The Perfect Programming Language
I know a lot of different programming languages, I've even built my own, and as a result, I have some opinions about what makes a programming language good or bad. I don't think that there exists a "perfect language", but there are bits and pieces of existing languages, which when combined, could be really interesting when put together.
Also, when I say "programming language", I will also delve into the compiler of said language, the runtime environment, and so on [1].
Let's begin! In no particular order...
Pattern Matching
If you haven't used pattern matching before, you should give it a try! Languages like Rust, F#, and Python (as of version 3.10) have pattern matching. They allow you to succinctly match based off of the structure of your data, while also destructuring it for use in the case body. For example, in Python:
point = (10, 42)
match point:
case (x, _) if x < 0:
print("x is negative")
case (_, y) if y < 0:
print("y is negative")
case (x, y):
print(f"x is {x}, y is {y}")
case _:
print("point is invalid")
This is a trivial example, but pattern matching can do a whole lot more than what I just showed you. Read this (Python) and this (Rust) for better examples of how pattern matching works in Python and Rust.
Static Types
The jury is in, and static types are cool again. This is evident given the rise in popularity of TypeScript, and type annotations in Python [2]. Using dynamic languages are great, and the added safety of type information makes them not a total pain to use.
TypeScript's type system is especially mature, and is almost an entire programming language in and of itself. Take a look at the Type Challenges repo on GitHub for some examples of what I'm talking about.
In general, better type systems allow for more safety when dealing with your data, and allow for leveraging types to do your validation for you, instead of having to do it all by yourself.
Trailing Types
Although not 100% required, trailing types are, IMO, much more readable compared to traditional ones:
// traditional types
int x = 123; // int x is equal to 123
int f(); // int returned by function f
// trailing types
x: int = 123; // x is an int equal to 123
f: () => int; // f is a function returning an int
As you can (hopefully) tell, the second group looks much more readable: It reads left to right, and the function/field names are both on the left, with they type info to the right.
Functional Programs With Automatic Concurrency
Functional programming is on the rise, along with employing functional programming methodologies
in existing languages. I can't even remember the last time I wrote a class
in JavaScript. In JavaScript, using filter
, map
and reduce
feel so
much better then looping over things with for-loops.
In addition, lots of languages are going immutable by default, such as Rust and F#. Being immutable by default doesn't automatically make a language functional, but it is a step in the right direction.
We are seeing wider adoption of languages like Haskell, Clojure, and F#, but none of them have hit the mainstream. I think we are one step away from closing this gap, and that is:
We must get rid of function colors.
TL;DR, red and blue functions represent async and non-async functions. We have special keywords we use to call these functions, and there are certain restrictions on how you are able to call them and so on.
What would be great (and maybe impossible) is a language which is automatically able to run your program concurrently.
This is really hard. You need to take into account the lifetime of all the variables passed to the function, whether the function has side effects or not, and so on. But, assuming you solve this, you will be able to achieve very fast programs without having to think about async/await/threads. One of the biggest pros of C# is their very mature async/await ecosystem. Python does support async, but I have found it a little bit more clunky to use compared to C#. Having the ability to automatically run long running tasks in the background would be a real game changer, though achieving it might be an impossible feat.
Domain Specific
We have seen the rise and fall of many programming languages. Some try to be general purpose, which means it must generalize to all use cases, which can be hard. Some languages are domain specific, and solve a particular problem very well. I would say that the general purpose programming language field has sorta burned out. We have enough of them already, and they already get the job done for the most part. We don't really need any more.
What we need are small, purpose-built languages that do one thing really well [3]. I would much rather pickup and start using a DSL with only the features that I need, as opposed to a general purpose language which has too many features, and not the ones I need.
Also, an entire language meant to cater a specific domain will (probably) feel a lot better then using an library in an existing language (if done correctly).
Uniform Style
Languages that don't specify a style guide inevitably end up with complex linting/formatting tools, such as C/C++/JavaScript. These languages have a "common" style associated with them, but still allow for a lot of wiggle room.
Tools like black, gofmt, and rustfmt are opinionated formatters that standardizes the look of your code for you. It might not be what you prefer, but having an easy way to format your code based on an agreed upon format is really nice.
The less time we spend on configuring linters, the more time we have to actually write code!
Elegant
This is very subjective, and hard to define, but a well written program should reward you with something which is elegant and easy to read, easy to understand, distilled to its truest form.
Fin
That's it! I am probably forgot a lot of important categories, but these are the ones I came up with off the top of my head.
I don't think a language can exist which has all of these features, but I can dream, can't I?
[1]: Languages don't exist in a vacuum, they exist amongst their compilers, the environments they run in, and the use cases people find for the language.
[2]: Obviously I am excluding popular languages which are already statically typed, such as C++, C#, Java, and so on.
[3]: See also, Unix Philosophy.