retroforth/literate/RetroForth.md
crc b8131ba437 fix a typo causing crashes on display of numbers
FossilOrigin-Name: 73f62c723c58a9c1acbd29fef76172f6caa37f20d6c0d74f398f760a84215522
2019-03-14 18:19:10 +00:00

46 KiB

RETRO FORTH

Background

Retro is a dialect of Forth. It builds on the barebones Rx core, expanding it into a flexible and useful language.

Over the years the language implementation has varied substantially. Retro began in 1998 as a 16-bit assembly implementation for x86 hardware, evolved into a 32-bit system with cmForth and ColorForth influences, and eventually started supporting mainstream OSes. Later it was rewritten for a small, portable virtual machine.

This is the twelfth generation of Retro. It targets a virtual machine (called Nga) and runs on a wide variety of host systems.

Namespaces

Various past releases have had different methods of dealing with the dictionary. I have settled on using a] single global dictionary, with a convention of using a short namespace prefix for grouping related words. This was inspired by Ron Aaron's 8th language.

The main namespaces are:

| namespace  | words related to   |
| ---------- | ------------------ |
| ASCII      | ASCII Constants    |
| array      | arrays             |
| c          | characters         |
| compile    | compiler functions |
| d          | dictionary headers |
| err        | error handlers     |
| n          | numbers            |
| s          | strings            |
| v          | variables          |

This makes it very easy to identify related words, especially across namespaces. E.g.,

c:put
c:to-upper
s:put
s:to-upper

Prefixes

Prefixes are an integral part of Retro. These are single symbol modifiers added to the start of a word which control how Retro processes the word.

The interpreter model is covered in Rx.md, but basically:

- Get a token (whitespace delimited string)
- Pass it to `interpret`
  + if the token starts with a known prefix then pass
    it to the prefix handler
  + if the initial character is not a known prefix,
    look it up
    - if found, push the address ("xt") to the stack
      and call the word's class handler
    - if not found call `err:not-found`
- repeat as needed

This is different than the process in traditional Forth. A few observations:

- there are no parsing words
- numbers are handled using a prefix
- prefixes can be added or changed at any time

The basic prefixes are:

| prefix | used for               |
| ------ | ---------------------- |
| :      | starting a definition  |
| &      | obtaining pointers     |
| (      | stack comments         |
| `      | inlining bytecodes     |
| '      | strings                |
| #      | numbers                |
| $      | characters             |
| @      | variable get           |
| !      | variable set           |

Naming and Style Conventions

  • Names should start with their namespace (if appropriate)
  • Word names should be lowercase
  • Variable names should be Title case
  • Constants should be UPPERCASE
  • Names may not start with a prefix character
  • Names returning a flag should end with a ?
  • Words with an effect on the stack should have a stack comment

Code Begins

Memory Map

This assumes that the VM defines an image as being 524,288 cells. Nga implementations may provide varying amounts of memory, so the specific addresses will vary.

| RANGE           | CONTAINS                     |
| --------------- | ---------------------------- |
| 0 - 1024        | rx kernel                    |
| 1025 - 1535     | token input buffer           |
| 1536 +          | start of heap space          |
| ............... | free memory for your use     |
| 506879          | buffer for string evaluate   |
| 507904          | temporary strings (32 * 512) |
| 524287          | end of memory                |

I provide a word, EOM, which returns the last addressable location. This will be used by the words in the s: namespace to allocate the temporary string buffers at the end of memory.

:EOM  (-n)  #-3 fetch ;

Stack Depth

depth returns the number of items on the data stack. This is provided by the VM upon reading from address -1.

:depth  (-n) #-1 fetch ;

Stack Comments

Stack comments are terse notes that indicate the stack effects of words. While not required, it's helpful to include these.

They take a form like:

(takes-returns)

I use a single character for each input and output item. These will often (though perhaps not always) be:

n, m, x, y  number
a, p        pointer
q           quotation (pointer)
d           dictionary header (pointer)
s           string
c           character (ASCII)

Dictionary Shortcuts

I define a few words in the d: namespace to make it easier to operate on the most recent header in the dictionary. These return the values in specific fields of the header.

:d:last        (-d) &Dictionary fetch ;
:d:last<xt>    (-a) d:last d:xt fetch ;
:d:last<class> (-a) d:last d:class fetch ;
:d:last<name>  (-s) d:last d:name ;

Changing A Word's Class Handler

I implement reclass to change the class of the most recent word.

:reclass    (a-) d:last d:class store ;

With this I can then define immediate (for state-smart words) and data to tag data words.

:immediate  (-)  &class:macro reclass ;
:data       (-)  &class:data  reclass ;
:primitive (-) &class:primitive reclass ;

Visual Grouping

Comments start with a ( and end at the first whitespace. It's useful to be able to punctuate a code block, providing some inline commentary and having a clear end point. We can provide this with a single word ).

Example:

:pre.min (a-an)
  (comparison  &lt? &lt-or-gt? set-hook )
  (begin_with  #-1 !Index n:min !Value dup array:length ) ;
:) ; immediate

Optimizations & Compiler Extensions

I have a compile namespace for some low level words that compile specific Nga bytecode. This is intended to aid in readability when constructing compiler extensions.

:compile:lit  (a-) (li...... #1    , , ) ;
:compile:jump (a-) (liju.... #1793 , , ) ;
:compile:call (a-) (lica.... #2049 , , ) ;
:compile:ret  (-)  (re...... #10   ,   ) ;

The compiler state is stored in a value named Compiler. I have an accessor word that aids in readability.

:compiling?  (-f)  &Compiler fetch ;

It's sometimes useful to inline values directly. I use a backtick prefix for this.

:prefix:`  (s-)
  s:to-number , ; immediate

It's traditional to have a word named here which returns the next free address in memory.

