You should care about colorForth!

July 22, 2023

Brad Nelson / @flagxor

Caveats

  • I'm hoping to talk about colorForth in the large
  • I want to talk about the potent ideas in Chuck's explorations
  • I have an incomplete view of Chuck's work

Introduction

  • Chuck's work has always been about "Decomplexifying"
  • Software and hardware is SO complex now
  • But kilobytes still matter, and everything is so slow
  • We really need to find a better way to write software!
  • The ideas in colorForth point the way

Note about Colors

  • word - define
  • word - compile
  • word - execute
  • 1234 - execute numbers
  • word - variable
  • word - force compile
  • word - decorate
  • word - comment

Note about Colors

For older slides
  • word - define
  • word - compile
  • word - execute
  • 1234 - execute numbers
  • word - comment

Don't get distracted by color

  • colorForth isn't about syntax highlighting or a neat alternate visualization
  • It's not about the eclectic editor
  • It's informed by making Forth even simpler

colorForth Chronology

  • Precursors (1970-1997)
  • iTV Era (1998-20??)
  • GreenArrays Era (20??-2015)
  • uhdForth (2016-present)

Precursors

What is Forth?

In 1999 Chuck said:
  • Highly factored code
  • Definitions
  • Stacks (to support definitions)

What is a definition?

In 1999 Chuck said:
  • : SOME ... ;
  • an abbreviation
  • should never have more than 1-2 arguments
  • stack should never be more than 3-4 deep

You can write FORTRAN in any language

  • Forth is powerful, it lets you define any syntax
  • There's a temptation to borrow from other languages
  • locals, exceptions, structures, scopes, types
  • "Find the hundred words to solve you problem in one line."
DEA- FIG,

I AM A FOR-- PRO------- CUR------
EMP----- BY FOR-- INC.  I HAV- NOT
WOR--- ON ANY FIG TYP- SYS---- AND I
AM EXC---- BY THE VAR----- LEN---
NAM- IDE-. PLE--- SEN- ME THE FIG
FOR-- MOD-- SO THA- I MAY TRY IT OUT
HER- AT FOR-- INC.

FRE- THO----
Forth Dimensions Volume 2 Number 6 
DEAR FIG,

I AM A FORTH PROGRAMMER CURRENTLY
EMPLOYED BY FORTH INC.  I HAVE NOT
WORKED ON ANY FIG TYPE SYSTEMS AND I
AM EXCITED BY THE VARIABLE LENGTH
NAME IDEA. PLEASE SEND ME THE FIG
FORTH MODEL SO THAT I MAY TRY IT OUT
HERE AT FORTH INC.

FRED THORTON
Forth Dimensions Volume 2 Number 6 
DEA- EDI---

I AM AFR--- THA- THE LET--- IN THE LAS- ISS-- ABO--
FOR-- INC- USI-- ONL- THR-- LET--- NAM- FIE--- HAS
HAD THE OPP----- EFF--- FRO- WHA- THE WRI--- WAN---

HIS LET--- ( LIK- THI- ONE ) SHO-- THA- SAV--- ONL-
THR-- LET---- AND COU-- IS JUS- ABO-- OPT---- IN
TER-- OF A TRA-- OFF BET---- SAV--- MEM--- AND
KEE---- LEG--------

YOU-- TRU--
CHU-- MOO--
FOR-- INC-

P.S-  MR. FRE- THO---- IS NOT AN EMP----- OF
      FOR-- INC-
Forth Dimensions Volume 3 Number 1 
DEAR EDITOR

I AM AFRAID THAT THE LETTER IN THE LAST ISSUE ABOUT
FORTH INC. USING ONLY THREE LETTER NAME FIELDS HAS
HAD THE OPPOSITE EFFECT FROM WHAT THE WRITER WANTED

HIS LETTER ( LIKE THIS ONE ) SHOWS THAT SAVING ONLY
THREE LETTERS AND COUNT IS JUST ABOUT OPTIMAL IN
TERMS OF A TRADE OFF BETWEEN SAVING MEMORY AND
KEEPING LEGIBILITY.

YOURS TRULY
CHUCK MOORE
FORTH INC.

P.S.  MR. FRED THORTON IS NOT AN EMPLOYEE OF
      FORTH INC.
Forth Dimensions Volume 3 Number 1

Why was standardizing Forth hard?

  • Forth is more complex than it seems
  • Dictionary: links, vocabulary hooks, variable length entries
  • You can't even define Forth dictionary entries statically in C!
  • Fig-Forth had 221 words, not including internals

1993

iTV Era

Early colorForth

  • 6 Colored spaces: define, compile, execute, excute number, comment preceed words
  • 4 x 3 "stack" of word colors
  • Interpret / compile is table dispatch
  • square dup * ;

Why is this simpler?

