Programming languages are too slow! I’m not talking about execution speed—I’m talking about evolution speed. Programmers are always building new libraries and embedded DSLs, but the host programming language—particularly its type system—doesn’t understand the domain-specific aspects of these things.
...The Tao gave birth to machine language. Machine language gave birth to the assembler.
The assembler gave birth to the compiler. Now there are ten thousand languages.
Each language has its purpose, however humble. Each language expresses the Yin and Yang of software. Each language has its place within the Tao.
But do not program in COBOL if you can avoid it.
I probably hate writing code in your favorite programming language, whatever it may be. This is because I get frustrated by basically all of the top 10 languages you’ll find listed anywhere for various reasons. Do I hate programming in Python? You bet I do. “But it’s Python! Python is the best programming language on earth!” I can hear you say. I grant that it has its place. Python wins because its ecosystem of libraries is so huge and because there are so many resources for new users. It’s also garbage collected, which means memory safety is not an issue. It the current hot thing, because there is so much support for machine learning in Python.
...Chorex is a brand-new Elixir library for choreographic programming [3]: Chorex provides a macro-based DSL that lets you describe how processes communicate to perform a computation. This top-down description of interacting processes is called a choreography. From this choreography, Chorex creates modules for each process that handle all the message-passing in the system. The interactions performed by the generated code will never deadlock by construction because the choreographic DSL ensures that no processes will be waiting on each other at the same time.
...This is the story of how I solved a problem (ugly, cumbersome boilerplate code) that I ran into while writing a program in a functional language (Elixir). Functional programming languages often pride themselves on expressiveness and elegance; but occasionally they are not amenable to the most obvious solutions to the problems we wish to solve. In this case, the simplest solution to my problem would have been to have a global mutable variable. But no one likes those.
...Somewhere in my adolescence I got stuck with the notion that functional languages were slow while languages like C were fast. Now, a good C programmer can eke more performance out of their code than probably anyone else, but the cost you pay to keep your code correct goes exponential as you get closer and closer to the machine.
Functional languages abstract a lot away from the machine. Higher languages in general abstract away the machine and make code easier to maintain. So I think I had it in my head that functional languages, being far away from the bare metal, must necessarily be slow. It didn’t help that I also thought of functional languages as being necessarily interpreted languages.
...Macros are tricky beasts. Most languages—if they have macros at all—usually include a huge “here there be dragons” warning to warn curious would-be macro programmers of the dangers that lurk ahead.
What is it about macros that makes them so dangerous and unwieldy? That’s difficult to answer in general: there are many different macro systems with varying degrees of ease-of-use. Moreover, making macros easy to use safely is an open area of research—most languages that have macros don’t have features necessary to implement macros safely. Hence, most people steer clear of macros.
...Something I’ve wondered about for a little while: why don’t more languages have a call/cc operator? Having first-class continuations in your programming language gives your programmers a powerful construct. So why do only a handful of languages have it?
The short answer is: it’s tricky to implement efficiently. One way to get call/cc is to convert your code into continuation-passing style. Then, call/cc simply takes the continuation in that representation and binds it to a variable. Most languages don’t seem to go through a continuation-passing style conversion pass though, so there’s no continuation to grab.
Or: Approaching the Y Combinator
These are some of my class notes. Learning to derive the Y Combinator from first principles is something I’ve always wanted to do. This isn’t quite the Y Combinator, but it’s very close and it still gets you recursion without relying on recursive structures to begin with.
In the beginning, we write a recursive function to compute the length of a list:
(let* ([len (λ (lst)
(if (null? lst)
0
(+ 1 (len (cdr lst)))))])
(len '(1 2 3)))
The let* syntax allows us to create local variable bindings that can reference themselves. But let’s suppose we don’t have let*—what do we do?
There’s a neat paper Type Systems as Macros by Chang, Knauth, and Greenman [1] that describes how to implement a typed language using an untyped host language and macro expansion. The paper is neat, but I found the code hard to follow—the paper uses a compact notation that’s convenient for print, but not so much for reproducing on one’s own. This post is my attempt to implement and explain in more accessible terms what’s presented in the paper.
...This is a question I’ve been wrestling with for a little bit. My first experience with a type system was with Java, and I didn’t like it. It just felt like an annoying constraint on the kinds of programs I could write. I was coming from Perl, which sports weak dynamic typing, so Java’s rigidity came as a bit of a shock.
After Java I learned some C, which too has types. C’s types are different from Java’s in a big way: in C they’re really just directives to the compiler on how to interpret some bytes. “Everything is just void *” is kind of true. In C, bytes can be interpreted however you wish.