:here  (-a)
  &Heap fetch ;

Variables

The next two are additional prefixes to make working with variables a bit less painful. By default you have to do things like:

&Name fetch #10 * &Name store

Or using combinators:

&Name [ fetch #10 * ] sip store

With the @ and ! prefixes this can become:

@Name #10 * !Name

When compiling, these will generate packed Nga instructions corresponding to:

lit + fetch + nop + nop    'life....   #3841
lit + store + nop + nop    'list....   #4097
:prefix:@  (s-n)
  d:lookup d:xt fetch
  compiling? [ (life.... #3841 , , ) ]
             [ fetch                 ] choose ; immediate

:prefix:!  (s-n)
  d:lookup d:xt fetch
  compiling? [ (list.... #4097 , , ) ]
             [ store                 ] choose ; immediate

The next few words aren't actually useful until the s: namespace is defined. With strings and the ' prefix they allow creation of variables and constants.

| To create a                  | Use a form like    |
| ---------------------------- | ------------------ |
| Variable                     | `'Base var`        |
| Variable, with initial value | `#10 'Base var<n>` |
| Constant                     | `#-1 'TRUE const`  |

The lower level kernel provides d:add-header to make a new header. This is a bit ugly to use as most of the time I don't need all of the flexibility it provides. Here I add a word to create a new header pointing to here. This is then used to build other data structures without invoking the : compiler.

:d:create (s-)
  &class:data #0 d:add-header
  here d:last d:xt store ;

And then the others are trivial.

:var<n> (ns-) d:create , ;
:var    (s-)  #0 swap var<n> ;
:const  (ns-) d:create d:last d:xt store ;

The const word is an example of using the dictionary and word classes to do some optimization.It creates a header, with a class of class:data, then sets the execution pointer to the desired value. Since the data class either leaves the word pointer on the stack or compiles it as a literal into a definition, this allows constants to exist as just a header with no extra runtime code.

Stack Shufflers

The core Rx language provides a few basic stack shuffling words: push, pop, drop, swap, and dup. There are quite a few more that are useful. Some of these are provided here.

Most of these are implemented as raw bytecodes which can be inlined when used in definitions. The high level definitions are:

:tuck      (xy-yxy)   dup push swap pop ;
:over      (xy-xyx)   push dup pop swap ;
:nip       (xy-y)     swap drop ;
:drop-pair (nn-)      drop drop ;

And the low level forms:

:tuck      (xy-yxy)   (dupuswpo `100926722 ) ; primitive
:over      (xy-xyx)   (puduposw `67502597  ) ; primitive
:nip       (xy-y)     (swdr.... `772       ) ; primitive
:drop-pair (nn-)      (drdr.... `771       ) ; primitive
:?dup      (n-nn|n-n) (duzr.... `6402      ) ; primitive
:dup-pair  (xy-xyxy)  over over ;

Combinators

Retro makes use of anonymous functions called quotations for much of the execution flow and stack control. The words that operate on these quotations are called combinators.

Combinators are a major part of using Retro. They help in reducing the use of lower level shuffling and allow for a greater overall consistency in the syntax. I also find them to help in reducing visual noise.

Combinators: Data

dip executes a quotation after moving a value off the stack. The value is restored after execution completes. These are equivilent:

#10 #12 [ #3 - ] dip
#10 #12 push #3 - pop
:dip  (nq-n)  swap push call pop ;

sip is similar to dip, but leaves a copy of the value on the stack while the quotation is executed. These are equivilent:

#10 [ #3 * ] sip
#10 dup push #3 * pop
:sip  (nq-n)  push dup pop swap &call dip ;

Apply each quote to a copy of x.

:bi  (xqq-)  &sip dip call ;

Apply q1 to x and q2 to y.

:bi*  (xyqq-) &dip dip call ;

Apply q to x and y.

:bi@  (xyq-)  dup bi* ;

Apply each quote to a copy of x.

:tri  (xqqq-)  [ &sip dip sip ] dip call ;

Apply q1 to x, q2 to y, and q3 to z.

:tri* (xyzqqq-)  [ [ swap &dip dip ] dip dip ] dip call ;

Apply q to x, y, and z.

:tri@ dup dup tri* ;

Combinators: Control

Execute quote until quote returns a flag of 0.

:while  (q-)
  [ repeat dup dip swap 0; drop again ] call drop ;

Execute quote until quote returns a non-zero flag.

:until  (q-)
  [ repeat dup dip swap #-1 xor 0; drop again ] call drop ;

The times combinator runs a quote (n) times.

:times  (nq-)
  swap [ repeat 0; #1 - push &call sip pop again ] call drop ;

case is a conditional combinator. It's actually pretty useful. What it does is compare a value on the stack to a specific value. If the values are identical, it discards the value and calls a quote before exiting the word. Otherwise it leaves the stack alone and allows execution to continue.

Example:

:c:vowel?
  $a [ TRUE ] case
  $e [ TRUE ] case
  $i [ TRUE ] case
  $o [ TRUE ] case
  $u [ TRUE ] case
  drop FALSE ;
:case
  [ over eq? ] dip swap
  [ nip call #-1 ] [ drop #0 ] choose 0; pop drop drop ;

:s:case
  [ over s:eq? ] dip swap
  [ nip call #-1 ] [ drop #0 ] choose 0; pop drop drop ;

A Shortcut

:prefix:|
  d:lookup [ d:xt fetch ] [ d:class fetch ] bi
  compiling? [ [ class:data ] dip compile:call ]
             [ call ] choose ; immediate

Conditionals

Taking a break from combinators for a bit, I turn to some words for comparing things. First, constants for TRUE and FALSE.

Due to the way the conditional execution works, only these values can be used. This is different than in a traditional Forth, where non-zero values are true.

:TRUE  (-n) #-1 ;
:FALSE (-n)  #0 ;

The basic Rx kernel doesn't provide two useful forms which I'll provide here.

:lteq?  (nn-f)  dup-pair eq? [ lt? ] dip or ;
:gteq?  (nn-f)  dup-pair eq? [ gt? ] dip or ;
:if;   (qf-)  over [  if ] dip         0; pop drop-pair ;
:-if;  (qf-)  over [ -if ] dip #-1 xor 0; pop drop-pair ;

And then some numeric comparators.

