In order to work around a KPI problem I wrote about here, I did some testing and discovered that KPI’s work against folders with meta data in the same way that they work against documents or list items.   I proved it out by creating a new content type based on the folder content type and then added a few fields.  I created some indicators and proved to myself that KPIs work as expected.  This was welcome news.  It’s not perfect, because the drill-down you get from the KPI against the folders is not exactly what you want.  This isn’t too much a drawback in my case because 1) the end users don’t know any better and 2) the drill-down goes to a folder.  They click the folder name and they are at the item.  It’s two clicks instead of one, which isn’t the end of the world.

This flowed nicely with the work I was doing.  I am creating a folder for every document that gets uploaded.  This is done via an event receiver.  As a result, it’s a piece of cake to keep the parent folder’s meta data in sync with the KPI-driven meta data from the file itself since the plumbing is already in place.  This allows me to have my KPI’s and eat them too 🙂

I modified the event receiver to add the folder and then set this new folder’s content type to my custom KPI-friendly content type.  This bit of code did the trick:

 SPFolderCollection srcFolders = targetWeb.GetFolder("Documents").SubFolders;
  SPFolder addedFolder = srcFolders.Add(properties.ListItem.ID.ToString());

  SPContentTypeId kpiCT = new SPContentTypeId("0x0120002A666CAA9176DC4AA8CBAA9DC6B4039F");
  addedFolder.Item["Content Type ID"] = kpiCT;
  addedFolder.Item.Update();

To locate the actual Content Type ID, I accessed that content type via site settings and copy/pasted it from the URL as shown:

</end>

Subscribe to my blog!

I hate to admit it, but I struggled with this one all day.  My event receiver needs to update a field of its parent folder.  This little bit shows how to do it:

        private void UpdateParentFolder(SPItemEventProperties properties)
        {

            SPFolder thisItemFolder = properties.ListItem.File.ParentFolder;
            thisItemFolder.Item["ZZ Approval Status"] = "Good news, everyone!";
            thisItemFolder.Item.Update();
           
           
        } // UpdateParentFolder

In this case, I’m working with a document library and the properties are coming from an ItemAdded event.

The trick is that you can’t get the SPFolder of the item directly from the item itself (i.e. properties.ListItem.Folder is null).  Instead, go to the list item’s associated File and get the File’s folder.

</end>

 Subscribe to my blog!

I’m sure I’m not the first person to come up with this.  However, I haven’t noticed anyone publish a trick like this since I started paying close attention to the community last July.  So, I thought I’d post it this quick and easy debug tip.

I’m working on an event receiver that started to generate this error in the 12 hive:

Error loading and running event receiver Conchango.xyzzyEventReceiver in xyzzy, Version=1.0.0.0, Culture=neutral, PublicKeyToken=blahbalhbalh. Additional information is below.  : Object reference not set to an instance of an object.    

I didn’t know where I had introduced this bug because I had done too many things in one of my code/deploy/test cycles. 

I tried this solution to get my pdb in there with hopes that SharePoint’s 12 hive would show the stack trace, but no luck.  I don’t know if it’s possible and if someone does, please let me know 🙂

I know it’s possible to write your own log messages to the 12 hive.  Frankly, I wanted something a little less scary and quicker to implement.

It occurred to me that I could at least get some basic trace information by catching and re-throwing generic exceptions like this:

  try {
    UpdateEditionDate(properties);
  }
  catch (Exception e)
  {
    throw new Exception("Dispatcher, UpdateEditionDate(): Exception: [" + e.ToString() + "].");
  }

This showed up in the 12 hive thusly:

Error loading and running event receiver Conchango.xyzzyEventReceiver in xyzzy, Version=1.0.0.0, Culture=neutral, PublicKeyToken=blahblahblah. Additional information is below.  : Dispatcher, UpdateEditionDate(): Exception: [System.NullReferenceException: Object reference not set to an instance of an object.     at Conchango.xyzzyManagementEventReceiver.UpdateEditionDate(SPItemEventProperties properties)     at Conchango.xyzzyManagementEventReceiver.Dispatcher(SPItemEventProperties properties, String eventDescription)].

That gave me all the detail I needed to track down that particular problem and I expect to use it a lot going forward.

</end>

Subscribe to my blog!

This past week I was implementing an "evolving" solution for a client that uses BDC and SPQuery and ran into some difficulty using SPQuery against a document library containing folders.  Bottom line: assign "recursive" to the view attribute of the query.

My scenario:

• On Monday, I upload a document and supply some meta data.
• The following week, I upload a new document.  Much of this new document’s meta data is based on the document I uploaded on Monday (which we call the "master document").
• We’ve created a web service façade that provides a BDC-friendly interface to the list so that users can easily locate that Monday document via a title search.
• A BDC data column provides a friendly user interface.  (This is part of my attempt at using BDC for a more friendly Lookup column).

