MVC Razor view nested foreach's model

David C picture David C · Jan 17, 2012 · Viewed 103.2k times · Source

Imagine a common scenario, this is a simpler version of what I'm coming across. I actually have a couple of layers of further nesting on mine....

But this is the scenario

Theme contains List Category contains List Product contains List

My Controller provides a fully populated Theme, with all the Categories for that theme, the Products within this categories and the their orders.

The orders collection has a property called Quantity (amongst many others) that needs to be editable.

@model ViewModels.MyViewModels.Theme

@Html.LabelFor(Model.Theme.name)
@foreach (var category in Model.Theme)
{
   @Html.LabelFor(category.name)
   @foreach(var product in theme.Products)
   {
      @Html.LabelFor(product.name)
      @foreach(var order in product.Orders)
      {
          @Html.TextBoxFor(order.Quantity)
          @Html.TextAreaFor(order.Note)
          @Html.EditorFor(order.DateRequestedDeliveryFor)
      }
   }
}

If I use lambda instead then then I only seem to get a reference to the top Model object, "Theme" not those within the foreach loop.

Is what I'm trying to do there even possible or have I overestimated or misunderstood what is possible?

With the above I get an error on the TextboxFor, EditorFor, etc

CS0411: The type arguments for method 'System.Web.Mvc.Html.InputExtensions.TextBoxFor(System.Web.Mvc.HtmlHelper, System.Linq.Expressions.Expression>)' cannot be inferred from the usage. Try specifying the type arguments explicitly.

Thanks.

Answer

J. Holmes picture J. Holmes · Jan 17, 2012

The quick answer is to use a for() loop in place of your foreach() loops. Something like:

@for(var themeIndex = 0; themeIndex < Model.Theme.Count(); themeIndex++)
{
   @Html.LabelFor(model => model.Theme[themeIndex])

   @for(var productIndex=0; productIndex < Model.Theme[themeIndex].Products.Count(); productIndex++)
   {
      @Html.LabelFor(model=>model.Theme[themeIndex].Products[productIndex].name)
      @for(var orderIndex=0; orderIndex < Model.Theme[themeIndex].Products[productIndex].Orders; orderIndex++)
      {
          @Html.TextBoxFor(model => model.Theme[themeIndex].Products[productIndex].Orders[orderIndex].Quantity)
          @Html.TextAreaFor(model => model.Theme[themeIndex].Products[productIndex].Orders[orderIndex].Note)
          @Html.EditorFor(model => model.Theme[themeIndex].Products[productIndex].Orders[orderIndex].DateRequestedDeliveryFor)
      }
   }
}

But this glosses over why this fixes the problem.

There are three things that you have at least a cursory understanding before you can resolve this issue. I have to admit that I cargo-culted this for a long time when I started working with the framework. And it took me quite a while to really get what was going on.

Those three things are:

  • How do the LabelFor and other ...For helpers work in MVC?
  • What is an Expression Tree?
  • How does the Model Binder work?

All three of these concepts link together to get an answer.

How do the LabelFor and other ...For helpers work in MVC?

So, you've used the HtmlHelper<T> extensions for LabelFor and TextBoxFor and others, and you probably noticed that when you invoke them, you pass them a lambda and it magically generates some html. But how?

So the first thing to notice is the signature for these helpers. Lets look at the simplest overload for TextBoxFor

public static MvcHtmlString TextBoxFor<TModel, TProperty>(
    this HtmlHelper<TModel> htmlHelper,
    Expression<Func<TModel, TProperty>> expression
) 

First, this is an extension method for a strongly typed HtmlHelper, of type <TModel>. So, to simply state what happens behind the scenes, when razor renders this view it generates a class. Inside of this class is an instance of HtmlHelper<TModel> (as the property Html, which is why you can use @Html...), where TModel is the type defined in your @model statement. So in your case, when you are looking at this view TModel will always be of the type ViewModels.MyViewModels.Theme.

Now, the next argument is a bit tricky. So lets look at an invocation

@Html.TextBoxFor(model=>model.SomeProperty);

It looks like we have a little lambda, And if one were to guess the signature, one might think that the type for this argument would simply be a Func<TModel, TProperty>, where TModel is the type of the view model and TProperty is inferred as the type of the property.