:n:MAX        (-n)    #2147483647 ;
:n:MIN        (-n)    #-2147483648 ;
:n:zero?      (n-f)   #0 eq? ;
:n:-zero?     (n-f)   #0 -eq? ;
:n:negative?  (n-f)   #0 lt? ;
:n:positive?  (n-f)   #-1 gt? ;
:n:strictly-positive?  (n-f)  #0 gt? ;
:n:even?      (n-f)  #2 /mod drop n:zero? ;
:n:odd?       (n-f)  #2 /mod drop n:-zero? ;

More Stack Shufflers.

rot rotates the top three values.

High level:

:rot  (abc-bca)   [ swap ] dip swap ;

And low level, for inlining:

:rot (abc-bca) `67503109 ; primitive

Numeric Operations

The core Rx language provides addition, subtraction, multiplication, and a combined division/remainder. Retro expands on this.

I implement the division and remainder as low level words so they can be inlined. Here's the high level forms:

:/         (nq-d)  /mod nip ;
:mod       (nq-r)  /mod drop ;
:/         (nq-d)  (diswdr.. `197652 ) ; primitive
:mod       (nq-r)  (didr.... `788    ) ; primitive
:not       (n-n)   #-1 xor ;
:n:pow     (bp-n)  #1 swap [ over * ] times nip ;
:n:negate  (n-n)   #-1 * ;
:n:square  (n-n)   dup * ;
:n:sqrt    (n-n)
  #1 [ repeat dup-pair / over - #2 / 0; + again ] call nip ;
:n:min     (nn-n)  dup-pair lt? [ drop ] [ nip ] choose ;
:n:max     (nn-n)  dup-pair gt? [ drop ] [ nip ] choose ;
:n:abs     (n-n)   dup n:negate n:max ;
:n:limit   (nlu-n) swap push n:min pop n:max ;
:n:inc     (n-n)   #1 + ;
:n:dec     (n-n)   #1 - ;
:n:between? (nul-) rot [ rot rot n:limit ] sip eq? ;

Some of the above, like n:inc, are useful with variables. But it's messy to execute sequences like:

@foo n:inc !foo

The v: namespace provides words which simplify the overall handling of variables. With this, the above can become simply:

&foo v:inc

:v:inc-by  (na-)   [ fetch + ] sip store ;
:v:dec-by  (na-)   [ fetch swap - ] sip store ;
:v:inc     (n-n)   #1 swap v:inc-by ;
:v:dec     (n-n)   #1 swap v:dec-by ;
:v:limit   (alu-)
  push push dup fetch pop pop n:limit swap store ;
:v:on      (a-)    TRUE swap store ;
:v:off     (a-)    FALSE swap store ;
:allot     (n-)    &Heap v:inc-by ;

v:preserve is a combinator that executes a quotation while preserving the contents of a variable.

E.g., instead of:

@Base [ #16 !Base ... ] dip !Base

You can do:

&Base [ #16 !Base ... ] v:preserve

This is primarily to aid in readability. I find it to be helpful when revisiting older code as it makes the intent a bit clearer.

:v:preserve (aq-)
  swap dup fetch [ [ call ] dip ] dip swap store ;

If you need to update a stored variable there are two typical forms:

#1 'Next var<n>
@Next #10 * !Next

Or:

#1 'Next var<n>
&Next [ fetch #10 * ] sip store

v:update-using replaces this with:

#1 'Next var<n>
&Next [ #10 * ] v:update-using

It takes care of preserving the variable address, fetching the stored value, and updating with the resulting value.

:v:update-using (aq-) swap [ fetch swap call ] sip store ;

I have a simple word copy which copies memory to another location.

:copy   (aan-) [ &fetch-next dip store-next ] times drop drop ;

Lexical Scope

Now for something tricky: a system for lexical scoping.

The dictionary is a simple linked list. Retro allows for some control over what is visible using the {{, ---reveal---, and }} words.

As an example:

{{
  :increment dup fetch n:inc swap store ;
  :Value `0 ; data
---reveal---
  :next-number @Value &Value increment ;
}}

Only the next-number function will remain visible once }} is executed.

It's important to note that this only provides a lexical scope. Any variables are global (though the names may be hidden), so use v:preserve if you need reentrancy.

:ScopeList `0 `0 ;
:{{            (-)
  d:last dup &ScopeList store-next store ;
:---reveal---  (-)
   d:last &ScopeList n:inc store ;
:}}            (-)
  &ScopeList fetch-next swap fetch eq?
  [ @ScopeList !Dictionary ]
  [ @ScopeList
    [ &Dictionary repeat
        fetch dup fetch &ScopeList n:inc fetch -eq? 0; drop
      again ] call store ] choose ;

Linear Buffers

A buffer is a linear memory buffer. Retro provides a buffer: namespace for working with them.

This is something I've used for years. It's simple, but makes it easy to construct strings (as it writes a trailing ASCII null) and other simple structures.

| word            | used for                               |
| --------------- | -------------------------------------- |
| buffer:start    | return the first address in the buffer |
| buffer:end      | return the last address in the buffer  |
| buffer:add      | add a value to the end of the buffer   |
| buffer:get      | remove & return the last value         |
| buffer:empty    | remove all values from the buffer      |
| buffer:size     | return the number of stored values     |
| buffer:set      | set an address as the start of the     |
|                 | buffer                                 |
| buffer:preserve | preserve the current buffer pointers & |
|                 | execute a quotation that may set a new |
|                 | buffer. restores the saved pointers    |
|                 | when done                              |
{{
  :Buffer `0 ; data
  :Ptr    `0 ; data
  :terminate (-) #0 @Ptr store ;
---reveal---
  :buffer:start  (-a) @Buffer ;
  :buffer:end    (-a) @Ptr ;
  :buffer:add    (c-) buffer:end store &Ptr v:inc terminate ;
  :buffer:get    (-c) &Ptr v:dec buffer:end fetch terminate ;
  :buffer:empty  (-)  buffer:start !Ptr terminate ;
  :buffer:size   (-n) buffer:end buffer:start - ;
  :buffer:set    (a-) !Buffer buffer:empty ;
  :buffer:preserve (q-)
    @Buffer @Ptr [ [ call ] dip !Buffer ] dip !Ptr ;
}}

Strings

Traditional Forth systems have a messy mix of strings. You have counted strings, address/length pairs, and sometimes other forms.