  • Compilation and interpretation is more uniform
  • How to parse: number / lookup is explicit
  • Ambivalence about IMMEDIATE words
      FRAME EMPTY VARIABL
      E BUF W H : ROW A! 1
      59 BEGIN @+ IF + ; T
      HEN DROP NEXT 0 ; RO
      WS DUP BUF ! ROW IF
      DROP DROP ; THEN DRO
      P -1  H +! DUP A + R
      OWS ; 448 H ! VGA 0
      OVER ROWS -1280 SWAP
      640 447 * + ROWS
          
      FRAME EMPTY VARIABL
      E BUF W H : ROW A! 1
      59 BEGIN @+ IF + ; T
      HEN DROP NEXT 0 ; RO
      WS DUP BUF ! ROW IF
      DROP DROP ; THEN DRO
      P -1  H +! DUP A + R
      OWS ; 448 H ! VGA 0
      OVER ROWS -1280 SWAP
      640 447 * + ROWS
          
FRAME EMPTY
VARIABLE BUF W H :
ROW A! 159 BEGIN @+ IF + ; THEN DROP NEXT 0 ;
ROWS DUP BUF ! ROW IF DROP DROP ; THEN
     DROP -1  H +! DUP A + ROWS ;
448 H ! VGA 0
OVER ROWS -1280 SWAP
640 447 * + ROWS

Address Registers

  • A! - Store to A
  • A - Get value of A
  • @+ - Read from A and increment
  • !+ - Store to A and increment

Less Greedy Conditionals

  • IF doesn't consume its argument
  • Avoids DUPs
  • -IF - Tests sign bit

Simpler flow control

  • DO .. LOOP - 2 parameters and just too complicated
  • FOR .. NEXT - 1 parameter, good for hardware
  • BEGIN .. UNTIL - variable num params
  • WORD: ... IF ... WORD ; THEN ... ; - seems to be adequate
  • Needs recursive definitions, no smudge
  • Tail recursion

GreenArrays Era

Howerd Oakford

  • Howerd worked hard to make colorForth easier to run and more accessible
  • https://www.inventio.co.uk/cf2023/index.html
  • Thanks Howerd!

Example: Fibonacci Numbers

fib ( n--n ) push 1 dup pop -2 +
f ( x...x ) -if drop nip ; then
            push swap over + pop -1 + f ;
8 fib

Example: Fibonacci Numbers

: fib ( n -- n ) >r 1 1 r>
   1- for swap over + next drop nip ;
8 fib

Pre-parsed Words

Shannon Coded Words

Why is this awesome?

  • One word = one machine word
  • Dispatch is now even more trivial, mask and call
  • Parse numbers at edit time
  • Dictionary is two arrays

The compiler / interpreter

: eval ( a n -- )
   for
     dup @ dup untag execute 1+
   next drop ;

Parts of the system

  • dispatcher
  • dispatch word per color: number unpacks, define adds to dict, compile adds a call, execute calls
  • editor happens beforehand
  • seed the dictionary with basic words

Why aren't we all doing this?

  • Chuck's implementation doesn't work where we want it
  • Which conventions do we adopt?
  • Which source conventions do we adopt?

Quirks of "standard" colorForth

  • Limitations on name encoding
  • Two vocabularies only (forth + macro)
  • Definitions drop through
  • Literal on execute to compile
  • No standard loops
  • IF preserves stack
  • -IF in place of 0< IF
  • No ELSE

MORE Quirks of "standard" colorForth

  • >R R> become push pop
  • A! A @+ etc.
  • @ ! use indexes instead of addresses
  • Requires special editor
  • Loss of CREATE DOES&rt;
  • Unclarity around allocation

Why have I been stuck?

  • My RainbowForth was too complex
  • I got stuck on trying to match colorForth conventions
  • How to sensibly interface with "normal" things?
  • Source control, github

Why have I been stuck?

  • Subroutine threading is hard
  • Color constrains what kind of environment
  • Is it a Turing Tar Pit?
  • Macro vocabulary or not?
  • Can it be even simpler? No compiler, just an editor

uhdForth Era

What is Chuck grappling with?

  • Running on Windows
  • Avoiding non-explicit macros
  • Wants to touch the machine code
  • Keeps vacilating on if source is needed

How can we take this somewhere?

  • Pick a source format + build something?
  • Pick a target? ESP32-colorForth?

What's great with it?

  • Editors really are easy to write
  • Forth interpreters are harder, less in Forth
  • Strings are messy, we haven't found a great vocabulary for them

Proposal

  • Try to write simple easy short definititions ;
  • Think about how to live without: vocabularies, locals, strings, files, unicode, complexity, CREATE DOES>, objects
  • Come back with something useful


flagxor.com
slides

⚘

Thank you