Search Results

Search found 55736 results on 2230 pages for 'asp net mvc'.

Page 47/2230 | < Previous Page | 43 44 45 46 47 48 49 50 51 52 53 54  | Next Page >

  • How do I maintain scroll position in MVC?

    - by Eric Brown
    Im working on a project in MVC and have enjoyed learning about it. There are a few growing pains but once you figure them out it's not bad. One thing that is really simple in the WebForms world is maintaining the scroll position on a page. All you do is set the MaintainScrollPositionOnPostback property to true. However, in MVC, Im not using postbacks so this will not work for me. What is the standard way of handling this? Edit: Ajax is acceptable, but I was also wondering how you would do it without AJAX.

    Read the article

  • Removing Database References from an ASP.NET MVC Site

    - by Maxim Z.
    I'm building a very simple ASP.NET MVC site that doesn't require a database and doesn't have user accounts or any other special ASP.NET stuff. When I first created the site with the ASP.NET MVC 1.0 site template in Visual Studio, the web.config page was configured to automatically connect to SQL Server Express as a User Instance. Can I completely remove this dependency? If so, how do I get rid of all this database stuff?

    Read the article

  • ASP.NET MVC: post-redirect-get pattern, with only two overloaded action methods

    - by Rafi
    Is it possible to implement post-redirect-get pattern, with two overloaded action methods(One for GET action and the other for POST action) in ASP.NET MVC. In all of the MVC post-redirect-get pattern samples, I have seen three different action methods for the post-redirect-get process, each having different names. Is this really required? For Eg:(Does the code shown below, follows Post-Redirect-Get pattern?) public class SalaryTransferController : Controller { // // GET: /SalaryTransfer/ [HttpGet] public ActionResult Index(int id) { SalaryTransferIndexViewModel vm = new SalaryTransferIndexViewModel(id) { SelectedDivision = DivisionEnum.Contracting }; //Do some processing here return View(vm); } // // POST: /SalaryTransfer/ [HttpPost] public ActionResult Index(SalaryTransferIndexViewModel vm) { bool validationsuccess = false; //validate if (validationsuccess) return RedirectToAction("Index", new {id=1234 }); else return View(vm); } } Thank you for your responses.

    Read the article

  • What is the order of execution when dealing with .NET MVC 2 Action Filters?

    - by user357933
    Say I have: [Attribute1(Order=0)] public class Controller1 { [Attribute2] [Attribute3] public ActionResult Action1() { ... } } The attributes get executed in the following order: 2, 3, 1 This makes sense because attributes 2 and 3 have an order of -1 and will be executed before attribute 1 which has an explicitly set order equal to 0. Now, lets say I have: [Attribute1] [Attribute2(Order=0)] public class Controller1 { [Attribute3] public ActionResult Action1() { ... } } The attributes get executed in the following order: 1, 2, 3 Why is it that attribute 2 in this case (which has an order equal to 0) is executed before attribute 3 (which has an order equal to -1)?

    Read the article

  • View artifacts leaking into the model of MVC

    - by Jono
    In an ASP.NET MVC application (which has very little chance of having its view technology ported to something non-HTML, but whose functional requirements evolve weekly,) how much HTML should ideally be allowed to be directly represented in the Model? I might come across as a design bigot for this, but I regard it as bad practice to allow any view constructs to "leak" into the model in an MVC application (and vice versa). For example, a Model that represents an item you're about to purchase should know nothing about the HTML check box that says "add giftwrap/message", nor should it know about any HTML drop down lists for payment card types. Conversely the View shouldn't be doing work like figuring out button text by translating keys into values (by looking in resource files.)

    Read the article

  • asp.mvc model design

    - by Radu D
    Hi, I am pretty new to MVC and I am looking for a way to design my models. I have the MVC web site project and another class library that takes care of data access and constructing the business objects. If I have in that assembly a class named Project that is a business object and I need to display all projects in a view ... should I make another model class Project? In this case the classes will be identical. Do I gain something from doing a new model class? I don't like having in views references to objects from another dll ... but i don't like duplicating the code neither. Did you encounter the same problem?

    Read the article

  • why developing ASP.NET - MVC?

    - by sam
    Hi Guys, I am new to web development, I am coding some ASP.NET, I checked a lot of examples using MVC in ASP.NET, But I am looking for verbal answers from senior programmers, about why using MVC? can U as seniors and team leaders show me the benefits?? and why not keeping using asp.net webforms? thanks

    Read the article

  • MVC 2 Beta DefaultControllerFactory with Areas

    - by stoto
    Why default factory WON'T return full name of the controllers (with namespaces)? I'm using Service Locator and autofac. using System.Web.Mvc; using Microsoft.Practices.ServiceLocation; namespace Application.Core.MVC { public override IController CreateController(System.Web.Routing.RequestContext requestContext, string **controllerName**) { return ServiceLocator.Current.GetInstance<IController>(controllerName); } } I had two home controllers (one under area Blog) http://localhost/Home http://localhost/Blog/Home controllerName return only "Home" without full qualified name for both in above code. This creates a problem when I try to regiser controllers' names for dependency injection. Here is how I register controllers right now according to this situation. Even this brings up the pages without exception. But When I access http://localhost/Home, both controllers invoked regardlessly. foreach (var tp in currentAssemblyControllersTypes) builder.Register(tp).FactoryScoped().Named(tp.Name.Replace("Controller", "")); Anyone can help?Thanks.

    Read the article

  • Server.Execute - render .ASP from MVC controller action

    - by David Lively
    I need to render an ASP page to a string from an MVC controller action. I can use Server.Execute() to render a .aspx page, but not a .asp page. Here's what I'm using: public ActionResult Index() { Server.Execute("/default.asp"); return new EmptyResult(); } which returns `No http handler was found for request type 'GET'` Any suggestions? I can do something similar with with a web request, but I'd rather avoid the overhead of a loopback request.

    Read the article

  • Who can change the View in MVC?

