Ask HN: How to start programming language design? | Hacker News Reader

Ask HN: How to start programming language design?

12 pointsposeid17 comments

Hello,

as far as I know, programming started out with translating expressions like:

    (number) (operation) (number) := (result)

Now, I want to understand a bit more on programming language principles, but reading references quickly results into discussion of abstract algebra.

So here is my question: How can I start designing simple programming languages?

What are some simple, but useful grammars to understand programming language principles?

What are these "treetop" like parsing libraries about, and are these for beginners or advanced users?

View on HN

You should probably spend some time browsing through Lambda the Ultimate [1], a discussion forum devoted to this kind of thing (it is very academic though).

There's a collectively authored set of guidelines "How to Design (Declarative) Programming Languages" [2] posted in comp.lang.lightweight, curated by Scott McKay that you might find useful.

[1] http://lambda-the-ultimate.org

[2] http://thread.gmane.org/gmane.comp.lang.lightweight/2547/foc...

You could take a look at Marc Cournoyer's course http://createyourproglang.com, check out the compiler course on Cousera https://www.coursera.org/course/compilers and watch Rob Pikes talk on lexical scanning http://www.youtube.com/watch?v=HxaD_trXwRE.b

That should get you going. Oh and don't forget to have a look at Go language specification as well.

thanks!

what about these "PEG" parsers? when/why do they make sense?

"PEG-based transformer provides front-, middle and back-end stages in a simple compiler"

http://www.vpri.org/pdf/tr2010003_PEG.pdf

what makes the go language a good example?

The language spec is very short compared to others.

For some quick hands-on learning, it might be interesting to read up on Resursive Descent Parsers (http://en.wikipedia.org/wiki/Recursive_descent_parser). They're easy to understand and easy to code in a language of your choice. Once you have a working one, you can expand it to experiment with your own grammar. Spend some time experimenting that way and you'll be able to get a lot of insight in what the pitfalls of language design are and how you can work with your parser to minimize them.

thanks! do you have some tool that you would recommend to run programs with these grammars. e.g. I have

   calculator.ycomb

and a ycomb grammar.

which programs would you recommend to get an output of calculator.ycomb ?

honestly, I would recommend you write your own parser for it. A grammar is meaningless by itself, and you'll learn a lot more about why yours might work or not work if you do the work yourself.

I can't say I know of any good PEG implementations, and even if there were, you'd be limited to what you can express in PEG. A real language can be far broader, encompassing inputs, outputs, complicated functions, etc.

interesting advice to get hands dirty with source code directly. I might think to experiment a bit with C/lexx/yacc then.

one thing, I am curious about, what are the kind of "hello_world()" functions, that would print something on a screen... hmm..

Getting in on the ground floor tends to be my advice for a lot things. It's hard to make something click if you can't see it work in the flesh.

For the typical functions that print output or take input or the likes, I'm always partial to keeping them separate from the grammar itself. You can see the distinction clearly between for example Python:

  print myVar

and Java:

  System.out.println(myVar);

In Python, the print verb is defined in the grammar, and as such needs to be parsed and handled properly. In Java's case, it's regarded as any other method. It's irrelevant to the parser what it does, it just needs to handle it like it would a normal method. This tends to make the whole easier to debug and adapt, since you never need to touch the parser to change behavior of one of these standard methods.

This does ofcourse lead to more complex applications, which may not be desirable, depending on the intent of the language. Just my 2 cents :)

maybe this is interesting too: https://github.com/jashkenas/coffeescript/wiki/%5BHowto%5D-H...

and http://zaach.github.io/jison/docs/#installation

Those are interesting resources, but do remember that a context-free grammar is exactly that: context-free. It might be a fun starting point, but it's much more limited than a grammar you would usually find a real programming language.

Check the next generation approach for a modern IDE:

http://blog.jetbrains.com/blog/2014/05/27/nitra-goes-open-so...

https://github.com/JetBrains/Nitra

with Nitra, how would a "hello_world()" of a programming language print something to the screen?

Here's the best "hands-on" introduction I've seen: http://nathansuniversity.com/

He walks you through the basics of compiling, parsing, and interpreting to create your own scheme-like language (complete with exercises along the way).

Reply on news.ycombinator.com