[PicForth] food for thought - picforth as compiler back-end

[David McNab]

Hi,

My recent quietness on this list can be explained by the fact that I'm 
trying out the CCS C compiler, and am quite impressed so far with its 
performance - seems to be the most practical C-like compiler for PIC. I 
might well make an exception to my usual 'free software only' principles 
and buy a copy.

Regarding PicForth - I reached a situation of getting frustrated with 
the Forth language in general. Annoyances such as:
  - constant need for stack accounting
  - ease of introducing fatal bugs
  - application code tends to be far less readable than its equivalent
    in other languages like C, even with prolific stack-effect comments
  - coding/debugging effort is (at least for me) greater with Forth than
    for traditional languages like C.
  - forth compilers are far less capable of picking up bugs (eg function
    argument/return mismatches) than compilers of other languages

All this said - PicForth does offer one beautiful thing - its capability 
to understand a procedural, stack-capable language higher-level than 
assembler, and generate efficient, compact and largely reliable code.

So a thought hit me - PicForth could be used as a back-end for compilers 
of other high-level languages such as Python, C, Ruby etc. A parser 
could be written with the usual tools (eg yacc/bison, PLY, Jack etc). 
Converting the parser's generated AST into PicForth source code would 
take a fraction of the work of the traditional practice of generating 
assembler or machine-code.

I'm going to attempt a Python to PicForth compiler, and see how far I get.

Thoughts anyone?

-- 
Cheers
David