Dapper Multipmapping collection within collections

Question

In Dapper, what's the best way to map and object that contains a list and each of those list items have their own lists?

class Object1 {
    int Object1Id;
    List<Object2> object2s;
}

class Object2 {
    int Object2Id;
    List<Object3> object3s;
}

class Object3 {
    int Object3Id;
}

Here's the SQL I wanted to use

SELECT *
  FROM [Object1] o1
  left join [Object2] o2 on o1.Object1Id = o2.Object1Id
  left join [Object3] o3 on o2.Object2Id = o3.Object2Id

Using the Left joins because it's possible Object1 has no Object2's and Object2's may not have any Object3's

I thought about using

connection.Query<Object1, Object2, Object3, Object1>

but keeping track of a dictionary within a dictionary to add them to the list seems like a lot of lines of code.

Would I be better off using QueryMultiple?

Answer 1

If you're using SQL Server 2016 or above I would go with JSON. You can return the data as a JSON document, like this:

USE tempdb
GO

DROP TABLE IF EXISTS dbo.[Object1] 
DROP TABLE IF EXISTS dbo.[Object2] 
DROP TABLE IF EXISTS dbo.[Object3] 
GO

CREATE TABLE dbo.[Object1] ([Object1Id] INT PRIMARY KEY, [Value1] NVARCHAR(100));
CREATE TABLE dbo.[Object2] ([Object2Id] INT PRIMARY KEY, [Object1Id] INT, [Value2] NVARCHAR(100));
CREATE TABLE dbo.[Object3] ([Object3Id] INT PRIMARY KEY, [Object2Id] INT, [Value3] NVARCHAR(100));
GO

INSERT INTO dbo.[Object1] VALUES (1, 'Object 1 Value 1')
INSERT INTO dbo.[Object1] VALUES (2, 'Object 1 Value 2')
INSERT INTO dbo.[Object1] VALUES (3, 'Object 1 Value 3')
GO                        

INSERT INTO dbo.[Object2] VALUES (10, 1, 'Object 2 Value 1')
INSERT INTO dbo.[Object2] VALUES (20, 1, 'Object 2 Value 2')
INSERT INTO dbo.[Object2] VALUES (30, 2, 'Object 2 Value 3')
GO                       

INSERT INTO dbo.[Object3] VALUES (100, 10, 'Object 3 Value 1')
INSERT INTO dbo.[Object3] VALUES (200, 10, 'Object 3 Value 2')
INSERT INTO dbo.[Object3] VALUES  (300, 30, 'Object 3 Value 3')
GO

SELECT 
    *
FROM 
    [Object1] 
LEFT JOIN
    [Object2] AS Object2s ON [Object1].Object1Id = [Object2s].Object1Id
LEFT JOIN 
    [Object3] AS Object3s ON [Object2s].Object2Id = [Object3s].Object2Id
WHERE
    [Object1].Object1Id = 1
FOR
    JSON AUTO,WITHOUT_ARRAY_WRAPPER

With a JSON as result it is easy to then deserialize it in a class you want:

class Program
{
    class Object1 {
        public int Object1Id;
        public List<Object2> Object2s;
    }

    class Object2 {
        public int Object2Id;
        public List<Object3> Object3s;
    }

    class Object3 {
        public int Object3Id;
    }

    static void Main(string[] args)
    {
        using(var conn = new SqlConnection("Data Source=localhost; Initial Catalog=tempdb; Integrated Security=SSPI"))
        {
            var json = conn.ExecuteScalar<string>(@"               
                SELECT 
                    *
                FROM 
                    [Object1] 
                LEFT JOIN
                    [Object2] AS Object2s ON [Object1].Object1Id = [Object2s].Object1Id
                LEFT JOIN 
                    [Object3] AS Object3s ON [Object2s].Object2Id = [Object3s].Object2Id
                WHERE
                    [Object1].Object1Id = 1
                FOR
                    JSON AUTO,WITHOUT_ARRAY_WRAPPER
            ");                                

            var result = JsonConvert.DeserializeObject<Object1>(json.Replace("\"", "'"));

            Console.WriteLine(result.Object2s.Count);
        }
    }
}

Performances are also very good.

Answer 2

You're absolutely right : this kind of parent-child handling isn't a scenario that dapper handles very well right now. It would be nice if it did, but it simply hasn't been the top priority to do something about it.

You have a choice of two inelegant solutions. Personally I think the QueryMultiple approach is proabbly cleaner.