    - by Luke
    I'm working on a thick client graph displaying and manipulation application. I'm trying to apply the MVC pattern to our 3D visualization component. Here is what I have for the Model, View, and Controller: Model - The graph and it's metadata. This includes vertices, edges, and the attributes of each. It does not contain position information, icons, colors, or anything display related. View - This would commonly be called a scene graph. It includes the 3D display information, texture information, color information, and anything else that is related specifically to the visualization of the model. Controller - The controller takes the view and displays it in a Window using OpenGL (but it could potentially be any 3D graphics package). The application has various "layouts" that change the position of the vertices in the display. For instance, one layout may arrange the vertices in a circle. Is it common for these layouts to access and change the view directly? Should they go through the Controller to access the View? If they go through the Controller, should they just ask for direct access to the View or should each change go through the controller? I realize this is a bit different from the standard MVC example where there a finite number of Views. In this case, the View can change in an infinite number of ways. Perhaps I'm shattering some basic principle of MVC here. Thanks in advance!

    Read the article

  • Question about modeling with MVC (the pattern, not the MS stuff / non web)

    - by paul
    I'm working on an application in which I'm looking to employ the MVC pattern, but I've come up against a design decision point I could use some help with. My application is going to deal with the design of state-machines. Currently the MVC model holds information about the machine's states, inputs, outputs, etc. The view is going to show a diagram for the machine, graphically allowing the user to add new states, establish transitions, and put the states in a pleasing arrangement, among other things. I would like to store part of the diagram's state (e.g. the x and y state positions) when the machine information is stored for later retrieval, and am wondering how best to go about structuring the model(s?) for this. It seems like this UI information is more closely related to the view than to the state-machine model, so I was thinking that a secondary model might be in order, but I am reluctant to pursue this route because of the added complexity. Adding this information to the current model doesn't seem the right way to go about it either. This is the my first time using the MVC pattern so I'm still figuring things out. Any input would be appreciated.

    Read the article

  • when to clear or make null asp .net mvc models?

    - by SARAVAN
    HI, I am working in an asp .net mvc application. I am using the model and storing some of the values which i need to preserve between the page posts, in the form of datacontexts. Say my model looks something like this: public SelectedUser SelectedUserDetails { //get and set has //this.datacontext.data.SelectedUser = ..... //return this.datacontext.data..... } Now when this model needs to be cleared? I have many such models with many properties and datacontext. But I don't have an idea on when to clear it. Is there a way or an event that can be triggered automatically when the model is not used for a long time? Oneway I thought is when i navigate away from a page which uses my underlying model, I can clear that model if its no longer used anywhere and initialise it back as needed. But I need to clear almost many models at many points. Is there an automatic way that can clear models when it is no longer used beacuse care can be taken by my code to initialise them when I need them, but I don't know when to clear them when I no longer need them. I need this to get rid of any memory related issues. Any thoughts or comments?

