Working with Entity Framework 4 and MS SQL Spatial Types

I’ve been involved in a project at work recently that required the ability to maintain records that contained spatial information. The ORM of choice for the project was Entity Framework 4. During the build I ran into several problems. In this post I’ll detail what barriers I found, the options available to work around them and the solution I chose.

What we intend to build here is an editor, so first things first, let’s create a table to hold our data.

CREATE TABLE [dbo].[Dinners](
	[DinnerId] [int] NOT NULL,
	[Title] [nvarchar](50) NOT NULL,
	[Location] [geography] NOT NULL

The next step is to create a VS2010 project, to which we’ll add a class to model our Dinners table.

using System;

namespace JB.DinnersWithSpatial.Core
	public class Dinner
		public int DinnerId { get; set; }
		public string Title { get; set; }
		public byte[] Location { get; set; }

Next we’ll add an ADO.NET Entity Data Model called DinnersModel.edmx. This will kick start a wizard that will ask you about database connectivity and which entities from the database you want to include in your model. Be sure to grab the Dinners table. After you click finish the model should be visible, but wait where is the Location property? Checking the warnings shows the following:

warning 6005: The data type 'geography' is not supported; the column
'Location' in table 'JBExamples.dbo.Dinners' was excluded.

Well that’s just not cricket! So EF4 doesn’t support spatial types, time for a workaround. After searching the deepest darkest corners of the internet we came across a simple solution using a view. So now we need to create a view to represent our Dinners table.

CREATE VIEW [dbo].[vw_DinnersAsBinary]
SELECT [DinnerId], [Title], CONVERT(VARBINARY(MAX), [Location], 0) AS [Location]
FROM [dbo].[Dinners]

With that done we need to edit the DinnersModel.edmx. The entity for Dinners should be replaced with the View we just created. Once the entity representing the View is added we need to open its properties and rename it to Dinner (the Entity Set Name should be Dinners). After saving you may notice that there is a new message in the errors list:

The table/view 'dbo.vw_DinnersAsBinary' does not have a primary key defined.
The key has been inferred and the definition was created as a read-only

EF4 is telling us that since Views don’t support primary keys it has helped us by creating one, however the one it defines is not correct. The only property that should have Entity Key setting equal to true is DinnerId, so lets update the others. With that done we can move onto the repository plumbing.

Setting up for POCOs

In order to get our clean class to be used we need to turn off the code generation that is being done. To do this view the properties of DinnersModel.edmx and remove the value for Custom Tool. Once deleted the .cs file should disappear.

Now when we try to insert a new record we get another error.

Update or insert of view or function 'dbo.vw_DinnersAsBinary' failed because
it contains a derived or constant field.

This rabbit hole is getting deeper and deeper and we’ve just hit a fork in the tunnel. We have two options here. Either create CRUD stored procedures or use Instead Of triggers.

Approach 1 - Stored Procedures

This approach means creating the CRUD procedures, opening your mapping file and adding those procedures to it. Once that is done you need to go into your repository implementation (or context) and modify the code to call the procedures instead.

I’m not a huge fan of this approach because there is the need to map extra objects from the DB and extra code to write.

Approach 2 - Instead Of Triggers

The second approach is to use Instead Of triggers. These types of trigger allow you to intercept an action. So taking our insert statement above we could employ an Instead Of trigger to intercept the action from the view and perform the insert correctly.

First we need to modify the edmx so that EF4 thinks the view is a table. To do this we need to open the edmx with the XML editor. Somewhere near the top of the file will be the definition of vw_DinnersAsBinary, you’ll know if because it will contain a DefiningQuery element. Here’s what mine looked like.

<EntitySet Name="vw_DinnersAsBinary" EntityType="DinnersModel.Store.vw_DinnersAsBinary" store:Type="Views" store:Schema="dbo" store:Name="vw_DinnersAsBinary">
		[vw_DinnersAsBinary].[DinnerId] AS [DinnerId],
		[vw_DinnersAsBinary].[Title] AS [Title],
		[vw_DinnersAsBinary].[Location] AS [Location]
		FROM [dbo].[vw_DinnersAsBinary] AS [vw_DinnersAsBinary]

After making the changes it now looks like this

<EntitySet Name="vw_DinnersAsBinary" EntityType="DinnersModel.Store.vw_DinnersAsBinary" store:Type="Tables" Schema="dbo" />

Now we add our INSERT trigger. Note the use of the STPointFromWKB method is used to convert the incoming binary stream into a SQL spatial object.

CREATE TRIGGER [dbo].[trg_vw_DinnersAsBinary_Insert]
ON [dbo].[vw_DinnersAsBinary]
	INSERT INTO [dbo].[Dinners](DinnerId, Title, Location)
	SELECT DinnerId, Title, geography::STPointFromWKB((SELECT Location FROM inserted), 4326)
	FROM inserted

I like this approach because there is no need to map extra objects from the DB and no extra code to write. Supporting other SRIDs might prove an issue, however I don’t think solving that is too hard.

Workarounds suck

The down side to all of this is that without native support for spatial types in EF4 you cannot use the spatial querying that MS SQL provides and when MS does support spatial types in EF you’ve got code changes to make (if you decide to).

If anyone knows any other ways of doing this please post, I’d love to hear about them.