RESTful iPhone client and Model hierarchy

问题:

I'm curious what strategy most of you are employing in order to build solid RESTful iPhone clients in respect to model hierarchies. What I mean by model hierarchies is that I have a REST server in which several resources have a correlation to each other. For instance, let's say for hypothetical purposes I have a REST server which spits out information about a school. If I want to grab all the students in a particular class, I first need to query the REST server for information on the school, then I need to query the server for information on all of the classes the school has to offer, followed by a subsequent request for all students in a particular class. At the end of the day, the client is bringing in 3 unique XML trees. At that point, do most of you folks write your own algorithms in order to build the final tree which will end up being your data source? Do you not aggregate XML trees in this respect and instead use a different approach?

How do you prefer taking a myriad of related resources on the server and getting them into one tree that just makes sense onto the client?

Thanks for the insight.

回答1:

How you choose to store manage your model data on the iPhone probably depends on how much the XML data you are dealing with is likely to vary, and how bloated it might get.

If this is a simple XML model that is unlikely to change, and doesn't carry terribly much redundant information with it, you might do well to just use the XML trees you describe as-is.

But for anything even slightly more complex, I prefer to translate XML representations into a format that is most easily manipulated by my Objective-C code. Consider for example the possibility that your application expands one day to support other web-based services that provide similar data, but as JSON or SOAP formatted data. Now you start to run into the headaches of maintaining 3 different types of models in your application, when it would be preferable to maintain just one.

I would treat the XML-based REST resources as "foreign data" that need to be massaged into a locally manipulatable format. If you adopt a local format that maximizes the ease of performing your app-specific operations, then you can adapt inputs from any other foreign format, and convert back to foreign formats as needed for uploads/edits/whatever.

Daniel

回答2:

There is a project called IPhone on Rails, where they created an Objective-C implementation of ActiveResource. I have not yet had a chance to play with it, but it may be worth checking out.

回答3:

I'm currently working on a project to solve this problem, we will be releasing the code when we get it to a stable point.

It's a very tricky problem if you don't have a standardized system for serializing all your objects coming back from the server. Luckily, on this project we do have a common serialization mechanism, so we built an abstract SAX parser which we inherit from and then use type introspection and NSCoding to bridge the gap between Objective-C and XML. You can do a similar concept on your own.

Before this project, I just built/reused a couple of classes to handle the NSURLConnection and NSXMLParser work but instead of type introspection, I used the Delegation Pattern.

回答4:

In my own case, I have been using wrappers around ASIHTTPRequest to build RESTful queries. I have been handling the results with libxml2, parsed directly into model objects rendered into the user interface.

I'm not sure I have a catchall solution for you, but I think you will gain a lot from separating as functionality into different, reusable methods. This way, you won't have to worry as much about the number of XML trees coming back. You just issue queries to whatever you have, so as to populate your interface.