Best practice with MVVM

[andreiim]

I'm on a project that will have about a dozen of commands. Where should I put them?

For each viewmodel I have a view and a main model.

My commands only show the views, so it would make sense to be in the View part of the MVVM. But, if I put the command in the view's codebehind, autocad would instantiate the view for the whole lifetime of the document and I don't want this.

If I create a separate class for each command using a naming system like NameOfTheViewCommand, then I would practically have an MVVMC pattern, where C stand for command.

Or maybe I should create a class which includes all my commands and nothing else?

Or maybe all these approaches are wrong and there's already a well-established pattern known only by Autocad .NET experts and one of these gurus is kind enough to provide a best practice guide for this matter?

[fenton.webb]

in terms of AutoCAD, we have a 3 level split...

1) Dependency on acad.exe

   - define your commands that rely on the AutoCAD UI here

2) Dependency on accore.dll

  - define your commands that only require the command line to work

3) Dependency on acdb19.dll

  - define your DWG manipulate code here

I recommend that all of your commands define a UI version and a non-UI version - e.g. LAYER (UI) -LAYER command line only. Your UI version should be in a DLL which references acmgd.dll (acad.exe), acdbmgd.dl (acdb19.dlll and accoremgd.dll (accore.dll)

I recommend that all of your -LAYER defined commands should be in a DLL which references accoremgd.dll, acdbmgd.dll

I recommend all of your DB related functionality reside in a DLL which references acdbmgd.dll.

Doing this will ensure:

1) Your commands can be scripted

2) Your commands will work on the cloud

3) Your code will work in other non-AutoCAD DWG hostable applications (RealDWG enabled)