But thats not quite right, if you look at the actual type of the argument its Expression<Func<TModel, TProperty>>.

So when you normally generate a lambda, the compiler takes the lambda and compiles it down into MSIL, just like any other function (which is why you can use delegates, method groups, and lambdas more or less interchangeably, because they are just code references.)

However, when the compiler sees that the type is an Expression<>, it doesn't immediately compile the lambda down to MSIL, instead it generates an Expression Tree!

What is an Expression Tree?

So, what the heck is an expression tree. Well, it's not complicated but its not a walk in the park either. To quote ms:

| Expression trees represent code in a tree-like data structure, where each node is an expression, for example, a method call or a binary operation such as x < y.

Simply put, an expression tree is a representation of a function as a collection of "actions".

In the case of model=>model.SomeProperty, the expression tree would have a node in it that says: "Get 'Some Property' from a 'model'"

This expression tree can be compiled into a function that can be invoked, but as long as it's an expression tree, it's just a collection of nodes.

So what is that good for?

So Func<> or Action<>, once you have them, they are pretty much atomic. All you can really do is Invoke() them, aka tell them to do the work they are supposed to do.

Expression<Func<>> on the other hand, represents a collection of actions, which can be appended, manipulated, visited, or compiled and invoked.

So why are you telling me all this?

So with that understanding of what an Expression<> is, we can go back to Html.TextBoxFor. When it renders a textbox, it needs to generate a few things about the property that you are giving it. Things like attributes on the property for validation, and specifically in this case it needs to figure out what to name the <input> tag.

It does this by "walking" the expression tree and building a name. So for an expression like model=>model.SomeProperty, it walks the expression gathering the properties that you are asking for and builds <input name='SomeProperty'>.

For a more complicated example, like model=>model.Foo.Bar.Baz.FooBar, it might generate <input name="Foo.Bar.Baz.FooBar" value="[whatever FooBar is]" />

Make sense? It is not just the work that the Func<> does, but how it does its work is important here.

(Note other frameworks like LINQ to SQL do similar things by walking an expression tree and building a different grammar, that this case a SQL query)

How does the Model Binder work?

So once you get that, we have to briefly talk about the model binder. When the form gets posted, it's simply like a flat Dictionary<string, string>, we have lost the hierarchical structure our nested view model may have had. It's the model binder's job to take this key-value pair combo and attempt to rehydrate an object with some properties. How does it do this? You guessed it, by using the "key" or name of the input that got posted.

So if the form post looks like

Foo.Bar.Baz.FooBar = Hello

And you are posting to a model called SomeViewModel, then it does the reverse of what the helper did in the first place. It looks for a property called "Foo". Then it looks for a property called "Bar" off of "Foo", then it looks for "Baz"... and so on...

Finally it tries to parse the value into the type of "FooBar" and assign it to "FooBar".

PHEW!!!

And voila, you have your model. The instance the Model Binder just constructed gets handed into requested Action.


So your solution doesn't work because the Html.[Type]For() helpers need an expression. And you are just giving them a value. It has no idea what the context is for that value, and it doesn't know what to do with it.

Now some people suggested using partials to render. Now this in theory will work, but probably not the way that you expect. When you render a partial, you are changing the type of TModel, because you are in a different view context. This means that you can describe your property with a shorter expression. It also means when the helper generates the name for your expression, it will be shallow. It will only generate based on the expression it's given (not the entire context).

So lets say you had a partial that just rendered "Baz" (from our example before). Inside that partial you could just say:

@Html.TextBoxFor(model=>model.FooBar)

Rather than

@Html.TextBoxFor(model=>model.Foo.Bar.Baz.FooBar)

That means that it will generate an input tag like this:

<input name="FooBar" />

Which, if you are posting this form to an action that is expecting a large deeply nested ViewModel, then it will try to hydrate a property called FooBar off of TModel. Which at best isn't there, and at worst is something else entirely. If you were posting to a specific action that was accepting a Baz, rather than the root model, then this would work great! In fact, partials are a good way to change your view context, for example if you had a page with multiple forms that all post to different actions, then rendering a partial for each one would be a great idea.


Now once you get all of this, you can start to do really interesting things with Expression<>, by programatically extending them and doing other neat things with them. I won't get into any of that. But, hopefully, this will give you a better understanding of what is going on behind the scenes and why things are acting the way that they are.