Menu Menu
Building reusable SQL queries with SQL query builder

Building reusable SQL queries with SQL query builder

This article describes one of the approaches to easily create and manage your SQL queries in ASP.NET Core applications. Examples given here were created in C#, but you can use any other programming language compatible with ASP.NET Core.

The full source code is located at the GitHub repository and there is also an ASP.NET Core NuGet package.

Introduction

Each time we need to create a data layer for an ASP.NET Core application, which is backed up by an SQL database, we're faced with a question how to design our SQL queries, to make them easy to use, maintain and extend. Of course, it's fairly easy to just implement SQL queries in ASP.NET. But, once implemented, the process of extending and maintaining those queries can quickly become difficult, especially if we implement those queries as simple string constants, which opens the door to SQL injections.

What we would like to implement, ideally, is something that is:

  • easy to maintain (which means we can easily add/delete/update/rename tables/columns)
  • easy to extend (which means we can quickly create new SQL queries on top of existing ones)
  • immune to security threats (like SQL injections)

Many of us usually decide to use an object-relational mapping (ORM) solution, like EntityFramework, NPoco, LINQ to SQL or similar, to deal with the data layer as a whole. This approach is pretty much a standard these days and this article does not offer an alternative to those solutions, but rather tries to make things even easier.

Note that SQL query builder, described in this article, is a standalone solution, without any dependency, and it is NOT a requirement to have an ORM solution in order to use it.

What is SQL query builder

Working on a web API project recently, I had an idea to create a small library, which could help us create SQL queries that are maintainable, reusable and easy to use. Part of the inspiration for this library was the .NET's Language Integrated Query facility (LINQ) and the elegance with which LINQ queries were extended.

The idea was to create reusable SQL queries and build any new queries on top of the existing ones. Ideally, we should be able to create a basic query, like SELECT * FROM [Table] and build any new queries on top of that basic query, just by extending it, just like in a LINQ query, where we start from an enumeration and keep extending the query until we shape it into what we need. For example:

IEnumerable<User> allUsers = GetAll<User>();

var activeGroups = allUsers
                      .Where(user => user.IsActive)
                      .Select(user => user.Id)
                      .Distinct()
                      .ToArray();

We start with an enumeration of all users, using GetAll<User>(), which we can reuse as many times as we need. Then we extend our query by adding a filter (Where()), mapping the result to the list of user ids (Select()), reducing the result further to the distinct enumeration of user ids. After we crafted our query, we materialize it with ToArray().

The SQL query builder implements the similar behavior, helping us to create our queries in a similar fashion as LINQ queries, materializing them in the end as simple SqlQuery objects which consist of a string (the actual SQL command) and an object array (the SQL parameters):

public class SqlQuery
{
    public string Command { get; }
    public object[] Parameters { get; }
}

That approach helps us create parameterized SQL queries to avoid being a victim of an SQL injection attack.

Usage

Let's take a look at some examples, which would explain it better, hopefully. First, we create an SqlQueryBuilder:

var builder = new SqlQueryBuilder();

which we'll use in all the following examples. For example, let's select everything from a table User:

var query = builder
    .From<User>()
    .SelectAll();

var sql = query.ToSqlQuery();

In our sql variable, we'll have an SqlQuery object, with the Command property set to:

SELECT *
FROM [User]

And if we extend that query, by adding a filter to our result set, using a WHERE clause:

var name = "John";

var newQuery = query
    .Where(user => $"{user.Name} LIKE '%' + @0 + '%'", name);

var newSql = newQuery.ToSqlQuery();

that will create an SqlQuery, with the Command property:

SELECT *
FROM [User]
WHERE ([User].[Name] LIKE '%' + @0 + '%')

and its Parameters property:

@0 = "John"

Note that we made use of the String.Format() method in order to leverage the help of IntelliSense, to help us write queries more conveniently. We also used the "string interpolation" feature of the C# language, to make things even easier.

The string $"{user.Name} LIKE '%' + @0 + '%'" is the same as string.Format("{0} LIKE '%' + @0 + '%'", user.Name).

The SqlQueryBuilder will parse this construct and enumerate all the classes and properties used and will map them to the appropriate tables/columns of the underlying SQL database. The default convention is to use the same naming for the C# classes and SQL tables, as well as the same naming for the C# properties on those classes and SQL columns of those tables. We can, of course, customize this mapping by providing our own mapper implementations (see section "Custom table/column names mapping").

