DEV Community

Muhammad Azeez
Muhammad Azeez

Posted on • Updated on • Originally published at mazeez.dev

How do Object Relational Mappers (like Entity Framework) work?

When connecting an app or website to a database, most of the time you're querying the database and mapping the result to strongly typed objects. If you've been using an ORM like Enitity Framework or a Micro-ORM like Dapper, they handle the mapping of the objects and translation of the queries between C# and SQL.

This article is a part of the 3rd annual C# advent calendar. I originally published it on my blog.

The idea of ORMs is simple, since most of the code mapping code is mechanical and repetitive, it's a very good candidate for automation. Lets assume we have a Person table with this schema:

Column DataType
Id int
Name nvarchar(50)
Height float
Birthdate datetime

And we have a corresponding Person class:

class Person
{
   public int Id { get; set; }
   public string Name { get; set; }
   public double Height { get; set; }
   public DateTime Birthdate { get; set; }
}

Here is the C# code if we manually retrieve the data and map it to a list of the Person class:

public async Task<List<Person>> GetPeople()
{
    var list = new List<Person>();

    var sql = "SELECT Id, Name, Height, Birthdate From Person";

    using (var connection = new SqlConnection(ConnectionString))
    using (var command = new SqlCommand(sql, connection))
    {
        await connection.OpenAsync();

        using (var reader = await command.ExecuteReaderAsync())
        {
            while (await reader.ReadAsync())
            {
                var person = new Person
                {
                    Id = (int)reader["Id"],
                    Name = (string)reader["Name"],
                    Height = (double)reader["Height"],
                    Birthdate = (DateTime)reader["Birthdate"],
                };

                list.Add(person);
            }
        }
    }

    return list;
}

The SQL query and the mapping of columns to properties are the only two things that differ from one table/class to another. If only we can get the list of properties of a class and their types... Well, it turns out that we can!

System.Reflection.Type represents a .NET type. It has a lot of metadata about the type. We can get the list of the properties of a type by using the GetProperties method:

PropertyInfo[] properties = typeof(Person).GetProperties();

The typeof keyword is used to get the type of a class at compile time. The PropertyInfo class has many useful properties and methods, but we are only interested in these:

Member Type Description
Name string Gets the name for a property.
GetMethod MethodInfo Gets the get method of a property.
SetMethod MethodInfo Gets the set method of a property.

For example, if we have this code:

var person = new Person
{
    Name = "Ahmed",
    Height = 180.4,
    Birthdate = new DateTime(1998, 1, 1),
    Id = 24
};

var heightProperty = typeof(Person).GetProperty(nameof(Person.Height));
heightProperty.SetValue(person, 185);

var properies = typeof(Person).GetProperties();
foreach (var prop in properies)
{
    Console.WriteLine($"{prop.Name} => {prop.GetValue(person)}");
}

This will be the output:

Height => 185
Id => 24
Name => Ahmed
Birthdate => 1/1/1998 12:00:00 AM

Note: It's interesting to see that because we fetched the metadata of Height above the loop using GetProperty, the order of returned properties from GetProperties has changed and Height is the first property.

As you can see, the value of height is changed from 180.4 to 185 by using reflection.

The second thing we needed was to generate an SQL query, which is very easy as well:

var columnNames = properties.Select(p => p.Name);
var columns = string.Join(", ", columnNames);

var sql = $"SELECT {columns} FROM {typeof(T).Name}";
Console.WriteLine(sql);

Note: If you are not sure why GetSelectQuery has a <T> at the end, it's because the method is generic. It allows GetSelectQuery<T> to be called with any type the programmer likes. In the example above, we have used it with Person. Generics is a very powerful concept. You can learn more about generics from Jeremy Clark.

This will be the output:

SELECT Id, Name, Height, Birthdate FROM Person

If you take close look at the generated SQL query, it's exactly like our own hand written query. So now we can combine our new found knowledge to write a method that can query the database for any table we like and map the results to c# types:

public async Task<List<T>> GetAll<T>() where T : new()
{
    var list = new List<T>();

    var properties = typeof(T).GetProperties();

    var columnNames = properties.Select(p => p.Name);
    var columns = string.Join(", ", columnNames);

    var sql = $"SELECT {columns} FROM {typeof(T).Name}";

    using (var connection = new SqlConnection(ConnectionString))
    using (var command = new SqlCommand(sql, connection))
    {
        await connection.OpenAsync();

        using (var reader = await command.ExecuteReaderAsync())
        {
            while (await reader.ReadAsync())
            {
                var model = new T();

                foreach (var property in properties)
                {
                    property.SetValue(model, reader[property.Name]);
                }

                list.Add(model);
            }
        }
    }

    return list;
}

Note: the new() constraint makes sure the T has a public constructor that takes no parameters. So that we can write var model = new T();.

And we can use it like so:

List<Person> people = GetAll<Person>();

Pretty neat right?

And here is what Insert and Update would look like:

public async Task Insert<T>(T item)
{
    var properties = typeof(T).GetProperties();

    var columns = string.Join(", ", properties.Select(p => p.Name));
    var columnParameters = string.Join(", ", properties.Select(p => $"@{p.Name}"));

    var sql = $"INSERT INTO {typeof(T).Name} ({columns}) VALUES ({columnParameters})";

    using (var connection = new SqlConnection(ConnectionString))
    using (var command = new SqlCommand(sql, connection))
    {
        await connection.OpenAsync();

        foreach (var property in properties)
        {
            command.Parameters.AddWithValue($"@{property.Name}", property.GetValue(item));
        }

        await command.ExecuteNonQueryAsync();
    }
}

public async Task Update<T>(string idPropertyName, T item)
{
    var properties = typeof(T).GetProperties();

    var columnUpdates = properties.Where(p => p.Name != idPropertyName)
                                  .Select(p => $"{p.Name} = @{p.Name}");
    var columns = string.Join(", ", columnUpdates);

    var sql = $"UPDATE {typeof(T).Name} SET {columns} WHERE {idPropertyName} = @{idPropertyName}";

    using (var connection = new SqlConnection(ConnectionString))
    using (var command = new SqlCommand(sql, connection))
    {
        await connection.OpenAsync();

        foreach (var property in properties)
        {
            command.Parameters.AddWithValue($"@{property.Name}", property.GetValue(item));
        }

        await command.ExecuteNonQueryAsync();
    }
}

As you can see, with a few lines of code we were able to implement a bare minimum Micro-ORM. But there are a lot of features in an ORM. One of them is that ORMs usually support multiple DBMSs, we only support MSSQL with our implementation here (Although for a simple mapper like this one, it's not very hard to support other DBMSs). Another thing would be providing a way to query the database (adding where clauses in the GetAll method). It's usually implemented using ExpressionTrees.

And the way we have implemented, it won't have a very good performance. Some easy tricks to improve performance would be cache the PropertyInfos for each type and reuse it, because reflection is very expensive. Actual ORMs much better optimized both in the c# side and in the way they generate SQL.

The code for this article is available on GitHub.

Disclaimer: The sample codes in this article are for demonstration purposes alone and should NOT be used in production environments.

Top comments (4)

Collapse
 
jamesmh profile image
James Hickey

Well done!

Collapse
 
mhmd_azeez profile image
Muhammad Azeez

Thanks!

Collapse
 
mhmd_azeez profile image
Muhammad Azeez

Thank you!