Retro uses zero terminated strings. Counted strings are better in many ways, but I've used these for years and they are a workable approach. (Though caution in needed to avoid buffer overflow).

Temporary strings are allocated in a circular pool (STRINGS). This space can be altered as needed by adjusting these variables.

:TempStrings ;   data  #32 !TempStrings
:TempStringMax ; data #512 !TempStringMax
:STRINGS   EOM @TempStrings @TempStringMax * - ;

{{
  :Current `0 ; data

  :s:pointer (-p)  @Current @TempStringMax * STRINGS + ;
  :s:next    (-)
    &Current v:inc
    @Current @TempStrings eq? [ #0 !Current ] if ;
---reveal---
  :s:temp (s-s) dup s:length n:inc s:pointer swap copy
                s:pointer s:next ;
  :s:empty (-s) s:pointer s:next #0 over store ;
}}

Permanent strings are compiled into memory. To skip over them a helper function is used. When compiled into a definition this will look like:

i lica....
r s:skip
d 98
d 99
d 100
d 0

It'd be faster to compile a jump over the string instead. I use this approach as it makes it simpler to identify strings when debugging.

s:skip is the helper function which adjusts the Nga instruction pointer to skip to the code following the stored string.

:s:skip (-)
  pop [ fetch-next n:-zero? ] while n:dec push ;

:s:keep (s-s)
  compiling? [ &s:skip compile:call ] if
  here [ s, ] dip class:data ;

And now a quick ' prefix. (This will be replaced later). What this does is either move the string token to the temporary buffer or compile it into the current definition.

This doesn't support spaces. I use underscores instead. E.g.,

'Hello_World!

Later in the code I'll add a better implementation which can handle conversion of _ into spaces.

:prefix:' compiling? [ s:keep ] [ s:temp ] choose ; immediate

s:chop removes the last character from a string.

:s:chop (s-s) s:temp dup s:length over + n:dec #0 swap store ;

s:reverse reverses the order of a string. E.g.,

'hello  ->  'olleh
:s:reverse (s-s)
  [ dup s:temp buffer:set &s:length [ dup s:length + n:dec ] bi swap
    [ dup fetch buffer:add n:dec ] times drop buffer:start s:temp ]
  buffer:preserve ;

Trimming removes leading (s:trim-left) or trailing (s:trim-right) spaces from a string. s:trim removes both leading and trailing spaces.

:s:trim-left (s-s)
  s:temp [ fetch-next [ #32 eq? ] [ n:-zero? ] bi and ] while
  n:dec ;
:s:trim-right (s-s) s:temp s:reverse s:trim-left s:reverse ;
:s:trim (s-s) s:trim-right s:trim-left ;

s:prepend and s:append for concatenating strings together.

:s:prepend (ss-s)
  s:temp [ dup s:length + [ dup s:length n:inc ] dip swap copy ] sip ;
:s:append (ss-s) swap s:prepend ;

s:for-each executes a quote once for each cell in string. It is a key part of building the other high-level string operations.

:s:for-each (sq-)
  [ repeat
      over fetch 0; drop
      dup-pair
      [ [ &fetch dip call ] dip ] dip
      [ n:inc ] dip
    again
  ] call drop-pair ;

Building on s:for-each, I am able to implement s:index-of, which finds the first instance of a character in a string.

:s:index-of (sc-n)
  swap [ repeat
           fetch-next 0; swap
           [ over -eq? ] dip
           swap 0; drop
         again
       ] sip
  [ - n:dec nip ] sip
  s:length over eq? [ drop #-1 ] if ;

s:contains-char? returns a flag indicating whether or not a given character is in a string.

:s:contains-char? (sc-f) s:index-of #-1 -eq? ;

s:contains-string? returns a flag indicating whether or not a given substring is in a string.

{{
  'Src var
  'Tar var
  'Pad var
  'I   var
  'F   var
  'At  var

  :terminate (-)
    #0 @Pad @Tar s:length + store ;

  :extract (-)
    @Src @I + @Pad @Tar s:length copy ;

  :compare (-)
    @Pad @Tar s:eq? @F or !F @F [ @I !At ] -if ;

  :next (-)
    &I v:inc ;
---reveal---
  :s:contains-string? (ss-f)
    !Tar !Src s:empty !Pad #0 !I #0 !F
    @Src s:length
    [ extract terminate compare next ] times
    @F ;

  :s:index-of-string (ss-a)
    !Tar !Src s:empty !Pad #0 !I #0 !F #-1 !At
    @Src s:length
    [ extract terminate compare next ] times
    @F [ @At ] [ #-1 ] choose ;
}}

s:filter returns a new string, consisting of the characters from another string that are filtered by a quotation.

'This_is_a_test [ c:-vowel? ] s:filter
:s:filter (sq-s)
  [ s:empty buffer:set swap
    [ dup-pair swap call
        [ buffer:add ]
        [ drop       ] choose
    ] s:for-each drop buffer:start
  ] buffer:preserve ;

s:map Return a new string resulting from applying a quotation to each character in a source string.

'This_is_a_test [ $_ [ ASCII:SPACE ] case ] s:map
:s:map (sq-s)
  [ s:empty buffer:set swap
    [ over call buffer:add ]
    s:for-each drop buffer:start
  ] buffer:preserve ;

s:substr returns a subset of a string. Provide it with a string, a starting offset, and a length.