Once we have our query built, we can use it, for example, directly using System.Data.SqlClient.SqlConnection, like this:

using (var connection = new SqlConnection(ConnectionString))
{
    IDbCommand cmd = connection.CreateCommand();
    cmd.CommandText = query.Command;

    var param = cmd.CreateParameter();
    param.ParameterName = "@0";
    param.Value = query.Parameters[0];

    connection.Open();
    ...
}

Of course, this is just a basic example how can we use the query that SqlQueryBuilder has generated for us. But we can also use our favorite ORM solution instead.

Reusing queries

If we want to create a simple SQL query and reuse it to construct more complex new queries, we could do it easily, writing something like this:

var name = "John";
var userGroupIds = new[] { 1, 2, 3 };

var allUsersQuery = builder
    .From<User>()
    .SelectAll();

var usersByNameQuery = allUsersQuery
    .Where(user => $"{user.Name} LIKE '%' + @0 + '%'", name);

var usersByNameExtendedQuery = usersByNameQuery
    .InnerJoin<Address>((user, address) => $"{user.AddressId} = {address.Id}")
    .InnerJoin<UserGroup>((user, address, userGroup) => $"{user.UserGroupId} = {userGroup.Id}")
    .Where((user, address, userGroup) => $"{user.UserGroupId} IN (@0)", userGroupIds)
    .Select((user, address, userGroup) => $"{user.Id}, {user.Name}, {user.Age}");

var usersByNameSql = usersByNameQuery.ToSqlQuery();
var usersByNameExtendedSql = usersByNameExtendedQuery.ToSqlQuery();

we would end up with 2 SqlQuery objects. The first one, usersByNameSql, would have a Command:

SELECT *
FROM [User]
WHERE ([User].[Name] LIKE '%' + @0 + '%')

and its Parameters set to:

@0 = "John"

The second SqlQuery, usersByNameExtendedSql, would have Command/Parameters properties set to:

SELECT [User].[Id], [User].[Name], [User].[Age]
FROM [User]
INNER JOIN [Address] ON [User].[AddressId] = [Address].[Id]
INNER JOIN [UserGroup] ON [User].[UserGroupId] = [UserGroup].[Id]
WHERE (([User].[Name] LIKE '%' + @0 + '%') AND ([User].[UserGroupId] IN (@1,@2,@3)))

@0 = "John"
@1 = 1
@2 = 2
@3 = 3

Note that, in the usersByNameExtendedSql, the first "SELECT *" got replaced with the second "SELECT [User].[Id]...", and all the WHERE clauses got merged.

A couple of more complex queries

We can create even more complex queries, expanding the list of joined tables with multiple WHERE statements, later combined into one:

var name = "John";
var userGroupIds = new[] { 1, 2, 3 };

var baseQuery = builder
    .From<User>()
    .Where(user => $"{user.Name} LIKE '%' + @0 + '%'", name)
    .SelectAll();

var joinQuery = baseQuery
    .InnerJoin<Address>((user, address) => $"{user.AddressId} = {address.Id}")
    .Where((user, address) => $"{user.UserGroupId} = 1")
    .InnerJoin<UserGroup>((user, address, userGroup) => $"{user.UserGroupId} = {userGroup.Id}")
    .Where((user, address, userGroup) => $"{user.UserGroupId} IN (@0)", userGroupIds)
    .Select((user, address, userGroup) => $"{user.Id}, {user.Name}, {user.Age}");

var baseSql = baseQuery.ToSqlQuery();
var joinSql = joinQuery.ToSqlQuery();

which would result in 2 SqlQuery objects. The first one, baseSql would have the Command/Parameters properties like:

SELECT *
FROM [User]
WHERE ([User].[Name] LIKE '%' + @0 + '%')

@0 = "John"

and the second query, joinSql, would look like:

SELECT [User].[Id], [User].[Name], [User].[Age]
FROM [User]
INNER JOIN [Address] ON [User].[AddressId] = [Address].[Id]
INNER JOIN [UserGroup] ON [User].[UserGroupId] = [UserGroup].[Id]
WHERE ((([User].[Name] LIKE '%' + @0 + '%') AND ([User].[UserGroupId] = 1)) AND ([User].[UserGroupId] IN (@1,@2,@3)))

@0 = "John"
@1 = 1
@2 = 2
@3 = 3

INSERT / UPDATE made easy

In order to create an INSERT SQL statement, it is just enough to write something like this:

var age = 10;
var addressId = 1;
var name = "John";

var insertSql = builder
    .Insert<User>(user => $"{user.Age}, {user.AddressId}, {user.Name}", age, addressId, name)
    .ToSqlQuery();

which would produce this SqlQuery as a result:

