Re: Trying to extend gforth's outer interpreter

[Anton Ertl]

On Fri, Apr 01, 2022 at 09:51:09AM +0100, Tristan Williams wrote:
> Hello,
> 
> I am trying to extend gforth's outer interpreter so that how numbers
> are treated is dependent on which mode (indicated by the value of
> mode?) my program is in. In the code below, if mode? is true the
> number is just dropped, otherwise it should be treated as usual and
> placed on the host data stack. This seems to work if I type
> definitions in interactively at the terminal, but if I include them at
> the end of the source file or interactively using include "filename"
> it does not. What am I doing wrong?

You hook into 'quit, which is only used for text-interpretation of
terminal input.

If you use a recent development version of Gforth, the way to go is to
define a new number recognizer, and replace the default number
recognizer REC-NUM with the new one (or just insert the new one in
front of REC-NUM in the recognizer stack).  The documentation is not
finished yet, but look at the source of REC-NUM, and use
GET-RECOGNIZERS and SET-RECOGNIZERS to change the recognizer stack.

If you use Gforth-0.7, things are more complex: You need to replace
INTERPRETER1 and COMPILER1.  These are not hooks, so you need to patch
the existing colon definitions.  To do that, use:

: >colon-body ( xt -- addr )
    dup @ docol: <> -12 and throw >body ;

: >prim-code ( xt -- x )
    \ converts xt of a primitive into a form usable in the code of
    \ colon definitions on the current engine
    threading-method 0= IF @ THEN ;

: replace-word ( xt1 xt2 -- ) \ gforth
  \G make xt2 do xt1, both need to be colon definitions
    swap >colon-body ['] branch >prim-code rot >colon-body 2! ;

- anton