How Does MVC Handle Missing Data Requirements

Question

I'm teaching myself MVC concepts in hopes of applying them to a non-OO/procedural development environment. I am pretty sure I understand simple View -> Request -> Controller -> Request -> Model -> Response -> Controller -> Response -> View flow.

What I am struggling with is understanding more complex scenarios. For instance, let's say I have a shopping cart form with a button for 'Calculate Shipping'. Normally a click on this button will follow the above flow. But what if there is missing data, like the zip code? Should the View verify this first and alert the user before making a 'Calculate Shipping' request? Or should the request be made and the Model returns a notification that critical data is missing? If the latter, does the Controller instruct the View to alert the user? What if I wanted to prompt the user for the missing zip code (perhaps in a popup input display) and then automatically request the 'Calculate Shipping' method again?

I suppose this gets into the question of how smart a View ought to be. It seems that MVC has evolved due to richer UI and automation (such as with data-binding) and this muddies the water from a purist MVC perspective. Any thoughts are greatly appreciated.

Answer 1

There are a fiew things you can do:

1. Validate the zip code on the client side (in the browser) using Javascript or jQuery
2. Send an Ajax request to the server to validate the zip code, and have the controller return a validation result to the browser, or
3. Post the whole page, validate it, and return the same page with an error indication to the user, if the zip code is missing.

A reminder: you will still have to validate the data again on the server, once the page is POSTed. Never trust data coming from a web page, even if it has already been validated in the client.

Answer 2

My own thought are that, for efficiency, as much as possible should be done in the view, as long as there are no data implications.

In other words, the view can check for missing zipcode or non-numeric zipcode (as long as you're certain the data model won't ever allow non-numerics).

But, the second you decide to check data (such as "Is the zipcode a valid one for the continental US?"), that should be referred to the data model itself.

I've seen situations where the view is created with a series of regular expressions for pre-validation (one per field) so as to not unnecessarily burden the other layers. This means that the data entered passes at least the first layer of sanity checks before being passed back for more complete validation.

And, as an aside, don't ever use the term "sanity check" in user documentation when referring to checking what the user has entered. You might think it's an innocent phrase but user don't always think so :-)