What are types and why do they matter?

Luna can process any kind of data. It can compute a car's velocity, reply to a user's questions via a chat-like interface or process all photos from your last trip to San Francisco.

Some operations, however, only make sense with very specific kinds of data. You can multiply numbers, however it's hard to imagine what the result of a cake multiplication would be. Other operations allow some degree of freedom. Reversing, for example, works with any list with no regard to what its individual elements are. Yet other operations provide more complex constraints. Summing a list will work with any non-empty list containing arithmetic elements only, but we don't really care about the exact structure of these elements. They may be integers, fractions or even matrices; the only requirement is having an addition operation defined.

In this chapter we introduce the type system, a way Luna handles such constraints. You will learn how types are defined, how to check a type of a given value and what happens when the constraints could not be resolved. The philosophy behind the type system is to provide as much safety and useful hints as possible without restricting your expressive powers.

How do I see types?

The type system in Luna is fundamentally different from any other type system used in programming languages nowadays. It was designed from scratch to be invisible until you need it, yet fully aware of the structure of the data being processed. You will never need to manually specify the types in Luna. The type inference engine computes every single type behind the scenes automatically for you. This computation ensures that every operation will be performed safely, without any contradictory properties arising.

Since the types are usually not expressed in code, we need a way to show them to you. You have probably already noticed that the nodes in our visual editor come in different, seemingly random, colors. Those colors are just one of the ways the visual editor communicates types. Each of the ports on the node is assigned a color corresponding to its type. All the connections are also colored; the colors correspond to the type of data "being sent" via this connection.

Colors provide a quick visual cue to tell if two values have the same or different types, but they are not enough. Just imagine referring to them like "the text "hello world!" has type magenta". That's why the colors serve only as visual aids, but the real types have a text representation. If you hover above a node, the real types are revealed. Each of the node inputs and outputs displays its corresponding type.

In the above image you can see the actual type names and thus understand the behavior of the node:

1. The nodes number2 and number3 don't have inputs – they are both constants of type Real,

2. The node sum3 has two Real inputs and a Real output – it transforms two Real numbers into another Real number,

3. The node just2 has a Real input and a Maybe Real output – it transforms a Real number into a Maybe Real value.

Oh, and the type of "hello world!" is not magenta. It's Text.

What happens when I mess up?

Besides providing a better understanding of program behavior, types have an even more important role to play: they make sure everything will work smoothly and safely. They are crucial in preventing many errors that may accidentally arise during development. So what happens when you make a mistake?

Let's try doing something clearly nonsensical: adding a text and a number. This is how Luna reacts:

Luna's typechecker has found out about our mistake. The + function works only with arguments of the same type, so it complained that Real and Text are not the same and we cannot safely proceed.