:s:substr (sfl-s)
  [ + s:empty ] dip [ over [ copy ] dip ] sip
  over [ + #0 swap store ] dip ;

s:right and s:left are similar to s:substr, but operate from fixed ends of the string.

:s:right (sn-s) over s:length over - swap s:substr ;
:s:left  (sn-s) #0 swap s:substr ;

Hash (using DJB2)

I use the djb2 hash algorithm for computing hashes from strings. There are better hashes out there, but this is pretty simple and works well for my needs. This was based on an implementation at http://www.cse.yorku.ca/~oz/hash.html

:s:hash (s-n) #5381 swap [ swap #33 * + ] s:for-each ;

Copy a string, including the terminator.

:s:copy (ss-) over s:length n:inc copy ;

RETRO provides string constants for several ranges of characters that are of some general interest.

:s:DIGITS          (-s)  '0123456789 ;
:s:ASCII-LOWERCASE (-s)  'abcdefghijklmnopqrstuvwxyz ;
:s:ASCII-UPPERCASE (-s)  'ABCDEFGHIJKLMNOPQRSTUVWXYZ ;
:s:ASCII-LETTERS   (-s)
  'abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ ;
:s:PUNCTUATION     (-s)
  '_!"#$%&'()*+,-./:;<=>?@[\]^`{|}~ $_ over store ;
's:WHITESPACE d:create
  #32, #9 , #10 , #13 , #0 ,

ASCII Constants

Not all characters can be obtained via the $ prefix. ASCII has many characters that aren't really intended to be printable. Retro has an ASCII namespace providing symbolic names for these.

#0  'ASCII:NUL const    #1   'ASCII:SOH const
#2  'ASCII:STX const    #3   'ASCII:ETX const
#4  'ASCII:EOT const    #5   'ASCII:ENQ const
#6  'ASCII:ACK const    #7   'ASCII:BEL const
#8  'ASCII:BS  const    #9   'ASCII:HT  const
#10 'ASCII:LF  const    #11  'ASCII:VT  const
#12 'ASCII:FF  const    #13  'ASCII:CR  const
#14 'ASCII:SO  const    #15  'ASCII:SI  const
#16 'ASCII:DLE const    #17  'ASCII:DC1 const
#18 'ASCII:DC2 const    #19  'ASCII:DC3 const
#20 'ASCII:DC4 const    #21  'ASCII:NAK const
#22 'ASCII:SYN const    #23  'ASCII:ETB const
#24 'ASCII:CAN const    #25  'ASCII:EM  const
#26 'ASCII:SUB const    #27  'ASCII:ESC const
#28 'ASCII:FS  const    #29  'ASCII:GS  const
#30 'ASCII:RS  const    #31  'ASCII:US  const
#32 'ASCII:SPACE const  #127 'ASCII:DEL const

These words operate on character values. Retro currently deals with ASCII, though cells are 32 bits in length, so Unicode values can be stored.

First are a bunch of words to help identify character values.

:c:letter?      (c-f) $A $z n:between? ;
:c:lowercase?   (c-f) $a $z n:between? ;
:c:uppercase?   (c-f) $A $Z n:between? ;
:c:digit?       (c-f) $0 $9 n:between? ;
:c:visible?     (c-f) #31 #126 n:between? ;
:c:vowel?       (c-f) 'aeiouAEIOU swap s:contains-char? ;
:c:consonant?   (c-f)
  dup c:letter? [ c:vowel? not ] [ drop FALSE ] choose ;

{{
  'WS d:create
    ASCII:SPACE , ASCII:HT , ASCII:LF , ASCII:CR , #0 ,
---reveal---
  :c:whitespace?  (c-f) &WS swap s:contains-char? ;
}}

And the inverse forms. (These are included for readability and orthiginal completion).

:c:-lowercase?  (c-f)  c:lowercase?  not ;
:c:-uppercase?  (c-f)  c:uppercase?  not ;
:c:-digit?      (c-f)  c:digit?      not ;
:c:-whitespace? (c-f)  c:whitespace? not ;
:c:-visible?    (c-f)  c:visible?    not ;
:c:-vowel?      (c-f)  c:vowel?      not ;
:c:-consonant?  (c-f)  c:consonant?  not ;

The next few words perform simple transformations.

:c:to-upper     (c-c) dup c:lowercase? 0; drop ASCII:SPACE - ;
:c:to-lower     (c-c) dup c:uppercase? 0; drop ASCII:SPACE + ;
:c:to-string    (c-s) '. s:temp [ store ] sip ;
:c:toggle-case  (c-c)
  dup c:lowercase? [ c:to-upper ] [ c:to-lower ] choose ;
:c:to-number    (c-n)
  dup c:digit? [ $0 - ] [ drop #0 ] choose ;

Back to Strings

With the character transformations a few more string words are possible.

:s:to-upper  (s-s)  [ c:to-upper ] s:map ;
:s:to-lower  (s-s)  [ c:to-lower ] s:map ;

Convert a decimal (base 10) number to a string.

{{
  'Value var
  :correct (c-c)
    dup $0 lt? [ $0 over - #2 * + ] if ; 
---reveal---
  :n:to-string  (n-s)
    [ here buffer:set dup !Value n:abs
      [ #10 /mod swap $0 + correct buffer:add dup n:-zero? ] while drop
      @Value n:negative? [ $- buffer:add ] if
      buffer:start s:reverse s:temp ] buffer:preserve ;
}}

Now replace the old prefix:' with this one that can optionally turn underscores into spaces.

TRUE 'RewriteUnderscores var<n>

{{
  :sub (c-c) $_ [ ASCII:SPACE ] case ;
  :rewrite (s-s)
    @RewriteUnderscores [ &sub s:map ] if ;
  :handle &prefix:' call ;
---reveal---
  :prefix:' rewrite handle ; immediate
  :prefix:" rewrite s:keep ; immediate
}}

The s:split splits a string on the first instance of a given character. Results are undefined if the character can not be located.