    Read the article

  • Integrating JavaScript Unit Tests with Visual Studio

    - by Stephen Walther
    Modern ASP.NET web applications take full advantage of client-side JavaScript to provide better interactivity and responsiveness. If you are building an ASP.NET application in the right way, you quickly end up with lots and lots of JavaScript code. When writing server code, you should be writing unit tests. One big advantage of unit tests is that they provide you with a safety net that enable you to safely modify your existing code – for example, fix bugs, add new features, and make performance enhancements -- without breaking your existing code. Every time you modify your code, you can execute your unit tests to verify that you have not broken anything. For the same reason that you should write unit tests for your server code, you should write unit tests for your client code. JavaScript is just as susceptible to bugs as C#. There is no shortage of unit testing frameworks for JavaScript. Each of the major JavaScript libraries has its own unit testing framework. For example, jQuery has QUnit, Prototype has UnitTestJS, YUI has YUI Test, and Dojo has Dojo Objective Harness (DOH). The challenge is integrating a JavaScript unit testing framework with Visual Studio. Visual Studio and Visual Studio ALM provide fantastic support for server-side unit tests. You can easily view the results of running your unit tests in the Visual Studio Test Results window. You can set up a check-in policy which requires that all unit tests pass before your source code can be committed to the source code repository. In addition, you can set up Team Build to execute your unit tests automatically. Unfortunately, Visual Studio does not provide “out-of-the-box” support for JavaScript unit tests. MS Test, the unit testing framework included in Visual Studio, does not support JavaScript unit tests. As soon as you leave the server world, you are left on your own. The goal of this blog entry is to describe one approach to integrating JavaScript unit tests with MS Test so that you can execute your JavaScript unit tests side-by-side with your C# unit tests. The goal is to enable you to execute JavaScript unit tests in exactly the same way as server-side unit tests. You can download the source code described by this project by scrolling to the end of this blog entry. Rejected Approach: Browser Launchers One popular approach to executing JavaScript unit tests is to use a browser as a test-driver. When you use a browser as a test-driver, you open up a browser window to execute and view the results of executing your JavaScript unit tests. For example, QUnit – the unit testing framework for jQuery – takes this approach. The following HTML page illustrates how you can use QUnit to create a unit test for a function named addNumbers(). <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN" "http://www.w3.org/TR/html4/loose.dtd"> <html> <head> <title>Using QUnit</title> <link rel="stylesheet" href="http://github.com/jquery/qunit/raw/master/qunit/qunit.css" type="text/css" /> </head> <body> <h1 id="qunit-header">QUnit example</h1> <h2 id="qunit-banner"></h2> <div id="qunit-testrunner-toolbar"></div> <h2 id="qunit-userAgent"></h2> <ol id="qunit-tests"></ol> <div id="qunit-fixture">test markup, will be hidden</div> <script type="text/javascript" src="http://code.jquery.com/jquery-latest.js"></script> <script type="text/javascript" src="http://github.com/jquery/qunit/raw/master/qunit/qunit.js"></script> <script type="text/javascript"> // The function to test function addNumbers(a, b) { return a+b; } // The unit test test("Test of addNumbers", function () { equals(4, addNumbers(1,3), "1+3 should be 4"); }); </script> </body> </html> This test verifies that calling addNumbers(1,3) returns the expected value 4. When you open this page in a browser, you can see that this test does, in fact, pass. The idea is that you can quickly refresh this QUnit HTML JavaScript test driver page in your browser whenever you modify your JavaScript code. In other words, you can keep a browser window open and keep refreshing it over and over while you are developing your application. That way, you can know very quickly whenever you have broken your JavaScript code. While easy to setup, there are several big disadvantages to this approach to executing JavaScript unit tests: You must view your JavaScript unit test results in a different location than your server unit test results. The JavaScript unit test results appear in the browser and the server unit test results appear in the Visual Studio Test Results window. Because all of your unit test results don’t appear in a single location, you are more likely to introduce bugs into your code without noticing it. Because your unit tests are not integrated with Visual Studio – in particular, MS Test -- you cannot easily include your JavaScript unit tests when setting up check-in policies or when performing automated builds with Team Build. A more sophisticated approach to using a browser as a test-driver is to automate the web browser. Instead of launching the browser and loading the test code yourself, you use a framework to automate this process. There are several different testing frameworks that support this approach: · Selenium – Selenium is a very powerful framework for automating browser tests. You can create your tests by recording a Firefox session or by writing the test driver code in server code such as C#. You can learn more about Selenium at http://seleniumhq.org/. LTAF – The ASP.NET team uses the Lightweight Test Automation Framework to test JavaScript code in the ASP.NET framework. You can learn more about LTAF by visiting the project home at CodePlex: http://aspnet.codeplex.com/releases/view/35501 jsTestDriver – This framework uses Java to automate the browser. jsTestDriver creates a server which can be used to automate multiple browsers simultaneously. This project is located at http://code.google.com/p/js-test-driver/ TestSwam – This framework, created by John Resig, uses PHP to automate the browser. Like jsTestDriver, the framework creates a test server. You can open multiple browsers that are automated by the test server. Learn more about TestSwarm by visiting the following address: https://github.com/jeresig/testswarm/wiki Yeti – This is the framework introduced by Yahoo for automating browser tests. Yeti uses server-side JavaScript and depends on Node.js. Learn more about Yeti at http://www.yuiblog.com/blog/2010/08/25/introducing-yeti-the-yui-easy-testing-interface/ All of these frameworks are great for integration tests – however, they are not the best frameworks to use for unit tests. In one way or another, all of these frameworks depend on executing tests within the context of a “living and breathing” browser. If you create an ASP.NET Unit Test then Visual Studio will launch a web server before executing the unit test. Why is launching a web server so bad? It is not the worst thing in the world. However, it does introduce dependencies that prevent your code from being tested in isolation. One of the defining features of a unit test -- versus an integration test – is that a unit test tests code in isolation. Another problem with launching a web server when performing unit tests is that launching a web server can be slow. If you cannot execute your unit tests quickly, you are less likely to execute your unit tests each and every time you make a code change. You are much more likely to fall into the pit of failure. Launching a browser when performing a JavaScript unit test has all of the same disadvantages as launching a web server when performing an ASP.NET unit test. Instead of testing a unit of JavaScript code in isolation, you are testing JavaScript code within the context of a particular browser. Using the frameworks listed above for integration tests makes perfect sense. However, I want to consider a different approach for creating unit tests for JavaScript code. Using Server-Side JavaScript for JavaScript Unit Tests A completely different approach to executing JavaScript unit tests is to perform the tests outside of any browser. If you really want to test JavaScript then you should test JavaScript and leave the browser out of the testing process. There are several ways that you can execute JavaScript on the server outside the context of any browser: Rhino – Rhino is an implementation of JavaScript written in Java. The Rhino project is maintained by the Mozilla project. Learn more about Rhino at http://www.mozilla.org/rhino/ V8 – V8 is the open-source Google JavaScript engine written in C++. This is the JavaScript engine used by the Chrome web browser. You can download V8 and embed it in your project by visiting http://code.google.com/p/v8/ JScript – JScript is the JavaScript Script Engine used by Internet Explorer (up to but not including Internet Explorer 9), Windows Script Host, and Active Server Pages. Internet Explorer is still the most popular web browser. Therefore, I decided to focus on using the JScript Script Engine to execute JavaScript unit tests. Using the Microsoft Script Control There are two basic ways that you can pass JavaScript to the JScript Script Engine and execute the code: use the Microsoft Windows Script Interfaces or use the Microsoft Script Control. The difficult and proper way to execute JavaScript using the JScript Script Engine is to use the Microsoft Windows Script Interfaces. You can learn more about the Script Interfaces by visiting http://msdn.microsoft.com/en-us/library/t9d4xf28(VS.85).aspx The main disadvantage of using the Script Interfaces is that they are difficult to use from .NET. There is a great series of articles on using the Script Interfaces from C# located at http://www.drdobbs.com/184406028. I picked the easier alternative and used the Microsoft Script Control. The Microsoft Script Control is an ActiveX control that provides a higher level