There's a very interesting thread over at HP Museum, where Thomas Klemm has noted that Python is a convenient high-level language which compiles to bytecode and which includes introspection - so you can automatically convert from python, via bytecode, to a machine language of choice. Python can disassemble its own bytecode so the transformations can be text-based.

The whole thing works well if the machine language of choice has a structural resemblance to python bytecode - in this case it's Focal, which runs on the HP-41C calculator. Perhaps one could convert to a virtual machine like Forth or Sweet 16, or perhaps to assembly language, perhaps in a threaded style.

Thomas put up an example you can run in your browser. His 300 line program converts
to

Very interesting.

_________________
Michael A.

I contribute to a java bytecode compiler project (java_grinder), which is based on a similar concept. Some random thoughts:

-It's totally worth doing this with Python, especially if you can work out a way to call ML routines. Then its just like mixing BASIC/ML back in the day - but with a much nicer & faster language.

-You can also interpret the bytecode as you see fit. For example, my efforts were limited to 16-bit integers and lack floating-point support.

-Using a separate data stack made life a lot easier, see:
http://wilsonminesco.com/stacks/virtualstacks.html

Working on this sort of stuff has been very rewarding for me, I highly encourage it.

I've actually used Python to compile my Forth-83 implementation for the VIC-20. The input source code is entirely Forth (based on the 6502 reference model) and the compiler generates a binary image which contains the elemental words (compiled from Forth ASSEMBLER words) and the rest as a list of constituent words.

Using Python meant building and testing the compiler was easy, I have classes for stacks, target code blocks etc and I can dump the contents of the dictionary to help with debugging.

Welcome, srowe. Nice work defying the "chicken or the egg" problem! As you know, it's usually necessary to use a conventional assembler to create the Forth kernal, which means low-level definitions in the reference code (written in the RPN syntax expected by Forth's own assembler) can't be used.

You may wish to elaborate on what you've done by starting a thread in the Forth section of the forum.

Cheers,
Jeff

_________________
In 1988 my 65C02 got six new registers and 44 new full-speed instructions!
https://laughtonelectronics.com/Arcana/ ... mmary.html

Yes, welcome! I see there is a thread elsewhere (V-FORTH - Forth-83 for the VIC) with an attachment too.

Thanks. That version is a little out of date (but the git URL is still valid). I'll create a post in the Forth section shortly.