I'd like to hear some reasons for using a ServerControl opposed to a UserControl. I've found that I probably overuse UserControls.
My list looks something like this:
Pro UserControl
Easily modified. Need to add a class attribute to an element, hack it out in html.
Quick and easy to create initial view. Everyone can write simple html right?
Pro ServerControl
Performance. No html parsing.
Flexibility. Control rendering down to a gnat's behind.
Reusability. Compile it and stick it in the GAC for later use. Or, sell it.
Anything that I'm missing?
The following two advantages of server controls come to mind:
support for inheritance (because the actual user control at runtime inherits from your code behind, you can't inherit from it anymore)
detailed control over the tag's parsing (should nested tags be treated as child controls or properties, etc.)
support for control adapters (declarative, e.g. in a .browsers file)
support for tag mapping (if you're making your own controls that's often very useful though)
However, user controls do have an additional edge because they're templated:
support for databinding expressions
One other consideration is whether you want visual studio designer support at the time you're building the control or at the time you're using the control on the page. It really irks me that you only get one or the other, but not both.
Performance. No html parsing
I wouldn't expect that to be much of a runtime difference - the .ascx should get compiled on first use as well. Perhaps there's some slight overhead saved by not checking the file for modifications - but I can't think of much else.
Related
Is there any reason use standart HTML controls (input type=text,input type=checkbox) instead of asp.net controls ( asp:TextBox, asp:CheckBox) to improve performance?
IMHO this would be a micro optimization. You will gain performance but it won't be noticeable. On the other hand you will loose much of the flexibility offered by the server controls.
You could try to reduce the size of the ViewState by disabling it for controls that don't need it. This will reduce the size of the generated pages and improve performance.
The ASP.NET user controls will all have ViewState associated with them unless you explicitly set
EnableViewState="False"
As such, you will bloat the size of the underlying page if you have a large number of controls. As a general rule, use what meets your needs and nothing more.
Do you need to access the user control in the code-behind?
Do you need the control to maintain value across post-backs etc?
In most cases, it won't make a difference, but it is good to keep your page clean if you don't need these features.
As always with performance optimizations: it depends on the situation. Test it in your project and see if it makes any difference.
Also with .net 4.0 another con of using server controls is gone, since you can set ClientIDMode to Static, which will give you full control over ID's on your controls. Previously using just a standard textbox or button (without viewstate) would still render crazy non-readable ID's because of the way Naming Containers work. Those days are over now though :)
Remember your three options are:
use regular html which can't be referenced on the server.
add runat="server" to your existing html-tags (ie. ) and you'll be able to access it as an HtmlControl.
use the asp.net tags (<asp:* runat="server" />)
The disadvantage of option 3 is that you don't always know what the rendered html-markup will be and you have less control over it. I personally only use option 3 for more advanced controls like , the button () and third party controls. For normal html-markup that I need to reference on the server I prefer option 2.
Regarding performance I would mainly look at the rendered output, how much extra bloat is rendered to the client and such. CPU-time on the server using one or the other approach I would say is secondary compared to the different caching techniques ASP.Net has already.
I am designing an ASP.NET v3.5 custom control based on CompositeControl.
However, I do NOT want to create my child controls via code, but rather as
embedded HTML tag elements within the CompositeControl in the ASPX page, such as:
<cc:MyCompositeControl ID="MyControl">
<asp:Label>Cat</asp:Label>
<asp:Label>Cat</asp:Label>
</cc:MyCompositeControl>
At run time, MyControl.Controls contains the two labels as expected.
At design time, MyControl.Controls is empty.
I need this information at design time for various reasons.
What am I doing wrong?
You might consider taking a look at the System.Web.UI.WebControls.Panel control in the System.Web DLL using Reflector. That will help you get some idea of what properties and what attribute decorations are necessary to provide the design-time support you need.
Would this code be considered bad practice:
<div id="sidebar">
<% =DisplayMeetings(12) %>
</div>
This is a snippet of code from the default.aspx of a small web app I have worked on. It works perfectly well, runs very fast, but as always, I am aware of the fact that just because it works, doesn't mean it is OK.
Basically, the DisplayMeetings subroutine outputs a bunch of formatted HTML (an unordered list actually), with no formatting, just the requisite html, and then my CSS performs all the necessary formatting.
The data for generating the list comes from an SQL server database (the parameter controls how many rows to return) and I am using stored procedures and datareader for fast access. This keeps my front-end extraordinary simple and clean, imho, and lets me do all the work in VB or C# in a separate module.
I could of course use a databound repeater (and probably 6 or more other methods) of accomplishing the same thing, but are they any better? Other than loosing the design-time features of VS2010?