abstraction over the Window Script Interfaces. You can download the Microsoft Script Control from here: http://www.microsoft.com/downloads/en/details.aspx?FamilyID=d7e31492-2595-49e6-8c02-1426fec693ac After you download the Microsoft Script Control, you need to add a reference to it to your project. Select the Visual Studio menu option Project, Add Reference to open the Add Reference dialog. Select the COM tab and add the Microsoft Script Control 1.0. Using the Script Control is easy. You call the Script Control AddCode() method to add JavaScript code to the Script Engine. Next, you call the Script Control Run() method to run a particular JavaScript function. The reference documentation for the Microsoft Script Control is located at the MSDN website: http://msdn.microsoft.com/en-us/library/aa227633%28v=vs.60%29.aspx Creating the JavaScript Code to Test To keep things simple, let’s imagine that you want to test the following JavaScript function named addNumbers() which simply adds two numbers together: MvcApplication1\Scripts\Math.js function addNumbers(a, b) { return 5; } Notice that the addNumbers() method always returns the value 5. Right-now, it will not pass a good unit test. Create this file and save it in your project with the name Math.js in your MVC project’s Scripts folder (Save the file in your actual MVC application and not your MVC test application). Creating the JavaScript Test Helper Class To make it easier to use the Microsoft Script Control in unit tests, we can create a helper class. This class contains two methods: LoadFile() – Loads a JavaScript file. Use this method to load the JavaScript file being tested or the JavaScript file containing the unit tests. ExecuteTest() – Executes the JavaScript code. Use this method to execute a JavaScript unit test. Here’s the code for the JavaScriptTestHelper class: JavaScriptTestHelper.cs   using System; using System.IO; using Microsoft.VisualStudio.TestTools.UnitTesting; using MSScriptControl; namespace MvcApplication1.Tests { public class JavaScriptTestHelper : IDisposable { private ScriptControl _sc; private TestContext _context; /// <summary> /// You need to use this helper with Unit Tests and not /// Basic Unit Tests because you need a Test Context /// </summary> /// <param name="testContext">Unit Test Test Context</param> public JavaScriptTestHelper(TestContext testContext) { if (testContext == null) { throw new ArgumentNullException("TestContext"); } _context = testContext; _sc = new ScriptControl(); _sc.Language = "JScript"; _sc.AllowUI = false; } /// <summary> /// Load the contents of a JavaScript file into the /// Script Engine. /// </summary> /// <param name="path">Path to JavaScript file</param> public void LoadFile(string path) { var fileContents = File.ReadAllText(path); _sc.AddCode(fileContents); } /// <summary> /// Pass the path of the test that you want to execute. /// </summary> /// <param name="testMethodName">JavaScript function name</param> public void ExecuteTest(string testMethodName) { dynamic result = null; try { result = _sc.Run(testMethodName, new object[] { }); } catch { var error = ((IScriptControl)_sc).Error; if (error != null) { var description = error.Description; var line = error.Line; var column = error.Column; var text = error.Text; var source = error.Source; if (_context != null) { var details = String.Format("{0} \r\nLine: {1} Column: {2}", source, line, column); _context.WriteLine(details); } } throw new AssertFailedException(error.Description); } } public void Dispose() { _sc = null; } } }     Notice that the JavaScriptTestHelper class requires a Test Context to be instantiated. For this reason, you can use the JavaScriptTestHelper only with a Visual Studio Unit Test and not a Basic Unit Test (These are two different types of Visual Studio project items). Add the JavaScriptTestHelper file to your MVC test application (for example, MvcApplication1.Tests). Creating the JavaScript Unit Test Next, we need to create the JavaScript unit test function that we will use to test the addNumbers() function. Create a folder in your MVC test project named JavaScriptTests and add the following JavaScript file to this folder: MvcApplication1.Tests\JavaScriptTests\MathTest.js /// <reference path="JavaScriptUnitTestFramework.js"/> function testAddNumbers() { // Act var result = addNumbers(1, 3); // Assert assert.areEqual(4, result, "addNumbers did not return right value!"); }   The testAddNumbers() function takes advantage of another JavaScript library named JavaScriptUnitTestFramework.js. This library contains all of the code necessary to make assertions. Add the following JavaScriptnitTestFramework.js to the same folder as the MathTest.js file: MvcApplication1.Tests\JavaScriptTests\JavaScriptUnitTestFramework.js var assert = { areEqual: function (expected, actual, message) { if (expected !== actual) { throw new Error("Expected value " + expected + " is not equal to " + actual + ". " + message); } } }; There is only one type of assertion supported by this file: the areEqual() assertion. Most likely, you would want to add additional types of assertions to this file to make it easier to write your JavaScript unit tests. Deploying the JavaScript Test Files This step is non-intuitive. When you use Visual Studio to run unit tests, Visual Studio creates a new folder and executes a copy of the files in your project. After you run your unit tests, your Visual Studio Solution will contain a new folder named TestResults that includes a subfolder for each test run. You need to configure Visual Studio to deploy your JavaScript files to the test run folder or Visual Studio won’t be able to find your JavaScript files when you execute your unit tests. You will get an error that looks something like this when you attempt to execute your unit tests: You can configure Visual Studio to deploy your JavaScript files by adding a Test Settings file to your Visual Studio Solution. It is important to understand that you need to add this file to your Visual Studio Solution and not a particular Visual Studio project. Right-click your Solution in the Solution Explorer window and select the menu option Add, New Item. Select the Test Settings item and click the Add button. After you create a Test Settings file for your solution, you can indicate that you want a particular folder to be deployed whenever you perform a test run. Select the menu option Test, Edit Test Settings to edit your test configuration file. Select the Deployment tab and select your MVC test project’s JavaScriptTest folder to deploy. Click the Apply button and the Close button to save the changes and close the dialog. Creating the Visual Studio Unit Test The very last step is to create the Visual Studio unit test (the MS Test unit test). Add a new unit test to your MVC test project by selecting the menu option Add New Item and selecting the Unit Test project item (Do not select the Basic Unit Test project item): The difference between a Basic Unit Test and a Unit Test is that a Unit Test includes a Test Context. We need this Test Context to use the JavaScriptTestHelper class that we created earlier. Enter the following test method for the new unit test: [TestMethod] public void TestAddNumbers() { var jsHelper = new JavaScriptTestHelper(this.TestContext); // Load JavaScript files jsHelper.LoadFile("JavaScriptUnitTestFramework.js"); jsHelper.LoadFile(@"..\..\..\MvcApplication1\Scripts\Math.js"); jsHelper.LoadFile("MathTest.js"); // Execute JavaScript Test jsHelper.ExecuteTest("testAddNumbers"); } This code uses the JavaScriptTestHelper to load three files: JavaScripUnitTestFramework.js – Contains the assert functions. Math.js – Contains the addNumbers() function from your MVC application which is being tested. MathTest.js – Contains the JavaScript unit test function. Next, the test method calls the JavaScriptTestHelper ExecuteTest() method to execute the testAddNumbers() JavaScript function. Running the Visual Studio JavaScript Unit Test After you complete all of the steps described above, you can execute the JavaScript unit test just like any other unit test. You can use the keyboard combination CTRL-R, CTRL-A to run all of the tests in the current Visual Studio Solution. Alternatively, you can use the buttons in the Visual Studio toolbar to run the tests: (Unfortunately, the Run All Impacted Tests button won’t work correctly because Visual Studio won’t detect that your JavaScript code has changed. Therefore, you should use either the Run Tests in Current Context or Run All Tests in Solution options instead.) The results of running the JavaScript tests appear side-by-side with the results of running the server tests in the Test Results window. For example, if you Run All Tests in Solution then you will get the following results: Notice that the TestAddNumbers() JavaScript test has failed. That is good because our addNumbers() function is hard-coded to always return the value 5. If you double-click the failing JavaScript test, you can view additional details such as the JavaScript error message and the line number of the JavaScript code that failed: Summary The goal of this blog entry was to explain an approach to creating JavaScript unit tests that can be easily integrated with Visual Studio and Visual Studio ALM. I described how you can use the Microsoft Script Control to execute JavaScript on the server. By taking advantage of the Microsoft Script Control, we were able to execute our JavaScript unit tests side-by-side with all of our other unit tests and view the results in the standard Visual Studio Test Results window. You can download the code discussed in this blog entry from here: http://StephenWalther.com/downloads/Blog/JavaScriptUnitTesting/JavaScriptUnitTests.zip Before running this code, you need to first install the Microsoft Script Control which you can download from here: http://www.microsoft.com/downloads/en/details.aspx?FamilyID=d7e31492-2595-49e6-8c02-1426fec693ac

