When MVC runs an ActionMethod
it will populate the ModelState
dictionary and uses a ModelBinder
to build the ActionMethod
parameter(s) if any. It does this for both GET
and POST
. Which makes sense.
After the ActionMethod
successfully has been ran, the view is rendered using the razor supplied, which in my case uses as much HtmlHelper
calls as possible. Until now, you might think, 'Yes I know how MVC works'. Hold on, I'm getting there.
When we use e.g. @Html.TextFor(m => m.Name)
MVC uses the code that can be here to render the tag.
The interesting part is at the end of the file where we find:
switch (inputType)
{
case InputType.CheckBox:
// ... removed for brevity
case InputType.Radio:
// ... removed for brevity
case InputType.Password:
// ... removed for brevity
default:
string attemptedValue =
(string)htmlHelper.GetModelStateValue(fullName, typeof(string));
tagBuilder.MergeAttribute("value", attemptedValue ??
((useViewData)
? htmlHelper.EvalString(fullName, format)
: valueParameter), isExplicitValue);
break;
}
What this means is that the Modelstate
is used to get a value over a supplied Model
. And this makes sense for a POST
because when there are validation errors you want MVC to render the view with the values already supplied by the user. The documentation also states this behavior and there are several posts here on StackOverflow confirming this. But it is stated only for POST
, as can be seen on this thread for example
However this code is also used when a view is rendered for a GET
! And this makes no sense at all to me.
Consider the following action method:
public ActionResult GetCopy(Guid id)
{
var originalModel = this.repository.GetById(id);
var copiedModel = new SomeModel
{
Id = Guid.NewGuid(),
Name = originalModel.Name,
};
return this.View(copiedModel);
}
with this simple razor markup:
@Html.TextBoxFor(m => m.Id)
@Html.TextBoxFor(m => m.Name)
I would expect the view to show the newly generated GUID
instead of the Id
of the original model. However the view shows the Id
passed to the actionmethod on the GET
request because the ModelState
dictionary contains a key with the name Id
.
My question is not on how to solve this, because that is fairly easy:
- rename the
ActionMethod
parameter to something different - Clear the
ModelState
before returning theActionResult
usingModelstate.Clear()
- Replace the value in the
ModelState
dictionary and don't supply aModel
to the view at all.
My question is:
Why is this proces the same for both GET
and POST
. Is there any valid use case for using the populated ModelState
over the supplied Model
on a GET
.
The DefaultModelBinder
does not make a distinction between GET and POST requests. While only the MVC team can confirm why they made that decision, there are valid use cases why ModelState
should be used in a GET (in the same way its used in a POST) since you can also submit a form to a GET method.
Take for example an airline booking app, where you have a model with properties to select departure and arrival locations, a date, and the number of passengers (an int
property), plus a collection property for the filtered results. The GET method signature might look like
public ActionResult Search(SearchViewModel model)
The view includes a that makes a GET back to the method and includes a textbox for the number of passengers.
A (not too bright) user who has disabled javascript enters "TWO" in the textbox and submits the form. Your method returns the view without populating the collection of available flights because ModelState
is invalid (the value "TWO" is not valid for an int
) and now the user sees an error message stating The field Passengers must be a number.
If the model value was used rather that the ModelState
value, then the associated textbox would be displaying 0
instead of "Two"
, making the error message confusing (but 0
is a number! - and what happened to what I entered?)
来源:https://stackoverflow.com/questions/48204441/why-does-mvc-use-the-modelstate-over-a-supplied-model-on-a-get