PartitionKey and RowKey can't be same property

[dvdstelt]

It's not unusual for us to have the same value for PartitionKey and RowKey. In our EntityMap we provide the same property for ParititionKey and RowKey. This throws an exception.

[dtretyakov]

Could you briefly describe what would you plan to reach when using same property as PK & RK?

[dvdstelt]

We have a lot of unique entities, which each have just a single row in TableStorage. Instead of relying on a fixed partitionkey like "AllCompanies" or something and have thousands and thousands of rows in a single partition, we'd like Azure to possibly partition all keys over unique partitions.

We just specify a single property in our model and specify it as both PartitionKey and RowKey.

[dtretyakov]

Ok, in this case just use a PartitionKeyAttribute without RowKeyAttribute.

[dvdstelt]

This also works with EntityTypeMap ?

[tdietrich513]

No, EntityTypeMap seems to throw an exception when either PartitionKey or RowKey is not set.

Setup:

public class KeyOnly
{
    public string Key { get; set; }
    public string Data { get; set; }
}

public class KeyOnlyMap : EntityTypeMap<KeyOnly>
{
    public KeyOnlyMap()
    {
        PartitionKey(o => o.Key);
    }
}

Test:

[TestFixture]
public class ASETests
{
    [Test]
    public void CanUseKeyOnly()
    {
        EntityTypeMap.RegisterAssembly(GetType().Assembly);
        var storageAccount = CloudStorageAccount.DevelopmentStorageAccount;
        var tableClient = storageAccount.CreateCloudTableClient();
        var keyOnlyTable = new TableSet<KeyOnly>(tableClient);

        keyOnlyTable.CreateIfNotExists();

        var newEntity = new KeyOnly()
        {
            Key = "1",
            Data = "SomeData"
        };

        keyOnlyTable.Add(newEntity);
    }
}

Error:
System.InvalidOperationException : PartitionKey or RowKey attribute should be defined for type 'AzureStorageExtensionsTesting.KeyOnly'.

The issue seems to be from this method in the EntityTypeMap class:

    private void CheckEntityMap()
    {
        // Check whether entity's composite key completely defined
        if (!_nameChanges.ContainsKey(PartitionKeyPropertyName) && !_nameChanges.ContainsValue(RowKeyPropertyName))
        {
            string message = string.Format(Resources.EntityTypeDataMissingKey, _entityType);
            throw new InvalidOperationException(message);
        }
    }

Which I'm not really understanding entirely (shouldn't it be ContainsValue in both cases? and || rather than &&?), but this seems to be by design, as there is a unit test to make sure an InvalidOperation is thrown for entities without a composite key:

    [Fact]
    public void CreateEntityTypeDataWithoutCompositeKey()
    {
        // Arrange
        EntityTypeData<EntityWithoutCompositeKey> typeData = null;

        // Act
        Assert.Throws<InvalidOperationException>(() => { typeData = new EntityTypeData<EntityWithoutCompositeKey>(); });

        // Assert
        Assert.Null(typeData);
    }

However you can just the attribute based mapping without a RowKey and it works just fine.

[dtretyakov]

But actually it's a bug which should be fixed.

[dtretyakov]

It was fixed in version 1.0.2