    Read the article

  • A Custom View Engine with Dynamic View Location

    - by imran_ku07
        Introduction:          One of the nice feature of ASP.NET MVC framework is its pluggability. This means you can completely replace the default view engine(s) with a custom one. One of the reason for using a custom view engine is to change the default views location and sometimes you need to change the views location at run-time. For doing this, you can extend the default view engine(s) and then change the default views location variables at run-time.  But, you cannot directly change the default views location variables at run-time because they are static and shared among all requests. In this article, I will show you how you can dynamically change the views location without changing the default views location variables at run-time.       Description:           Let's say you need to synchronize the views location with controller name and controller namespace. So, instead of searching to the default views location(Views/ControllerName/ViewName) to locate views, this(these) custom view engine(s) will search in the Views/ControllerNameSpace/ControllerName/ViewName folder to locate views.           First of all create a sample ASP.NET MVC 3 application and then add these custom view engines to your application,   public class MyRazorViewEngine : RazorViewEngine { public MyRazorViewEngine() : base() { AreaViewLocationFormats = new[] { "~/Areas/{2}/Views/%1/{1}/{0}.cshtml", "~/Areas/{2}/Views/%1/{1}/{0}.vbhtml", "~/Areas/{2}/Views/%1/Shared/{0}.cshtml", "~/Areas/{2}/Views/%1/Shared/{0}.vbhtml" }; AreaMasterLocationFormats = new[] { "~/Areas/{2}/Views/%1/{1}/{0}.cshtml", "~/Areas/{2}/Views/%1/{1}/{0}.vbhtml", "~/Areas/{2}/Views/%1/Shared/{0}.cshtml", "~/Areas/{2}/Views/%1/Shared/{0}.vbhtml" }; AreaPartialViewLocationFormats = new[] { "~/Areas/{2}/Views/%1/{1}/{0}.cshtml", "~/Areas/{2}/Views/%1/{1}/{0}.vbhtml", "~/Areas/{2}/Views/%1/Shared/{0}.cshtml", "~/Areas/{2}/Views/%1/Shared/{0}.vbhtml" }; ViewLocationFormats = new[] { "~/Views/%1/{1}/{0}.cshtml", "~/Views/%1/{1}/{0}.vbhtml", "~/Views/%1/Shared/{0}.cshtml", "~/Views/%1/Shared/{0}.vbhtml" }; MasterLocationFormats = new[] { "~/Views/%1/{1}/{0}.cshtml", "~/Views/%1/{1}/{0}.vbhtml", "~/Views/%1/Shared/{0}.cshtml", "~/Views/%1/Shared/{0}.vbhtml" }; PartialViewLocationFormats = new[] { "~/Views/%1/{1}/{0}.cshtml", "~/Views/%1/{1}/{0}.vbhtml", "~/Views/%1/Shared/{0}.cshtml", "~/Views/%1/Shared/{0}.vbhtml" }; } protected override IView CreatePartialView(ControllerContext controllerContext, string partialPath) { var nameSpace = controllerContext.Controller.GetType().Namespace; return base.CreatePartialView(controllerContext, partialPath.Replace("%1", nameSpace)); } protected override IView CreateView(ControllerContext controllerContext, string viewPath, string masterPath) { var nameSpace = controllerContext.Controller.GetType().Namespace; return base.CreateView(controllerContext, viewPath.Replace("%1", nameSpace), masterPath.Replace("%1", nameSpace)); } protected override bool FileExists(ControllerContext controllerContext, string virtualPath) { var nameSpace = controllerContext.Controller.GetType().Namespace; return base.FileExists(controllerContext, virtualPath.Replace("%1", nameSpace)); } } public class MyWebFormViewEngine : WebFormViewEngine { public MyWebFormViewEngine() : base() { MasterLocationFormats = new[] { "~/Views/%1/{1}/{0}.master", "~/Views/%1/Shared/{0}.master" }; AreaMasterLocationFormats = new[] { "~/Areas/{2}/Views/%1/{1}/{0}.master", "~/Areas/{2}/Views/%1/Shared/{0}.master", }; ViewLocationFormats = new[] { "~/Views/%1/{1}/{0}.aspx", "~/Views/%1/{1}/{0}.ascx", "~/Views/%1/Shared/{0}.aspx", "~/Views/%1/Shared/{0}.ascx" }; AreaViewLocationFormats = new[] { "~/Areas/{2}/Views/%1/{1}/{0}.aspx", "~/Areas/{2}/Views/%1/{1}/{0}.ascx", "~/Areas/{2}/Views/%1/Shared/{0}.aspx", "~/Areas/{2}/Views/%1/Shared/{0}.ascx", }; PartialViewLocationFormats = ViewLocationFormats; AreaPartialViewLocationFormats = AreaViewLocationFormats; } protected override IView CreatePartialView(ControllerContext controllerContext, string partialPath) { var nameSpace = controllerContext.Controller.GetType().Namespace; return base.CreatePartialView(controllerContext, partialPath.Replace("%1", nameSpace)); } protected override IView CreateView(ControllerContext controllerContext, string viewPath, string masterPath) { var nameSpace = controllerContext.Controller.GetType().Namespace; return base.CreateView(controllerContext, viewPath.Replace("%1", nameSpace), masterPath.Replace("%1", nameSpace)); } protected override bool FileExists(ControllerContext controllerContext, string virtualPath) { var nameSpace = controllerContext.Controller.GetType().Namespace; return base.FileExists(controllerContext, virtualPath.Replace("%1", nameSpace)); } }             Here, I am extending the RazorViewEngine and WebFormViewEngine class and then appending /%1 in each views location variable, so that we can replace /%1 at run-time. I am also overriding the FileExists, CreateView and CreatePartialView methods. In each of these method implementation, I am replacing /%1 with controller namespace. Now, just register these view engines in Application_Start method in Global.asax.cs file,   protected void Application_Start() { ViewEngines.Engines.Clear(); ViewEngines.Engines.Add(new MyRazorViewEngine()); ViewEngines.Engines.Add(new MyWebFormViewEngine()); ................................................ ................................................ }             Now just create a controller and put this controller's view inside Views/ControllerNameSpace/ControllerName folder and then run this application. You will find that everything works just fine.       Summary:          ASP.NET MVC uses convention over configuration to locate views. For many applications this convention to locate views is acceptable. But sometimes you may need to locate views at run-time. In this article, I showed you how you can dynamically locate your views by using a custom view engine. I am also attaching a sample application. Hopefully you will enjoy this article too. SyntaxHighlighter.all()  

