Build Your Own Forth Interpreter
codingchallenges.fyiI've been on/off working on a homebrew NES game. Pretty much the go-to environment for that is assembly, which I'm sure I could write if I were motivated, but I find assembly considerably un-fun so I wanted to use a higher-level language.
I had been looking for an excuse to learn Forth, and its use in classic computing meant that it had a shot of being workable on the NES.
I was initially using IceForth but I had trouble getting that working, and so I got Codex to generate something that works, but then I also that building your own Forth is sort of a rite of passage for a software engineer, so I have been building my own Forth from scratch.
My custom hack-job isn't ready yet, but I was extremely impressed at the performance I was able to get on the NES with compiled Forth from the Codex thing on the NES. I'm getting roughly 80% of the speed for equivalent programs written in assembly, with much less code and this is without advanced optimizations. I do plan on finishing my custom one because I think I can build what I want a bit better than Codex, and I'm optimistic I can get the performance reasonable.
Forth is such a fascinating language, because it sort of enables you to work at any level of the program. You can write it super high level, almost like Lisp, but you can also poke around and create mappings to assembly, and you can do all this with decent performance no less! It has quickly become one of my more favorite scripting languages, though that might be because I have always had a soft spot for RPN.
thats so awesome. i wonder what a live forth interpreter would look like on NES.
A live interpreter turns out to be very slow! It's still useful but it's about 1/10th the speed of the compiled version. I've mostly just been using it to play with placement and test out my ppu logic. Honestly a lot of the work is being handled by BizHawk.
The compiler is pretty neat though; my current system is doing a straightforward conversion of the threaded-code quasi-AST thing that Forth does and convert it all to JSR calls, where each name is a subroutine. I haven't done a ton of optimizations yet, so I'm actually surprised it's going as fast as it is. I am just going to say this is yet another "Forth is awesome" thing.
Once I get it into a more stable state, I'll push it up to SourceHut.
Video where I demonstrate how I explore JONESFORTH using GDB:
I enjoyed that! Nice intro to JonesForth and how it maps to assembly instructions.
I've already done that---ANS Forth for the 6809 (https://github.com/spc476/ANS-Forth).
Advanced challenge: make it self-hosting.
I like building little throwaway FORTH interpreters as an exercise when learning a new language. They tend to touch just enough common programming needs to be interesting: read user input and/or a file, parse the input, run a state machine with two stacks, write output.
In the past years, I developed a forth like language as an intermediate language for a C compiler. For debuging purposes, I also implemented a memory safe interpreter, besides a compiler that generates assembly output.
See my profile for link to my github repositories and look under MES-replacement for stack_c
Here’s a link to the right directory at least: https://github.com/FransFaase/MES-replacement/tree/main/src
I’ve been working on a homemade CPU (in simulator) and I’m in the middle of implementing an ISA and assembler as a bug step up from working in machine code to working in assembly. I’ve been looking at Forth as a good option for a next-level-up language which is relatively easy to implement and easy to script with.
I wrote one some time ago, and it took me a lot longer than I thought it should. (it's core is finished, but dictionary is limited) https://ntavish.in/projects/emforth/
You might also enjoy this tutorial, which started out being based upon a hacker news thread:
I go through stages to implement minimal language.
For some reason this site feels AI generated to me.
"if you know one forth, you know one forth"
So implement four of them, and you will know them all! First Forth with indirect threaded code, second Forth with direct threaded code, third Forth with subroutine threaded code, and the final fourth with token threaded code.
You jest, but I did end up doing just that in my implementation (https://github.com/romforth/romforth) trying to shoehorn a Forth implementation into a MSP430 device with just 2KB ROM + 128 byte RAM
Interesting. I have a slightly off topic question: do you agree that there are no good newbie programming manual for msp430? in comparison to Arduino or PIC etc
I thought this was going to be a pun on the word "fourth", disappointed when I got to "final".
I doubt you will want to code professionally in Forth unless you work on embedded, so the dialect you learn doesn't matter too much. But it is interesting to implement a small interpreter and play with it.
(I'm saying this with love)
This is a strange article imo.
I was expecting to see FORTH in bare metal C or ASM.
There is a common myth about newbie programmers that FORTH is write-only and that you need to type everything in one line, without comments or function calls etc.
Writing forth is super easy especially if you have a stack machine at your disposal. For example when you are building your own virtual cpu/architecture with assembler and compiler.
It's more trivial than to understand any JavaScript framework lol
Research FORTH more guys - it doesn't need to be strange and hard :)
ps. Lisp SUCKS
/rant
I was with you 'till the last line. :P
IMO Lisp is harder to implement than Forth, and LESS readable, butt MAYBE i fell into the same trap as others with Forth. hahaha
actually I think Lisp is easier to implement than Forth, buth is it really Forth if the internals aren’t discussed? (E.g. Word secondary, threaded code, etc)
Forth does not specify threaded code. Implementation is left to the implementor.
Internally Forth can be direct threaded code, indirect threaded code, byte code, sub-routine calls or optimized native code.
> There is a common myth about newbie programmers that FORTH is write-only and that you need to type everything in one line, without comments or function calls etc.
By contrast, in APL it's not a myth at all.