:s:split (sc-ss)
  dup-pair s:index-of nip dup-pair s:left [ + ] dip ;

:s:split-on-string (ss-ss)
  dup-pair s:index-of-string n:inc nip dup-pair s:left [ + ] dip ;

:s:replace (sss-s)
  over s:length here store
  [ s:split-on-string swap here fetch + ] dip s:prepend s:append ;

s:tokenize takes a string and a character to use as a separator. It splits the string into a set of substrings and returns a set containing pointers to each of them.

{{
  'Split-On var
  :match?    (c-f) @Split-On eq? ;
  :terminate (s-s) #0 over n:dec store ;
  :step      (ss-s)
    [ n:inc ] dip match? [ dup , terminate ] if ;
---reveal---
  :s:tokenize (sc-a)
    !Split-On s:keep
    here #0 , [ dup , dup [ step ] s:for-each drop ] dip
    here over - n:dec over store ;
}}

s:tokenize-on-string is like s:tokenize, but for strings.

{{
  'Tokens var
  'Needle var
  :-match?    (s-sf)
    dup @Needle s:contains-string? ;
  :save-token (s-s)
    @Needle s:split-on-string s:keep buffer:add n:inc ;
  :tokens-to-set (-a)
    here @Tokens buffer:size dup , [ fetch-next , ] times drop ;
---reveal---
  :s:tokenize-on-string (ss-a)
    [ s:keep !Needle here #8192 + !Tokens
      @Tokens buffer:set
      [ repeat -match? 0; drop save-token again ] call
      s:keep buffer:add tokens-to-set ] buffer:preserve ;
}}

Use s:format to construct a string from multiple items. This can be illustrated with:

#4 #6 #10  '%n-%n=%n\n  s:format

The format language is simple:

| \r | Replace with a CR                         |
| \n | Replace with a LF                         |
| \t | Replace with a TAB                        |
| \\ | Replace with a single \                   |
| \  | Replace with an underscore (_)            |
| \0 | Replace with NUL                          |
| %c | Replace with a character on the stack     |
| %s | Replace with a string on the stack        |
| %n | Replace with the next number on the stack |
{{
  :char (c-)
    ASCII:SPACE [ $_ buffer:add ] case
    $r [ ASCII:CR  buffer:add ] case
    $n [ ASCII:LF  buffer:add ] case
    $t [ ASCII:HT  buffer:add ] case
    $0 [ ASCII:NUL buffer:add ] case
    buffer:add ;

  :string (a-a)
    repeat fetch-next 0; buffer:add again ;

  :type (aac-)
    $c [ swap buffer:add              ] case
    $s [ swap string drop             ] case
    $n [ swap n:to-string string drop ] case
    drop ;

  :handle (ac-a)
    $\ [ fetch-next char ] case
    $% [ fetch-next type ] case
    buffer:add ;
---reveal---
  :s:format (...s-s)
    [ s:empty [ buffer:set
      [ repeat fetch-next 0; handle again ]
      call drop ] sip ] buffer:preserve ;
}}
:s:const (ss-)  [ s:keep ] dip const ;

The Ultimate Stack Shuffler

Ok, This is a bit of a hack, but very useful at times.

Assume you have a bunch of values:

#3 #1 #2 #5

And you want to reorder them into something new:

#1 #3 #5 #5 #2 #1

Rather than using a lot of shufflers, reorder simplfies this into:

#3 #1 #2 #5
'abcd  'baddcb reorder
{{
  'Values var #27 allot
  :from s:length dup [ [ &Values + store ] sip n:dec ] times drop ;
  :to dup s:length [ fetch-next $a -  n:inc &Values + fetch swap ] times drop ;
---reveal---
  :reorder (...ss-?) [ from ] dip to ;
}}

Extending The Language

does is intended to be paired with d:create to attach an action to a newly created data structure. An example use might be something like:

:constant (ns-)  d:create , [ fetch ] does ;

In a traditional Forth this is similar in spirit to DOES>.

:curry (vp-p)
  here [ swap compile:lit compile:call compile:ret ] dip ;
:does  (q-)
  d:last<xt> swap curry d:last d:xt store &class:word reclass ;

d:for-each is a combinator which runs a quote once for each header in the dictionary. A pointer to each header will be passed to the quote as it is run.

This can be used for implementing words:

[ d:name s:put sp ] d:for-each

Or finding the length of the longest name in the dictionary:

#0 [ d:name s:length n:max ] d:for-each

It's a handy combinator that lets me quickly walk though the entire dictionary in a very clean manner.

:d:for-each (q-)
  &Dictionary [ repeat fetch 0;
 dup-pair [ [ swap call ] dip ] dip again ] call drop ;

Using d:for-each, I implement a means of looking up a dictionary header by the d:xt field.

:d:lookup-xt (a-d)
 #0 swap [ dup-pair d:xt fetch eq?
           [ swap &nip dip ] &drop choose ] d:for-each drop ;

Arrays

Retro provides words for statically sized arrays. They are represented in memory as:

count
data #1 (first)
...
data #n (last)

Since the count comes first, a simple fetch will suffice to get it, but for completeness (and to allow for future changes), we wrap this as array:length:

:array:length (a-n) fetch ;

The first couple of words are used to create arrays. The first, array:counted-results executes a quote which returns values and a count. It then creates an array with the provided data.

:array:counted-results (q-a)
  call here [ dup , &, times ] dip ;

The second, array:from-string, creates a new string with the characters in given a string.

:array:from-string (s-a)
  here [ dup s:length , [ , ] s:for-each ] dip ;

A very crucial piece is array:for-each. This runs a quote once against each value in an array. This is leveraged to implement additional combinators.

{{
  'Q var
---reveal---
  :array:for-each (aq-)
    &Q [ !Q fetch-next
         [ fetch-next swap [ @Q call ] dip ] times drop
       ] v:preserve ;
}}

With this I can easily define array:dup to make a copy of an array.

