 |
privacy |
||
|
Hangout for experimental confirmation and demonstration of software, computing, and networking. The exercises don't always work out. The professor is a bumbler and the laboratory assistant is a skanky dufus.
Blog Feed Recent Items 
The nfoCentrale Blog Conclave nfoCentrale Associated Sites |
2007-02-03Is There an MVC in the House?I think I need to understand Model-View-Controller. No, I’m pretty certain that I do. {tags: orcmid Java Windows MVC Model-View-Controller modal dialogs Swing JNI} My MVC ignorance came up because I have to build Java GUI applications to test some features of a middleware package that impact GUI behavior. This is a consequence of a highly-successful separation of high-level concerns that leads to a little low-level coordination breakiness. The two ends of the middleware connection don’t have to know anything about each other, but the back end (which runs in the same process as the front) can put up modal dialogs that apply to the window that belongs to the application on the front end. The modal part means that the main Window is supposed to not accept input until the dialog is dismissed. The dialog is supposed to stay atop the application Window until that happens. It’s not working properly right now, and that is a very serious usability matter. The breakiness comes in because the front end is a Java application and the back end is operating in native Windows. The middleware functions are delivered to the application by Java components using JNI. At the back-end connection, it is all native Windows non-GUI code. I needed a simple GUI application to use as a front end so I could observe the difficulties with this arrangement and also confirm the repair that I have in mind. I started out fussing with the first real GUI application in the Swing Tutorial. As far as GUIs go, the little application with one button and one label is enough. But I have more elaborate things happening under the covers than just updating a counter every time the button is pressed. I thought I needed to develop a more elaborate GUI too. Fortunately, I started explaining this in a buddy call with Bill Anderson. As the words came out of my mouth, I realized that I was making it too complicated. I could do just fine with one button and one label to provide a test of what happens when the application underneath the Java GUI does something that provokes a dialog from the back end. I still needed to refactor the Swing Tutorial example so that I could make sense of it for myself and be clear on where it is appropriate to add the application and extend the Controller part. I often learn an application by refactoring until it makes sense to me and I can explain it to myself and others. I did that with the single Java file and it is now larger by the amount of annotation I added, along with the logic that my test application requires. One problem with the example being in a single Java file is that it obscures the separation among the methods and my (or your) understanding which of them are only used in particular levels of the GUI operation. A large part of the code is for initialization, and only one method and some instance data are involved in operation of the running application. Sorting out the roles of the code is exacerbated by the little trick of having a static method of the single-class application fire up an instance of that very class and use it to constitute the view and also provide the Controller (i.e., the ActionListener implementation). This is an useful device, here, but it further obscures the separation of function and the relevance of individual methods. (People who see this trick tend to use it everywhere, and that makes for some really stupid Hello World demonstrations where it adds no value to understanding Java whatsoever.) With the refactoring completed, I think I have a sense for a simple Model-View-Controller situation. At least my test application works properly and demonstrates the known coordination problem in the middleware. My buddy Bill says MVC breaks his brain, and examples seem to fall down somehow. He’s not the only one that is puzzled when MVC is claimed yet you can’t find one of the pieces (usually the M) nicely separated out. I think I get it for my simple case. The sense I make of it also seems to conform to the situations Ayende describes. Having struggled to separate out the parts of an actual, simple case, even the Smalltalky explanation kinda sorta makes sense, although that is very abstract in contrast with the perhaps-overly-concrete instances that come up in comments to Ayende’s posts. In contrast, the effort to characterize MVC in Design Patterns leaves me with my eyes crossed and ringing in my ears. I am happy for now, although I think I will have to know far more when I write the guidance for use of my middleware fixture by Java applications. I already see where I am going to have a serious challenge in raising my understanding to a level where I can provide realistic examples that others can grok. Because operations of the capital-M Model can lead to dialogs and user interaction at the back end, calls into the Model from event-loop entries of the Controller (although nicely thread safe) can take an indefinite time before returning control to the event loop. This will also block the Java event-handler thread while a back-end dialog is up, and that means no front-end windows will be refreshed. So I might have to provide guidance on having Model operations carried out on a separate thread from the Controller procedures. In that event I’ll get to worry about providing guidance on the thread-death oblivion chasm and how to avoid falling into it while also preserving dialog modality. Aieeeeeeee. You might surmise that I am not a GUI kinda guy. You’d be right. Web pages, yes. GUI, no. All of my work on this project has been tested and demonstrated using console applications, whether Java-based or for native Windows. I have used command-line compiles and utility operations for the entire project. My one concession to the modern era is to use jEdit for my Java, C++, and batch scripts. I’m using the Visual C++ 2005 Express Edition command-line compiler and running naked through the Windows Platform SDK. I could have used the VC++ IDE, but I was using jEdit for the Java bits anyhow and it was simpler to stay in one editor. I also use Visual Source Safe and the VS 2005 Express Editions don’t integrate with VSS so there is not much gain to using VC++ projects (although my file organization is compatible with someone doing that). When I put the software up on SourceForge, I will add Subversion to the mix. I always wanted to do GUI work, and I will be doing some more-serious forms of it. I just didn’t want that learning curve in the middle of this project. I am thankful that, for my testing requirements, the modified one-button tutorial example is the least that actually works. I shall now dive into my hand-rolled custom-factory COM code where I am completely at home. Different strokes for different folks, as they say. [update: 2007-02-04T17:57Z the last paragraph needed some tweaking and I couldn’t stand to leave it alone.]
|
||
|
You are navigating Orcmid's Lair. |
template
created 2004-06-17-20:01 -0700 (pdt)
by orcmid |