The final BDC façade service uses a query like this to do the lookup:

 // Used U2U tool to assist in generating this CAML query.
      oQuery.Query =
        "<Where>";

      if (titleFilter.Length > 0)
        oQuery.Query +=
          "  <And>";

      oQuery.Query +=
        "    <And>" +
        "      <Geq>" +
        "        <FieldRef Name=\"DocumentId\" />" +
        "        <Value Type=\"Text\">" + minId + "</Value>" +
        "      </Geq>" +
        "      <Leq>" +
        "        <FieldRef Name=\"DocumentId\" />" +
        "        <Value Type=\"Text\">" + maxId + "</Value>" +
        "      </Leq>" +
        "    </And>";

      if (titleFilter.Length > 0)
        oQuery.Query +=
          "    <Contains>" +
          "      <FieldRef Name=\"Title\" />" +
          "      <Value Type=\"Text\">" + titleFilter + "</Value>" +
          "    </Contains>" +
          "  </And>";
      oQuery.Query +=
        "</Where>";

During the initial stage of development, this worked great.  However, we introduced folders into the directory to solve some problems and suddenly, my BDC picker wouldn’t return any results.  I tracked this down to the fact that the SPQuery would never return any results.  We used folders primarily to allow multiple files with the same name to be uploaded but with different meta data.  When the file is uploaded, we create a folder based on the list item’s ID and then move the file there (I wrote about that here; we’ve had mixed results with this approach but on the whole, it’s working well).  The user don’t care about folders and in fact, don’t really understand that there are any folders.  We have configured all the views on the library to show items without regard to folders.

I hit this problem twice as the technical implementation evolved and solved it differently each time.  The first time, I wasn’t using the CONTAINS operator in the query.  Without a CONTAINS operator, I was able to solve the problem by specifying the view on the SPQuery’s contructor.   Instead of using the default constructor:

SPList oList = web.Lists["Documents"];

SPQuery oQuery = new SPQuery();

I instead used a constructor that specified a view:

SPList oList = web.Lists["Documents"];

SPQuery oQuery = new SPQuery(oList.Views["All Documents"]);

That solved the problem and I started to get my results.

I then added the CONTAINS operator into the mix and it broke again.  It turns out that the CONTAINS operator, so far as I can tell, does not work with the view the same way as the a simpler GEQ / LEQ operators.  I did some searching and learned that the query’s ViewAttributes should be set to "Recursive", as in:

oQuery.ViewAttributes = "Scope=\"Recursive\"";

That solved the problem for CONTAINS.  In fact, this also solved my original search problem and if I had specified the recursive attribute the first time, I would not have run into the issue again.

The fact that a view-based SPQuery works for some operators (GEQ/LEQ) and not others (CONTAINS), coupled with the fact that KPIs don’t seem to work at all with folder-containing document libraries leads me to believe that SPQuery has some orthogonality issues.

Special Thanks:

• The good folks at U2U and their query tool.
• Michael Hoffer’s great "learning by doing" blog post, comments and responses.

</end>

 Subscribe to my blog!

UPDATE 02/29/08: I solved this problem by creating a folder and then assigning a content type to the folder which has the meta data I need for the KPIs.  I described that in a little more detail here.

We have implemented a technical solution where users upload documents to a document library.  An event receiver creates a directory and moves the file to that directory (using a technique similar to what I wrote about here).  We’ve successfully navigated around the potential issues caused by event receivers that rename uploaded files (mainly because users never start their document by clicking on "New" but instead create the docs locally and then upload them).

The meta data for these documents includes a Yes/No site column called "Urgent" and another site column called "Status".  We need to meet a business requirement that shows the percentage of "Urgent" documents whose status is "Pending".

This is usually simple to do and I described something very much like this at the SharePoint Beagle with lots of screen shots if you’re interested.

In a nutshell, I did the following:

• Create a view on the doc library called "Pending".
• Configure the view to ignore folder structure.
• Create a KPI List.
• Create an indicator in the list that points to the doc lib and that "Pending" view.

This simply does not work.  The KPI shows my target (e.g. five urgent documents) but always shows the actual number of urgent documents as zero.  Paradoxically, if you drill down to the details, it shows the five urgent documents in the list.  I created a very simple scenario with two documents, one in a folder and one not.  Here is the screen shot:

The above screen shot clearly shows there are two documents in the view but the "value" is one.  The "CamlSchema" with blank document Id is in the root folder and the other is in a folder named "84".

It appears to me that even though you specify a view, the KPI doesn’t honor the "show all items without folders" setting and instead, confines itself to the root folder.

If I’m wrong, please drop me a line or leave a comment.

</end>

Subscribe to my blog!