Ed Giardina's .NET Blog

Embedding .NET code and Windows Forms controls into ASPX pages

Sometimes you can't trust the users of your web application to be savvy enough to navigate away from your application. Most likely, what happens is that a user must do a third party process in order to pass data to your web application, such as scanning in an image before uploading it. Having the user need to minimize the browser window, then open a scanning application, then save the file to a remember-able place, then restore the browser window and 'Browse' to that location can be very overwhelming to people not necessarily technical.

Eliminating the steps above and encompassing as much of the functionality within the browser can be very convienient. As a primer, see this article: http://msdn.microsoft.com/msdnmag/issues/02/06/rich/default.aspx

There are a few interesting hiccups I ran into trying to get this to work, and I wanted to share these with you:

• Getting the control to load can be difficult. To make sure the control itself is loading, add a simple button on the user control. If the control loads with the 'image' type square in the top left, and you don't see your button, check the path to the DLL. make sure the DLL is NOT in the Bin folder (IIS will not serve files from the Bin folder).
• If you get an 'Object does not support this property or method' error client side in your javascript when trying to access public methods of your class, be sure you have ComVisible set to true above your definition (see below). This most often happens with .NET 2.0
  

The ComVisible attribute was the hardest nut to crack. Above your class definition, do as such:

1 2	[ComVisible(true)] public class MyClass : UserControl{...}

This will get rid of the accesibility problems for javascript and the pesky 'object does not support' messages.

What did I use this all for? Well, we support a large base of users who are not technically savvy. As part of their compliance operations they need to have a document management solution that keeps a digital copy of their documents and letters. Our web application has a form to fill out document information, but instead of a 'browse' button for the file itself, we have a 'scan' button in the Windows forms control inside the webpage. When the user clicks that button, the Twain interface takes over, letting them crop the document. When they finish the scanning, Twain passes the image back to our control, which uploads it to the Web Server along with the other form information.

What? No Columns in my GridView?

Well, lesson learned. the GridView is not infalliable.

So I am building another nightmare report that will sift through reams of data. This one has two Gridviews, one nested inside of the other. The point of the Gridview will be to display an asset and have the child gridview's rows display the locations where the asset exists. Aside from the entire table being a Pivot/Crosstab, and that I have to expect, on minimum, 100 assets to be returned for any search, I am also working with a custom data layer written for the project. So binding to these grids is, well, challenging.

To keep the amount of database traffic down, I only make two queries: One to get the asset list, and one to get all locations for those assets. Storing the locations in a datatable, I can use the datatable's 'Select' method to populate the subgrid. One problem we have is that the key field we 'Select' on, the ID of the asset, is already listed in the parent table, so I want to hide it.

    <asp:TemplateField>
        <ItemTemplate>
            <td>&nbsp;<td colspan="5">
            </asp:GridView runat="server" ID="SubViewContainer" CssClass="Grid"/>
        </ItemTemplate>
    </asp:TemplateField>
</Columns>


Here is the nested grid, inside a template field in the parent grid. On the RowDataBound event of the parent grid, I bind the child grid by extracting the control from the parent grid's columns. With the grid extracted, I can bind it individually.

GridView SubView = (GridView)e.Row.Cells[6].Controls[1];
DataRow[] SubData = WeeklySubData.Tables[0].Select("ID = '" + ID + "'");
DataTable Tbl = WeeklySubData.Tables[0].Clone();

foreach (DataRow D in SubData)
{
    Tbl.ImportRow(D);
}
SubView.DataSource = Tbl;
SubView.DataBind();

So this gets us our sub-data bound to the sub-grid. Great! Now let's hide the first column. That first column is only there so we can use the Select method, we don't need to display it. So let's just do
SubView.Columns[0].visible = false;
, right?

Actually, no. For some reason, I don't have access to the Columns and that line would throw a runtime exception. The columns will have a Count of 0 and I can't do anything. So what's the solution?

Eventually I broke down and added an event handler to handle the 'OnRowCreated' event of the sub-grid. When this event fires, we're able to see the row we're in and hide the first cell of that row. I'm sure this is probably more expensive performance-wise; but it also works, which is the primary feature.

The event handler attachment code looks like 

SubView.RowCreated += new GridViewRowEventHandler(SubView_RowCreated);

And the event handler looks like:

void SubView_RowCreated(object sender, GridViewRowEventArgs e)
{
    e.Row.Cells[0].Visible = false;
}

Problem solved. now to explain to my boss what I've been doing all morning.

"Failed to Enable Constraints" Error, and the Zen of DataSets

If you're like me, you love the TableAdapters feature of Visual Studio 2005. Simply right-click on the App_Code folder, add a new item, and choose DataSet. Then pick a database to model it against, and BAM! A Strongly typed data object is built for you on the fly. This model enhances portability for the code we write at my company, since we have a core set of tools that look at different models depending on where you are in my company. (Texas has a different application running their operations versus us here in Boston. Therefore, their data model is different. I could write two datasets and one presentation layer to cover both.)

Anyways, So I went a little DataSet-crazy by making a TableAdapter for every View we use (essentially, every business object has its own Table Adapter). Then I wrote accessor-methods to get the data out of the TableAdapter depending on the Report (some reports filtered by X, others by Y, etc).

However, one of the methods I was calling was failing pretty badly.

1	ReportDataIndex.DataSource = T.GetDataGroupByWeek(ReportDate.SelectedDates[0]);

Seems pretty simple, right? ReportDataIndex is a GridView, and T is an instance of my TableAdapter. But every time I ran it, I got an exception: Failed to enable constraints. One or more rows  contain values violating non-null, unique, or foreign-key  constraints.

I read and read and read, and couldn't figure it out. A few forums recognized the problem, such as this article about DataSets, but none specifically mentioned the problem as it could have related to TableAdapters. I couldn't turn off 'EnforceConstraints' on my GridView, and I really struggled. I tried wrapping the Query Columns with 'ISNULL()' but that didnt' work either. I restricted my data results from the DB to only include non-nulls, and it was still failing.

After much soul-searching, I found the issue: If you create additional accessor-methods to your DataSet, their column count has to match the TableAdapter column count. Any columns not mentioned in your addtional method is passed back into the DataSet object as NULL. That's why the Fill was failing. Now, when I build my accessor-method and Query in the designer, it should throw an error and not let me save the DataSet / Compile. However, no pre-compilation is done with this regard on the XML document generated by the DataSet designer.

This is kind of limiting and if you have a TableAdapter that has two methods, one to get rows and one to get DISTINCT subset of rows, it can really cause problems. However, this small flaw is no reason to abandon DataSets and TableAdapters, as they are still great for rapidly prototyping data structures in your application