:array:dup (a-a)
  here [ dup fetch , [ , ] array:for-each ] dip ;

Next is array:filter, which is extracts matching values from an array. This is used like:

{ #1 #2 #3 #4 #5 #6 #7 #8 }
[ n:even? ] array:filter

It returns a new array with the values that the quote returned a TRUE flag for.

:array:filter (aq-)
  [ over [ call ] dip swap [ , ] [ drop ] choose ] curry
  here [ over fetch , array:for-each ] dip
  here over - n:dec over store ;

Next are array:contains? and array:contains-string? which compare a given value to each item in the array and returns a flag.

{{
  'F var
---reveal---
  :array:contains? (na-f)
    &F v:off
    [ over eq? @F or !F ] array:for-each
    drop @F ;

  :array:contains-string? (sa-f)
    &F v:off
    [ over s:eq? @F or !F ] array:for-each
    drop @F ;
}}

I implemented array:map to apply a quotation to each item in an array and construct a new array from the returned values.

Example:

{ #1 #2 #3 }
[ #10 * ] array:map
:array:map (aq-a)
  [ call , ] curry
  here [ over fetch , array:for-each ] dip ;

You can use array:reverse to make a copy of an array with the values reversed.

:array:reverse (a-a)
  here [ fetch-next [ + n:dec ] sip dup ,
         [ dup fetch , n:dec ] times drop
       ] dip ;

array:nth provides a quick means of adjusting an array and offset into an address for use with fetch and store.

:array:nth (an-a)
  + n:inc ;

array:reduce takes an array, a starting value, and a quote. It executes the quote once for each item in the array, passing the item and the value to the quote. The quote should consume both and return a new value.

:array:reduce (pnp-n)
  [ swap ] dip array:for-each ;

When making an array, I often want the values in the original order. The array:counted-results array:reverse is a bit long, so I'm defining a new array:make which wraps these.

:array:make (q-a)
  array:counted-results array:reverse ;

:{ (-)  |[ |depth |[ ; immediate
:} (-a) |] |dip |depth |swap |- |n:dec |] |array:make ; immediate

Muri: an assembler

Muri is my minimalist assembler for Nga. This is an attempt to implement something similar in Retro.

This requires some knowledge of the Nga architecture to be useful. The major elements are:

Instruction Set

Nga has 27 instructions. These are:

0  nop        7  jump      14  gt        21  and
1  lit <v>    8  call      15  fetch     22  or
2  dup        9  ccall     16  store     23  xor
3  drop      10  return    17  add       24  shift
4  swap      11  eq        18  sub       25  zret
5  push      12  neq       19  mul       26  end
6  pop       13  lt        20  divmod

The mnemonics allow for each name to be reduced to just two characters. In the same order as above:

no  ju  gt  an
li  ca  fe  or
du  cc  st  xo
dr  re  ad  sh
sw  eq  su  zr
pu  ne  mu  en
po  lt  di

Up to four instructions can be packed into a single memory location. (You can only use nop after a jump, call, ccall, ret, or zret as these alter the instruction pointer.)

So a bundled sequence like:

lit 100
lit 200
add
ret

Would look like:

'liliadre i
100       d
200       d

And:

lit s:eq?
call

Would become:

'lica.... i
's:eq?    r

Note the use of .. instead of no for the nop's; this is done to improve readability a little.

Instruction bundles are specified as strings, and are converted to actual instructions by the i word. As in the standard Muri assembler, the RETRO version uses d for decimal values and r for references to named functions.

This is kept in the global namespace, but several portions are kept private.

{{

I allocate a small buffer for each portion of an instruction bundle.

  'I0 d:create #3 allot
  'I1 d:create #3 allot
  'I2 d:create #3 allot
  'I3 d:create #3 allot

The opcode word maps a two character instruction to an opcode number.

  :opcode (s-n)
    '.. [ #0  ] s:case  'li [ #1  ] s:case
    'du [ #2  ] s:case  'dr [ #3  ] s:case
    'sw [ #4  ] s:case  'pu [ #5  ] s:case
    'po [ #6  ] s:case  'ju [ #7  ] s:case
    'ca [ #8  ] s:case  'cc [ #9  ] s:case
    're [ #10 ] s:case  'eq [ #11 ] s:case
    'ne [ #12 ] s:case  'lt [ #13 ] s:case
    'gt [ #14 ] s:case  'fe [ #15 ] s:case
    'st [ #16 ] s:case  'ad [ #17 ] s:case
    'su [ #18 ] s:case  'mu [ #19 ] s:case
    'di [ #20 ] s:case  'an [ #21 ] s:case
    'or [ #22 ] s:case  'xo [ #23 ] s:case
    'sh [ #24 ] s:case  'zr [ #25 ] s:case
    'en [ #26 ] s:case  'ie [ #27 ] s:case
    'iq [ #28 ] s:case  'ii [ #29 ] s:case
    drop #0 ;

I use pack to combine the individual parts of the instruction bundle into a single cell.

  :pack (-n)
    &I0 opcode  &I1 opcode
    &I2 opcode  &I3 opcode
    #-24 shift   swap
    #-16 shift + swap
    #-8  shift + swap + ;

Switch to the public portion of the code.

---reveal---

With this it's pretty easy to implement the instruction bundle handler. Named i, this takes a string with four instruction names, splits it into each part, calls opcode on each and then pack to combine them before using , to write them into the Heap.

  :i (s-)
    dup &I0 #2 copy #2 +
    dup &I1 #2 copy #2 +
    dup &I2 #2 copy #2 +
        &I3 #2 copy
    pack , ;

The d word inlines a data item.

  :d (n-)  , ;

And r inlines a reference (pointer).

  :r (s-)  d:lookup d:xt fetch , ;

The final bits are as{ and }as, which start and stop the assembler. (Basically, they just turn the Compiler on and off, restoring its state as needed).

  :as{ (-f)
    @Compiler &Compiler v:off ; immediate

  :}as (f-?)
    !Compiler ; immediate

This finishes by sealing off the private words.