The only thing "wrong" with your approach is that it mixes code and display, which you typically want to avoid wherever possible. If you have to have a procedurally generated portion of HTML (because it's just that difficult to do with controls, or whatever), create a control responsible for generating that HTML and embed it in the larger page/control that contains it.
is the "wrong" part, that the subroutine is returning HTML, or is it that I have a code-snippet executed from within the HTML markup?
Both, in a way. And also neither.
There's nothing directly "wrong" with using <%= foo %>; if there were, it wouldn't be part of the framework. What's "wrong" with it is that it sets up a two-way dependency that you don't necessarily want. Your HTML markup is dependent on and has to know about the code-behind. It has to call the method, which in turn is dedicated to the markup and nothing else. If you want to make changes to your output, you may have to change the method, the markup, or both.
What I'm trying to say is that what you're doing makes your code less maintainable, less flexible to modify the next time you have to open the hood.
If it's the only way to solve a problem, then it's the only way, and there's nothing wrong with that. But it's something that should be avoided, in general, if possible.
How would I have done it? That depends on the situation, frankly. If I could have used a standard component with databinding, I'd have done that. That is always preferred. If the HTML was just too complex to use a component, I'd use a Literal control as a placeholder for the markup, and generate the markup in a separate component--likely a User Control.
The point is that the markup knows nothing about the method used to generate the other markup, it just says "something goes here" and relies on its code-behind to handle deciding which markup to put in there.
It's not something I'd do. I'm not a fan of using <%= %> in Webforms. Sadly, it's also common practice.
What I'd probably do instead is a build a user control to represent the markup I wanted and put that on the form. This has the advantages of making the element more re-usable, testable, and gives you more control over where in the page life cycle the code is called.
A databound repeater can make it somewhat simpler to modify your html when you need to change the layout. It is also nice if you have separate people who work on html vs people who work on the server side code.
My usual rule is to use a repeater for any even slightly complex html. And I do not call methods from my aspx/ascx file - I only insert the values of a protected string variable which I have populated in the code. However, these are personal preferences, and I don't see anything really wrong with what you are doing.
Your code (without seeing it) will likely be faster than a repeater, since there is no data binding involved, but it probably wouldn't be a huge thing unless the page is very, very popular.
I know why runat="server" is currently required (ASP.NET why runat="server"), but the consensus is that it should not be required if you incorporate a simple default into the design (I agree of course).
Would it be possible to modify, extend, decompile and recreate, intercept or otherwise change the behavior of how ASP.NET parses ASPX and ASCX files so that runat="server" would no longer be required? For instance, I assume that a version of Mono could be branched to accomplish this goal.
In case specific requirements are helpful, the following highlights one design:
During parsing, when configured namespace tags are encountered (such as "asp"), default the element's runat property to "server"
During parsing, when configured namespace tags are encountered (such as "asp"), if the element's runat property value is available, then that value should be used in place of the default
New page-level setting introduced (can be set in the page directive or web.config) that specifies the default runat value for a specific namespace tag
I'm afraid you'd have to modify the entire page parser to accomplish this, and I don't think that's possible.
On the other hand, you should be able to create your own. See the buildProviders Element and the BuildProvider class. You should be able to create your own build provider for .aspx pages and use it to replace the built-in provider.
Unfortunately, the PageBuildProvider class used by ASP.NET is internal, and the PageParser class that it uses to parse pages is sealed. You'll be entirely on your own.
Consider that runat="server" has been in ASP.NET for a decade now, and I think you'll see that this won't change anytime soon.
You'd also lose Designer support, but maybe you don't care about that.
So far as I know, there are no hooks deep enough in the ASP.NET Page handling process that would allow this. I know of no way to override or extend the parsing or processing of actual aspx/ascx code.
While ASP.NET is fairly flexible and lets your override many default behaviors (like how ViewState is saved/loaded, where Session is stored, etc) this is not one of them.
However... technically the Page object is just another HttpHandler. You could write your handler and do anything you wanted with it. All you have to do is implement everything the Page class does and then throw in this extra functionality. :) Alternately, pull out Reflector and dig through the Page object's ProcessRequest method and see where it is actually parsing/initializing the objects declared in aspx and you might get a clue how to implement the functionality you're looking for. But I suspect you'd be wasting your time.
In terms of performance. Is it a good idea to build your own data grid (assuming you have the time) having in mind that you will create one that match perfectly your need? Getting rid of many unnecessary properties and methods. The rendered page might be smaller making the download faster?
thanks
The properties and methods of a server control have no direct relationship with the size of the rendered HTML. Simply removing methods won't make the slightest difference in the size of your download. In my experience, the .NET controls work well enough. If you encounter a performance problem, you can always drop in a different implementation later.
As ASP.NET already includes the Repeater control and the DataList control, both of which allow you to specify the HTML output of the control, there is little need to create a new control for performance reasons.