    Read the article

  • ASP.net Ajax tab container not appearing

    - by Eyla
    I created new web project using VS 2008 with enabled Ajax template with C# and Framework 3.5. I added Ajax reference to the project and I can see all Ajax toolkit in my tool box. The problem that when I add tab container with Tab Panels then run the projects nothing appear on the browser and I tried few browsers. I'm including my code and I wish that someone would help me. Regards, My Code: ................................................................ <%@ Page Language="C#" AutoEventWireup="true" CodeBehind="Default.aspx.cs" Inherits="Contacts._Default" %> <%@ Register assembly="AjaxControlToolkit" namespace="AjaxControlToolkit" tagprefix="asp" %> <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd"> <html xmlns="http://www.w3.org/1999/xhtml" > <head runat="server"> <title>Untitled Page</title> </head> <body> <form id="form1" runat="server"> <asp:ScriptManager ID="ScriptManager1" runat="server" /> <div> <asp:TabContainer ID="TabContainer1" runat="server" ActiveTabIndex="0"> <asp:TabPanel runat="server" HeaderText="TabPanel1" ID="TabPanel1"> <ContentTemplate> tab 1 </ContentTemplate> </asp:TabPanel> <asp:TabPanel runat="server" HeaderText="TabPanel2" ID="TabPanel2"> <ContentTemplate> tab 2 </ContentTemplate> </asp:TabPanel> <asp:TabPanel runat="server" HeaderText="TabPanel3" ID="TabPanel3"> <ContentTemplate> tab 3 </ContentTemplate> </asp:TabPanel> </asp:TabContainer> </div> </form> </body> </html>

    Read the article

  • How to specify an area name in an action link?

    - by Jeremy
    I have a shared master page which I am using from 2 different areas in my mvc 2 app. The master page has an action link which currently specifies the controller and action, but of course the link doesn't work if I'm in the wrong area. I see no overload for actionlink that takes an area parameter, is it possible to do?

    Read the article

  • Where does ASP.NET Web API Fit?

