TypeSql

The TypeSql project is hosted on GitHub

TypeSql is a compiler. It compiles annotated SQL into strongly-typed data-access-objects.

The inspiration was TypeScript. But don't try to take the analogy too far. TypeScript takes annotated javascript and compiles it into plain javascript. Its primary purpose is verification. TypeSQL takes annotated SQL and compiles it into a different, general-purpose programming language (C# only, for the moment). Our primary purpose is for you to write less code. Less code means less bugs and more functionality. It means you can ask for more money and go home earlier. Your family will love you more, and you will live longer.

Are you talking to me?

Do you believe that a strongly-typed, object-orientated programming language is the best way to write applications?

Do you believe that SQL is still the best language for querying relational databases?

Good for you. Here at TypeSql HQ, we feel the same.

Bridging the O-O relational-databases divide has been attempted many times, in many ways. Frankly, we don't love any of them. Here's our wish-list:

• We want to use SQL. Not Linq, not Hibernate Query Language. SQL.
• We want the data returned to be typed. Call us old-fashioned if you will, but we like types. It's a little piece of certainty in an uncertain world.
• We want getting our data to be minimum friction. We don't want to repeat ourselves. The typical data-retrieval ceremony (connection, command, reader, casting, etc) is tedious, and surface-area for bugs.

The closest we've found are micro-ORMs (eg. Dapper). Our issue with using these is how to manage the classes created to hold query results (we like types, remember). In a large application, it is not uncommon for many queries to differ only slightly in the results they return. Which tends to lead to either:

Class explosion

UserUserWithAddressUserWithImageUserAccounts

or

Super DTO!

One class that contains all possible properties that can be returned. Consumers of the class accessing, say user.Address.Streetthen play Russian-Roulette, hoping that the data-access operation they just performed resulted in the Address property being populated. Otherwise, its that most dreaded of exceptions.

So that's where we're coming from.

I already have [insert ORM of choice], why do I need TypeSql?

Maybe you don't. If your team is happy and productive using ORM X, then carry on. Full-service ORMs like Entity Framework andNHibernate are cruise-ships; TypeSQL is a surfboard. We just want to make travelling through the water a little easier and more fun. They do their best to hide the fact the ocean even exists (until the abstraction leaks, and sinks the ship...forgive me). But seriously, we approach the problem from different angles. TypeSql knows nothing about your database schema, and why should it? We care about the shape of the data you are retrieving, not the shape the data is stored. That is the fundamental difference. EF & friends map things that live in the database (like tables, views, and stored-procedures) to objects. They then attempt to translate interactions with those objects (ie. accessing properties and calling methods) into the appropriate SQL (this often happens at runtime). With TypeSQL, youwrite your SQL (plus a few annotations), and at compile time the SQL is translated to data-access objects that represent your query(not stuff that lives in the database). EF & friends also deal with concerns like caching and object-identity; TypeSql does not.

Show me some code!

Here's how it works. We have some data in our relational DB. We want to query it. We write our parameterised SQL ( let's suppose its in a file called 'TurtlesByColor.sql' ).

SELECT Id, Name FROM TeenageMutantNinjaTurtlesWHERE SashColor= @Color 

Now we annotate it with types.

SELECT Id:int, Name:string FROM TeenageMutantNinjaTurtlesWHERE SashColor= @Color:string

Compile with TypeSql. This produces a class called TurtlesByColor. TurtlesByColor has two constructors (in C#):

TurtlesByColor(string connectionStringName)TurtlesByColor(IDbConnection connection, IDbTransaction transaction=null) 

And a method:

IEnumerable<TurtlesByColorResult> Execute(string color, bool buffered=true); 

The 'TurtlesByColorResult' class looks like:

public partial class TurtlesByColorResult{    public TurtlesByColorResult(int id, string name)    {        Id = id;        Name = name;    }    public int Id { get; private set; }     public string Name{ get; private set; }}

Accessing your data then looks like:

IEnumerable<TurtlesByColorResult> redTurtles = new TurtlesByColor("turtlesDb").Execute("Red");

Availability

Currently, TypeSQL is available in C# flavor only. Hopefully, we will be coming soon to a language near you.

Installation

Currently, the supported method for using TypeSql is via a VisualStudio extension in Visual Studio 2010 or Visual Studio 2012.

• Install the TypeSql Visual Studio extension. This can be done via Tools -> Extensions and Updates in Visual Studio.
• Install the TypeSql nuget package.

Type that SQL baby!

• Add a new text file to your project; you can give it whatever extension you like (eg: OverdueBooks.tsql).
• Enter your type-annotated SQL.
• Right click on the file in the Visual Studio Solution Explorer, click 'Properties', and set the 'Custom Tool' property to 'TypeSql'.
• Two dependent files should be added under your TypeSql file: One will contain the source code of your generated data-access class; the other will contain the raw, unadorned SQL (this is just for convenience if you want to execute the SQL directly).

Its not you, its me

This is a work-in-progress. There are an infinite number of scenarios we have not thought of. We welcome your feedback. But please be gentle, our egos are fragile and we cry easily.

Performance

Internally, TypeSql uses Dapper for data-access. Its performance is effectively identical, which is itself effectively identical to raw ADO.NET. Checkout the benchmarks.

Credits

TypeSql stands on the shoulders of two brilliant projects.

ANTLR

If you need to parse text, ANTLR is your tool. Terence Parr has made language-recognition possible for mere mortals. And thanks to Sam Harwell for the C# version.

Dapper

Dapper has been called a micro-ORM. Whatever you call it, Sam Saffron and Marc Gravell have created a super-useful component.

USAGE

Outputs

Just append the items in your SELECT list with the types you expect them to be. e.g

SELECT Id:int, Name:stringFROM Smurfs

Aggregates

Yeah, you can do:

SELECT COUNT(*) AS SmurfCount:intFROM Smurfs

Inputs

Type your input parameters like so:

SELECT Name:stringFROM SmurfsWHERE Id= @SmurfId:int

Nullable types

Sure.

SELECT BirthDate:DateTime?FROM Smurfs

Namespaces

The USING statement is used to import namespaces. See Enums section below for an example.

Enums

Yup. Say you have:

namespace SmurfSpace {    public enum SmurfSex {        Female,        Male    }}

Then you can do:

USING SmurfSpace    SELECT Id:int, Gender:SmurfSex