UI Mappers: An Introduction

14 Comments

• I wrote a UI mapper in ASP, which generated XSLT (and applied the transform client-side or server-side) (we used SqlXml), and whilst it worked very well, the hardest thing was form layout. PITA.
  
  
  
  Probably easier now that I understand CSS.
  
  
  
  I must admit to thinking that ASP.NET kind of makes it unnecessary in a lot of cases. Building a form/grid isn't too tricky now (as it could be with ASP).
  

• <i> I've got my own in beta now -- what would you like to see in order to make it truly worthwhile?</i>
  
  
  
  I'd like to see it work in a production environment. ;)
  
  
  
  

• I'd love to see this.
  

• I'd like to see it based around the Observer pattern, personally - I've found that using the MVC pattern with the observer pattern works REALLY well..... tho I guess all of these could be used together..... :)
  

• I think it is great idea. It's another must have part of truly solid framework(like o/r mapper). The big question is does asp.net 2.0 with ObjectDataSource and Grid/Form View with two-way databindings eliminates this need?
  

• Well, I would say that this sounds like a very waterdowned version of XAML. The alternative, what you are trying to achieve is to have two providers for the rendering of the UI file: one for ASP.NET applications and one for Windows Forms.
  

• UI Mapper: Meta-data driven, enables personalization and localization, avoid boring data-hookup, ultimate reuse
  
  ASP.NET Designer: Total control of layout, but requires lots of boring data-hookup code, and little to no reuse
  
  ASP.NET v2.0: Many enhancements make easier, but not revolutionary changes, so still need code and little reuse
  
  UIP AB: Concerned with navigation, not layout of individual list/edit screens, same for other workflow engines
  
  XAML: XAML is like ASP.NET markup for Windows, and still requires boring data-hookup code, and very little reuse
  
  WebMaestro: Appears to be just another template/skinning system, no list/edit screens, nor any data integration
  

• To Phil:
  
  
  
  I got an email from a Phil in Switzerland, but I was not able to reply to the email address -- so maybe you can send me another email address instead.
  
  
  
  Thanks, Paul Wilson
  

• The main thing would be performance I think. I can generate list and edit usercontrols very easily using generation. The benefit with using a template generation (shameless plug of CodeSmith) is that I control the implementation and the style of the code.
  
  
  
  I think using a UI Mapper would negatively affect performance, but I am open to the idea I suppose.
  

• Me and my team researched O/R UI Mappers for one of our projects and we are now testing a solution from Modularis. Do any of you have experience with this product?
  

• Performance -- My initial data suggests that my list is at least no worse than using a datagrid, and probably better actually. Why? The datagrid (and 3rd party ones only get worse) is very very heavy, and creates a heck of a lot of controls. My ASP.NET list is just an html table that is rendered much more efficiently, and my mappings are cached. The edit "control" does perform slightly less, but its a fraction of a second that is insignificant to me, just like any list gain I have is insignificant to me. In other words, there are some tiny performance gains (list) or losses (edit), but they are miniscule when compared with working your database and sending the response to the client, so . . .
  
  
  
  Anyhow, I have nothing against code generation, and heartily recommend CodeSmith for it, if you need complete control, as opposed to additional flexibility driven by meta-data.
  
  
  
  Modularis -- never heard of it, but looks interesting if you are going down the code generation path -- but make sure you look at what it generates and hopefully you will also be able to modify the templates it uses. DeKlarit is also a very good tool if you want to code gen an entire app, but I would personally use CodeSmith if I wanted to code gen since it gives you complete control -- I have used DeKlarit though as an excellent RAD prototyper.
  

• I actually wrote something very similar. I had my business classes that implemented a fairly standard intereface. I then added a few attributes that controlled how they were displayed. The attributes could be applied either on the original class or on any class that inherits it. (For my purposes, I usually created a class that inherited my original class, and called it View).
  
  
  
  So for example, my code would look like this:
  
  public class Data
  
  {
  
  // fields
  
  }
  
  
  
  [Exposed(true)]
  
  [ExpectRole(Members.Editors]
  
  public class DataView
  
  {
  
  [Validators(NotNull)]
  
  [ReadOnly(false)]
  
  public string Field
  
  {
  
  //...
  
  }
  
  }
  
  
  
  Anyways, I expanded out the library a bit, and I have used it for several clients.
  

• I've written both an O/R mapper and UI Mapper using CodeSmith templates. We use it at the company for which I work to generate business objects and simple list/edit views on those objects.
  
  
  
  The UI Mapper is only a decent headstart though. It's not configurable and only emits text boxes for input fields and no validators. It also only targets web forms. It only takes a few minutes to change certain text boxes to drop down lists if needed, and add validation.
  
  
  
  It's definitely something worthwhile creating. An xml-based UI Mapper with all the features that you mentioned would be awsome. I've never looked into a third party option.
  

• Thanks for making-up the name "UI mappers". I've been writing "UI mappers" for the last 5 years and my best at naming those was "metadata UI frameworks for database applications" which is way too nerdy.