    - by Rick Strahl
    With the pending release of ASP.NET MVC 4 and the new ASP.NET Web API, there has been a lot of discussion of where the new Web API technology fits in the ASP.NET Web stack. There are a lot of choices to build HTTP based applications available now on the stack - we've come a long way from when WebForms and Http Handlers/Modules where the only real options. Today we have WebForms, MVC, ASP.NET Web Pages, ASP.NET AJAX, WCF REST and now Web API as well as the core ASP.NET runtime to choose to build HTTP content with. Web API definitely squarely addresses the 'API' aspect - building consumable services - rather than HTML content, but even to that end there are a lot of choices you have today. So where does Web API fit, and when doesn't it? But before we get into that discussion, let's talk about what a Web API is and why we should care. What's a Web API? HTTP 'APIs' (Microsoft's new terminology for a service I guess)  are becoming increasingly more important with the rise of the many devices in use today. Most mobile devices like phones and tablets run Apps that are using data retrieved from the Web over HTTP. Desktop applications are also moving in this direction with more and more online content and synching moving into even traditional desktop applications. The pending Windows 8 release promises an app like platform for both the desktop and other devices, that also emphasizes consuming data from the Cloud. Likewise many Web browser hosted applications these days are relying on rich client functionality to create and manipulate the browser user interface, using AJAX rather than server generated HTML data to load up the user interface with data. These mobile or rich Web applications use their HTTP connection to return data rather than HTML markup in the form of JSON or XML typically. But an API can also serve other kinds of data, like images or other binary files, or even text data and HTML (although that's less common). A Web API is what feeds rich applications with data. ASP.NET Web API aims to service this particular segment of Web development by providing easy semantics to route and handle incoming requests and an easy to use platform to serve HTTP data in just about any content format you choose to create and serve from the server. But .NET already has various HTTP Platforms The .NET stack already includes a number of technologies that provide the ability to create HTTP service back ends, and it has done so since the very beginnings of the .NET platform. From raw HTTP Handlers and Modules in the core ASP.NET runtime, to high level platforms like ASP.NET MVC, Web Forms, ASP.NET AJAX and the WCF REST engine (which technically is not ASP.NET, but can integrate with it), you've always been able to handle just about any kind of HTTP request and response with ASP.NET. The beauty of the raw ASP.NET platform is that it provides you everything you need to build just about any type of HTTP application you can dream up from low level APIs/custom engines to high level HTML generation engine. ASP.NET as a core platform clearly has stood the test of time 10+ years later and all other frameworks like Web API are built on top of this ASP.NET core. However, although it's possible to create Web APIs / Services using any of the existing out of box .NET technologies, none of them have been a really nice fit for building arbitrary HTTP based APIs. Sure, you can use an HttpHandler to create just about anything, but you have to build a lot of plumbing to build something more complex like a comprehensive API that serves a variety of requests, handles multiple output formats and can easily pass data up to the server in a variety of ways. Likewise you can use ASP.NET MVC to handle routing and creating content in various formats fairly easily, but it doesn't provide a great way to automatically negotiate content types and serve various content formats directly (it's possible to do with some plumbing code of your own but not built in). Prior to Web API, Microsoft's main push for HTTP services has been WCF REST, which was always an awkward technology that had a severe personality conflict, not being clear on whether it wanted to be part of WCF or purely a separate technology. In the end it didn't do either WCF compatibility or WCF agnostic pure HTTP operation very well, which made for a very developer-unfriendly environment. Personally I didn't like any of the implementations at the time, so much so that I ended up building my own HTTP service engine (as part of the West Wind Web Toolkit), as have a few other third party tools that provided much better integration and ease of use. With the release of Web API for the first time I feel that I can finally use the tools in the box and not have to worry about creating and maintaining my own toolkit as Web API addresses just about all the features I implemented on my own and much more. ASP.NET Web API provides a better HTTP Experience ASP.NET Web API differentiates itself from the previous Microsoft in-box HTTP service solutions in that it was built from the ground up around the HTTP protocol and its messaging semantics. Unlike WCF REST or ASP.NET AJAX with ASMX, it’s a brand new platform rather than bolted on technology that is supposed to work in the context of an existing framework. The strength of the new ASP.NET Web API is that it combines the best features of the platforms that came before it, to provide a comprehensive and very usable HTTP platform. Because it's based on ASP.NET and borrows a lot of concepts from ASP.NET MVC, Web API should be immediately familiar and comfortable to most ASP.NET developers. Here are some of the features that Web API provides that I like: Strong Support for URL Routing to produce clean URLs using familiar MVC style routing semantics Content Negotiation based on Accept headers for request and response serialization Support for a host of supported output formats including JSON, XML, ATOM Strong default support for REST semantics but they are optional Easily extensible Formatter support to add new input/output types Deep support for more advanced HTTP features via HttpResponseMessage and HttpRequestMessage classes and strongly typed Enums to describe many HTTP operations Convention based design that drives you into doing the right thing for HTTP Services Very extensible, based on MVC like extensibility model of Formatters and Filters Self-hostable in non-Web applications  Testable using testing concepts similar to MVC Web API is meant to handle any kind of HTTP input and produce output and status codes using the full spectrum of HTTP functionality available in a straight forward and flexible manner. Looking at the list above you can see that a lot of functionality is very similar to ASP.NET MVC, so many ASP.NET developers should feel quite comfortable with the concepts of Web API. The Routing and core infrastructure of Web API are very similar to how MVC works providing many of the benefits of MVC, but with focus on HTTP access and manipulation in Controller methods rather than HTML generation in MVC. There’s much improved support for content negotiation based on HTTP Accept headers with the framework capable of detecting automatically what content the client is sending and requesting and serving the appropriate data format in return. This seems like such a little and obvious thing, but it's really important. Today's service backends often are used by multiple clients/applications and being able to choose the right data format for what fits best for the client is very important. While previous solutions were able to accomplish this using a variety of mixed features of WCF and ASP.NET, Web API combines all this functionality into a single robust server side HTTP framework that intrinsically understands the HTTP semantics and subtly drives you in the right direction for most operations. And when you need to customize or do something that is not built in, there are lots of hooks and overrides for most behaviors, and even many low level hook points that allow you to plug in custom functionality with relatively little effort. No Brainers for Web API There are a few scenarios that are a slam dunk for Web API. If your primary focus of an application or even a part of an application is some sort of API then Web API makes great sense. HTTP ServicesIf you're building a comprehensive HTTP API that is to be consumed over the Web, Web API is a perfect fit. You can isolate the logic in Web API and build your application as a service breaking out the logic into controllers as needed. Because the primary interface is the service there's no confusion of what should go where (MVC or API). Perfect fit. Primary AJAX BackendsIf you're building rich client Web applications that are relying heavily on AJAX callbacks to serve its data, Web API is also a slam dunk. Again because much if not most of the business logic will probably end up in your Web API service logic, there's no confusion over where logic should go and there's no duplication. In Single Page Applications (SPA), typically there's very little HTML based logic served other than bringing up a shell UI and then filling the data from the server with AJAX which means the business logic required for data retrieval and data acceptance and validation too lives in the Web API. Perfect fit. Generic HTTP EndpointsAnother good fit are generic HTTP endpoints that to serve data or handle 'utility' type functionality in typical Web applications. If you need to implement an image server, or an upload handler in the past I'd implement that as an HTTP handler. With Web API you now have a well defined place where you can implement these types of generic 'services' in a location that can easily add endpoints (via Controller methods) or separated out as more full featured APIs. Granted this could be done with MVC as well, but Web API seems a clearer and more well defined place to store generic application services. This is one thing I used to do a lot of in my own libraries and Web API addresses this nicely. Great fit. Mixed HTML and AJAX Applications: Not a clear Choice  For all the commonality that Web API and MVC share they are fundamentally different platforms that are independent of each other. A lot of people have asked when does it make sense to use MVC vs. Web API when you're dealing with typical Web application that creates HTML and also uses AJAX functionality for rich functionality. While it's easy to say that all 'service'/AJAX logic should go into a Web API and all HTML related generation into MVC, that can often result in a lot of code duplication. Also MVC supports JSON and XML result data fairly easily as well so there's some confusion where that 'trigger point' is of when you should switch to Web API vs. just implementing functionality as part of MVC controllers. Ultimately there's a tradeoff between isolation of functionality and duplication. A good rule of thumb I think works is that if a large chunk of the application's functionality serves data Web API is a good choice, but if you have a couple of small AJAX requests to serve data to a grid or autocomplete box it'd be overkill to separate out that logic into a separate Web API controller. Web API does add overhead to your application (it's yet another framework that sits on top of core ASP.NET) so it should be worth it .Keep in mind that MVC can generate HTML and JSON/XML and just about any other content easily and that functionality is not going away, so just because you Web API is there it doesn't mean you have to use it. Web API is not a full replacement for MVC obviously either since there's not the same level of support to feed HTML from Web API controllers (although you can host a RazorEngine easily enough if you really want to go that route) so if you're HTML is part of your API or application in general MVC is still a better choice either alone or in combination with Web API. I suspect (and hope) that in the future Web API's functionality will merge even closer with MVC so that you might even be able to mix functionality of both into single Controllers so that you don't have to make any trade offs, but at the moment that's not the case. Some Issues To think about Web API is similar to MVC but not the Same Although Web API looks a lot like MVC it's not the same and some common functionality of MVC behaves differently in Web API. For example, the way single POST variables are handled is different than MVC and doesn't lend itself particularly well to some AJAX scenarios with POST data. Code Duplication I already touched on this in the Mixed HTML and Web API section, but if you build an MVC application that also exposes a Web API it's quite likely that you end up duplicating a bunch of code and - potentially - infrastructure. You may have to create authentication logic both for an HTML application and for the Web API which might need something different altogether. More often than not though the same logic is used, and there's no easy way to share. If you implement an MVC ActionFilter and you want that same functionality in your Web API you'll end up creating the filter twice. AJAX Data or AJAX HTML On a recent post's comments, David made some really good points regarding the commonality of MVC and Web API's and its place. One comment that caught my eye was a little more generic, regarding data services vs. HTML services. David says: I see a lot of merit in the combination of Knockout.js, client side templates and view models, calling Web API for a responsive UI, but sometimes late at night that still leaves me wondering why I would no longer be using some of the nice tooling and features that have evolved in MVC ;-) You know what - I can totally relate to that. On the last Web based mobile app I worked on, we decided to serve HTML partials to the client via AJAX for many (but not all!) things, rather than sending down raw data to inject into the DOM on the client via templating or direct manipulation. While there are definitely more bytes on the wire, with this, the overhead ended up being actually fairly small if you keep the 'data' requests small and atomic. Performance was often made up by the lack of client side rendering of HTML. Server rendered HTML for AJAX templating gives so much better infrastructure support without having to screw around with 20 mismatched client libraries. Especially with MVC and partials it's pretty easy to break out your HTML logic into very small, atomic chunks, so it's actually easy to create small rendering islands that can be used via composition on the server, or via AJAX calls to small, tight partials that return HTML to the client. Although this is often frowned upon as to 'heavy', it worked really well in terms of developer effort as well as providing surprisingly good performance on devices. There's still plenty of jQuery and AJAX logic happening on the client but it's more manageable in small doses rather than trying to do the entire UI composition with JavaScript and/or 'not-quite-there-yet' template engines that are very difficult to debug. This is not an issue directly related to Web API of course, but something to think about especially for AJAX or SPA style applications. Summary Web API is a great new addition to the ASP.NET platform and it addresses a serious need for consolidation of a lot of half-baked HTTP service API technologies that came before it. Web API feels 'right', and hits the right combination of usability and flexibility at least for me and it's a good fit for true API scenarios. However, just because a new platform is available it doesn't meant that other tools or tech that came before it should be discarded or even upgraded to the new platform. There's nothing wrong with continuing to use MVC controller methods to handle API tasks if that's what your app is running now - there's very little to be gained by upgrading to Web API just because. But going forward Web API clearly is the way to go, when building HTTP data interfaces and it's good to see that Microsoft got this one right - it was sorely needed! Resources ASP.NET Web API AspConf Ask the Experts Session (first 5 minutes) © Rick Strahl, West Wind Technologies, 2005-2012Posted in Web Api   Tweet !function(d,s,id){var js,fjs=d.getElementsByTagName(s)[0];if(!d.getElementById(id)){js=d.createElement(s);js.id=id;js.src="//platform.twitter.com/widgets.js";fjs.parentNode.insertBefore(js,fjs);}}(document,"script","twitter-wjs"); (function() { var po = document.createElement('script'); po.type = 'text/javascript'; po.async = true; po.src = 'https://apis.google.com/js/plusone.js'; var s = document.getElementsByTagName('script')[0]; s.parentNode.insertBefore(po, s); })();

