Comments on: Build a Compiler

By: kyunghoon

kyunghoon — Mon, 19 Apr 2010 00:12:46 +0000

Lisp like languages are too easy to parse, therefore would not be a very good exercise in that area. My vote is for JavaScript. Building something on par with the other speedy JavaScript engines out there would be very interesting.

By: Troy Straszheim

Troy Straszheim — Sun, 03 Jan 2010 20:28:40 +0000

While I’m at it, that’s from the ‘major goals for D’ section:

- Make D substantially easier to implement a compiler for than C++.
- Be compatible with the local C application binary interface.
- Have a context-free grammar.

of this page: http://www.digitalmars.com/d/2.0/overview.html

By: Troy Straszheim

Troy Straszheim — Sun, 03 Jan 2010 20:16:15 +0000

There’s “D”, which was designed from the beginning to be C-like with a context-free grammar….

By: Jesse

Jesse — Wed, 30 Dec 2009 17:43:05 +0000

lol! I used to know a fellow (Brian Prentice) who had worked for Boeing, and he told me that whenever they got a new system the first thing he’d do was to implement a Forth for it, since as far as he was concerned that was the quickest way to bootstrap pretty much everything.

By: Joel de Guzman

Joel de Guzman — Wed, 30 Dec 2009 01:07:30 +0000

Ah Forth! Brings in fond memories Do you need a parser for that?

By: Jesse

Jesse — Wed, 30 Dec 2009 00:46:14 +0000

I guess it comes back to the goal of your project. If the goal is to introduce Spirit by example, I think something like Scheme that is quite small yet fairly widely-known is not a bad idea.

On the other hand, something more ALGOL like could have more appeal, since a lot of people use those kinds of languages and could see familiar constructs. The advantage might be that learners could focus less on trying to figure out the language being implemented, and more on how Spirit is used to implement familiar constructs.

On the other other hand (am I out of hands, yet?) I don’t think anyone’s mentioned Forth.

By: Glen

Glen — Sun, 27 Dec 2009 03:51:46 +0000

Hey Joel

I’ve only just casually come across your site, so I might not understand what you are thinking about doing, but you said:

“C++ would be the ultimate. But that would mean spending the next N years of my life working on it ”

Well Bjarne has just published this work which seems like a paper that might help?:
http://www2.research.att.com/~bs/macis09.pdf

Wouldn’t his paper make your job a heck of a lot easier? So maybe you can do it after all, waddya think? Maybe he’d even help

Anyway, the link had a security server error when I tried to open it but I saw an older version of this document (I think) so when it’s back up take a look. I found it via Bjarne’s usual papers page:

http://www2.research.att.com/~bs/papers.html

Thanks.

By: Ben Allan

Ben Allan — Sat, 19 Dec 2009 11:19:24 +0000

I’ll nominate Java, since I’d really love to see a grammar I can then use to derive open-source source-to-source tools for taking code out of Java and several other java-like languages and into C/C++ without a tool-chain dependence on llvm or a java compiler/runtime.

Or +1 for C, and particularly for just updating Hartmut’s example from http://spirit.sourceforge.net/ as a full example of up-porting to 2.1 from 1.x. Separately, the tutorial aspects of walking a C ast, handling crufty bits in “real” languages, etc could be built on top of the updated base of parsing C.

By: bytecolor

bytecolor — Thu, 17 Dec 2009 14:31:03 +0000

Joel
What did you have in mind when you said:

interpreter that can be immediately executable

I’ve been working on an ISO EBNF parser. It works pretty good. I think I have all
of the syntax parsing correctly, including the alternate symbols (/ /) / ! etc.
I’ve got a visitor written that just writes the data structures back out as
semi-formatted EBNF.

?> foo = ["this" | "that"], 4*”x”;
foo
    = ['this' | 'that'], 4 * ‘x’
    ;
?> foo = (/’this’ / “that”/), 4*”x” .
foo
    = ['this' | 'that'], 4 * ‘x’
    ;

By: Jonas Persson

Jonas Persson — Tue, 15 Dec 2009 15:11:57 +0000

RAD spirit would be cool. A small language where you could connect a qi grammar and a karma grammar and have it typechecked for compatible attributes could make for a really nice and simple text transformation tool.