INSERT INTO [User] ([User].[Age], [User].[AddressId], [User].[Name])
VALUES (@0, @1, @2)

@0 = 10
@1 = 1
@2 = "John"

In order to create an INSERT SQL statement with multiple rows of data at once, we would write:

var users = new[]
{
    new User(Name: "John", Age: 10, AddressId: 1),
    new User(Name: "Jane", Age: 20, AddressId: 2),
    new User(Name: "Smith", Age: 30, AddressId: 3),
};

var parameters = users
    .Select(u => new object[] { u.Age, u.AddressId, u.Name })
    .ToArray();

var insertMultipleSql = builder
    .InsertMultiple<User>(user => $"{user.Age}, {user.AddressId}, {user.Name}", parameters)
    .ToSqlQuery();

which would create an SqlQuery like this:

INSERT INTO [User] ([User].[Age], [User].[AddressId], [User].[Name])
VALUES (@0, @1, @2), (@3, @4, @5), (@6, @7, @8)

@0 = 10
@1 = 1
@2 = "John"
@3 = 20
@4 = 2
@5 = "Jane"
@6 = 30
@7 = 3
@8 = "Smith"

For the UPDATE statement, it's quite similar:

var age = 10;
var addressId = 1;
var name = "John";

var updateByNameSql = builder
    .Update<User>(user => $"{user.Age} = @0, {user.AddressId} = @1", age, addressId)
    .Where(user => $"{user.Name} LIKE '%' + @0 + '%'", name)
    .ToSqlQuery();

and the result would be as expected:

UPDATE [User]
SET [User].[Age] = @0, [User].[AddressId] = @1
WHERE ([User].[Name] LIKE '%' + @2 + '%')

@0 = 10
@1 = 1
@2 = "John"

Note that we don't have to keep track of the last parameter index used in previous statements/clauses, because each new statement/clause will start enumerating its parameters from a zero-based index.

That's why, in the previous query in the Where() method, we didn't use the index "@2" for the user's name placeholder, but we rather used the parameter with index "@0".

Custom table/column names mapping

If we have a scenario where our table/column names are not exactly "one-to-one" mapped to our classes/properties, we can specify our custom table/column mappers, when creating a new instance of an SqlQueryBuilder.

For example, if we create our ITableNameResolver like this:

public class NPocoTableNameResolver : ITableNameResolver
{
    private readonly IDatabase _database;

    public NPocoTableNameResolver(IDatabase database)
    {
        _database = database ?? throw new ArgumentNullException(nameof(database));
    }

    public string Resolve(Type type)
    {
        if (type == null) throw new ArgumentNullException(nameof(type));

        var tableName = _database
            .PocoDataFactory
            .ForType(type)
            .TableInfo
            .TableName;

        return $"[{tableName}]";
    }
}

and our IColumnNameResolver like this:

public class NPocoColumnNameResolver : IColumnNameResolver
{
    private readonly IDatabase _database;

    public NPocoColumnNameResolver(IDatabase database)
    {
        _database = database ?? throw new ArgumentNullException(nameof(database));
    }

    public string Resolve(Type type, string memberName)
    {
        if (type == null) throw new ArgumentNullException(nameof(type));
        if (memberName == null) throw new ArgumentNullException(nameof(memberName));

        var data = _database.PocoDataFactory.ForType(type);
        var tableName = data.TableInfo.TableName;
        var columnName = data
            .Members
            .First(x => x.Name == memberName)
            .PocoColumn
            .ColumnName;

        return $"[{tableName}].[{columnName}]";
    }
}

then we can make use of the NPoco's mapping feature and have even more customized SQL strings. We just need to create an instance of an SqlQueryBuilder like this:

var db = new NPoco.Database("connectionString");
var tableNameResolver = new NPocoTableNameResolver(db);
var columnNameResolver = new NPocoColumnNameResolver(db);

var builder = new SqlQueryBuilder(tableNameResolver, columnNameResolver);

and we can reuse all the examples given here, in this document, the same way.

Feedback

Please provide your comments/feedback, if you like, via GitHub project page.

Latest blog posts
ITkonekt 2019 Overview
Front-End Day 2019: Our Impressions
Workshop with Uncle Bob
International Umbraco Conference in Novi Sad: Umbraco Roadshow 2019
How To Send an Email with Attachments Using EPiServer Forms