormous collection of protocols, then you might need a Suite with SelectionExt* which would preprocess all those notifications and reduce them to a single custom protocol. You'll probably need a Suite anyway unless you can reuse an existing one, to obtain a value for your widget.
In other words, all the _DOCUMENT are not related to the selection, they are just side effects. IID_ISELECTIONFILTER makes no sense. All those PATHSELECTION are only relevant if you want to watch every change to shapes. The reference point again is a sideeffect - I think of scrolling / active spread, it is very unlikely that you really need that notification.
Btw, do not check for IsAttached() during AutoAttach. It just suppresses an ASSERT, better fix the reason (unbalanced Attach / Detach).
* = search the SDK.
Dirk