    Read the article

  • Do I lose anything by coding in c# and using free online vb.net code convertors?

    - by Gullu
    The company I work for uses vb.net since there are many programmers who moved up from vb6 to vb.net. Basically more vb.net resources in the company for support/maintenance vs c#. I am a c# coder and was wondering if I could just continue coding in c# and just use the many online free c# to vb.net code convertors. That way, I will be more productive and also be more marketable since there are more c# jobs compared to vb.net jobs. I have done vb6 many years ago and I am comfortable debugging vb.net code. It's just the primary coding language. I am more comfortable in c#. Will I lose anything if I use this approach. (code conversion). Based on what i read online the future of vb.net is really "Dim". Please advise. thank you

    Read the article

  • Active Server Pages error 'ASP 0126' classic asp pls help

    - by sagarmatha
    our company have a a old classic asp application, we have no choice but to host it. I just moved it to another server. It was perfectly running fine in old server but in this new server it's continuously giving this error. I am running windows 2003 server with IIS 6. Why I am geting this error ? please help. Active Server Pages error 'ASP 0126' Include file not found /application/unprocessed_application.asp, line 56 The include file '../../_fplclass/pdblib.inc' was not found.

    Read the article

  • Shallow Copy vs DeepCopy in C#.NET

    Hope below example helps to understand the difference. Please drop a comment if any doubts. using System; using System.IO; using System.Runtime.Serialization.Formatters.Binary; namespace ShallowCopyVsDeepCopy {     class Program     {         static void Main(string[] args)         {             var e1 = new Emp { EmpNo = 10, EmpName = "Smith", Department = new Dep { DeptNo = 100, DeptName = "Finance" } };             var e2 = e1.ShallowClone();             e1.Department.DeptName = "Accounts";             Console.WriteLine(e2.Department.DeptName);             var e3 = new Emp { EmpNo = 10, EmpName = "Smith", Department = new Dep { DeptNo = 100, DeptName = "Finance" } };             var e4 = e3.DeepClone();             e3.Department.DeptName = "Accounts";             Console.WriteLine(e4.Department.DeptName);         }     }     [Serializable]     class Dep     {         public int DeptNo { get; set; }         public String DeptName { get; set; }     }     [Serializable]     class Emp     {         public int EmpNo { get; set; }         public String EmpName { get; set; }         public Dep Department { get; set; }         public Emp ShallowClone()         {             return (Emp)this.MemberwiseClone();         }         public Emp DeepClone()         {             MemoryStream ms = new MemoryStream();             BinaryFormatter bf = new BinaryFormatter();             bf.Serialize(ms, this);             ms.Seek(0, SeekOrigin.Begin);             object copy = bf.Deserialize(ms);             ms.Close();             return copy as Emp;         }     } } span.fullpost {display:none;}

    Read the article

  • How should data be passed between client-side Javascript and C# code behind an ASP.NET app?

    - by ctck
    I'm looking for the most efficient / standard way of passing data between client-side Javascript code and C# code behind an ASP.NET application. I've been using the following methods to achieve this but they all feel a bit of a fudge. To pass data from Javascript to the C# code is by setting hidden ASP variables and triggering a postback: <asp:HiddenField ID="RandomList" runat="server" /> function SetDataField(data) { document.getElementById('<%=RandomList.ClientID%>').value = data; } Then in the C# code I collect the list: protected void GetData(object sender, EventArgs e) { var _list = RandomList.value; } Going back the other way I often use either ScriptManager to register a function and pass it data during Page_Load: ScriptManager.RegisterStartupScript(this.GetType(), "Set","get("Test();",true); or I add attributes to controls before a post back or during the initialization or pre-rendering stages: Btn.Attributes.Add("onclick", "DisplayMessage("Hello");"); These methods have served me well and do the job, but they just dont feel complete. Is there a more standard way of passing data between client side Javascript and C# backend code? Ive seen some posts like this one that describe HtmlElement class; is this something I should look into?

    Read the article

< Previous Page | 43 44 45 46 47 48 49 50 51 52 53 54  | Next Page >