}}

Evaluating Source

The standard interfaces have their own approaches to getting and dealing with user input. Sometimes though it'd be nicer to have a way of evaluating code from within RETRO itself. This provides an evaluate word.

{{

First, create a buffer for the string to be evaluated. This is sized to allow for a standard FORTH block to fit, or to easily fit a traditional 1024 character block. It's also long enough for most source lines I expect to encounter when working with files.

I allocate this immediately prior to the temporary string buffers.

  :current-line (-a) STRINGS #1025 - ;

To make use of this, we need to know how many tokens are in the input string. count-tokens takes care of this by filtering out anything other than spaces and getting the size of the remaining string.

  :count-tokens (s-n)
    [ ASCII:SPACE eq? ] s:filter s:length ;

The next-token word splits the remainimg string on SPACE and returns both parts.

  :next-token (s-ss)
    ASCII:SPACE s:split ;

And then the process-tokens uses next-token and interpret to go through the string, processing each token in turn. Using the standard interpret word allows for proper handling of the prefixes and classes so everything works just as if entered directly.

  :process-tokens (sn-)
    [ next-token swap
      [ dup s:length n:-zero?
            &interpret &drop choose ] dip n:inc
    ] times interpret ;
---reveal---

And finally, tie it all together into the single exposed word evaluate.

  :s:evaluate (s-...)
    current-line s:copy
    current-line dup count-tokens process-tokens ;
}}

Loops, continued

Sometimes it's useful to be able to access a loop index. The next word, times<with-index> adds this to RETRO. It also provides I, J, and K words to access the index of the current, and up to two outer loops as well.

{{
  'LP var
  'Index d:create #128 allot
  :next (-) @LP &Index + v:inc ;
  :prep (-) &LP v:inc #0 @LP &Index + store ;
  :done (-) &LP v:dec ;
---reveal---
  :I (-n) @LP &Index + fetch ;
  :J (-n) @LP &Index + n:dec fetch ;
  :K (-n) @LP &Index + #2 - fetch ;
  :times<with-index>
    prep swap
      [ repeat 0; n:dec push &call sip pop next again ] call
    drop done ;
}}

Hooks

In RETRO 11, nearly all definitions could be temporarily replaced by leaving space at the start for compiling in a jump. In the current RETRO I do not do this, though the technique is still useful, especially with I/O. These next few words provide a means of doing this in RETRO 12.

To allow a word to be overridden, add a call to hook as the first word in the definition. This will compile a jump to the actual definition start.

:hook (-)  #1793 , here n:inc , ; immediate

set-hook takes a pointer to the new word or quote and a pointer to the hook to replace. It alters the jump target.

:set-hook (aa-) n:inc store ;

The final word, unhook, resets the jump target to the original one.

:unhook (a-) n:inc dup n:inc swap store ;

I/O

:io:enumerate (-n)   as{ 'ie...... i }as ;
:io:query     (n-mN) as{ 'iq...... i }as ;
:io:invoke    (n-)   as{ 'ii...... i }as ;

{{
  'Slot var
---reveal---
  :io:scan-for  (n-m)
    #-1 !Slot
    io:enumerate [ I io:query nip over eq?
                   [ I !Slot ] if ] times<with-index>
    drop @Slot ;
}}

A Retro system is only required to provide a single I/O word to the user: a word to push a single character to the output log. This is always mapped to device 0, and is exposed as c:put.

:c:put (c-) hook #0 io:invoke ;

This can be used to implement words that push other items to the log.

:nl    (-)   ASCII:LF    c:put ;
:sp    (-)   ASCII:SPACE c:put ;
:tab   (-)   ASCII:HT    c:put ;
:s:put (s-)  &c:put s:for-each ;
:n:put (n-)  n:to-string s:put ;

An interface layer may provide additional I/O words, but these will not be implemented here as they are not part of the core language.

Debugging Aids

I provide a few debugging aids in the core language. The examples provide much better tools, and interface layers can provide much more than I can do here.

:d:words    (-)  [ d:name s:put sp ] d:for-each ;
:reset      (...-) depth repeat 0; push drop pop #1 - again ;
:dump-stack (-)  depth 0; drop push dump-stack pop dup n:put sp ;

From Kiyoshi Yoneda, this is a variant of d:words which displays words containing a specific substring. It's useful to see words in a specific namespace, e.g., by doing 's: d:words-with, or words that likely display something: ':puts d:words-with.

{{
  :display-if-matched  (s-)
    dup here s:contains-string? [ s:put sp ] [ drop ] choose ;
---reveal---
  :d:words-with (s-)
    here s:copy [ d:name display-if-matched ] d:for-each ;
}}
:FREE (-n) STRINGS #1025 - here - ;

The End

Legalities

Permission to use, copy, modify, and/or distribute this software for any purpose with or without fee is hereby granted, provided that the copyright notice and this permission notice appear in all copies.

THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.

Copyright (c) 2008 - 2019, Charles Childers
Copyright (c) 2012 - 2013, Michal J Wallace
Copyright (c) 2009 - 2011, Luke Parrish
Copyright (c) 2009 - 2010, JGL
Copyright (c) 2010 - 2011, Marc Simpson
Copyright (c) 2011 - 2012, Oleksandr Kozachuk
Copyright (c) 2018,        Kiyoshi Yoneda
Copyright (c) 2010,        Jay Skeer
Copyright (c) 2010,        Greg Copeland
Copyright (c) 2011,        Aleksej Saushev
Copyright (c) 2011,        Foucist
Copyright (c) 2011,        Erturk Kocalar
Copyright (c) 2011,        Kenneth Keating
Copyright (c) 2011,        Ashley Feniello
Copyright (c) 2011,        Peter Salvi
Copyright (c) 2011,        Christian Kellermann
Copyright (c) 2011,        Jorge Acereda
Copyright (c) 2011,        Remy Moueza
Copyright (c) 2012,        John M Harrison
Copyright (c) 2012,        Todd Thomas