Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.lang.forth > #39542

Re: Macros based [CONTEXT]-EVALUATE

Show all headers | View raw

on 03/06/2015 08:12:11,  wrote:
> Using EVALUATE for macros is either buggy, because the EVALUATE
> happens in a different context than the definition of the macro, or
> cumbersome, because you have to set up the context for EVALUATE
> manually.
> 
> I have thought about how setting up the context could be automated, at
> least partially: for the dictionary and BASE:
> 
> One would define a macro as follows:
> 
>: foo s" 16 +" [context]-evaluate ; immediate
> 
> During compilation, [CONTEXT]-EVALUATE saves the BASE, the search
> order, and HERE.  During run-time, [CONTEXT]-EVALUATE sets the BASE
> and the search order to the saved values, and sets the FIND limit to
> HERE, then EVALUATEs, and then restores the previous values of BASE,
> search order and FIND limit.  The FIND limit influences dictionary
> searches such that only words below the FIND limit are found (so that
> later redefinitions of a word used inside the macro are not found).

That won't work for systems where HERE (or a plurality of HEREs for multi
section Forths) don't conform to a increments-thru-time model. For
instance, the dictionary in Win32Forth and others may be external; a
library, an SQL database and so on. It may be relocated, optimised, built
out of non-ascending lumps, garbage collected or even poofed out of
existence for turnkeys.

As was made clear in the case of TRAVERSE-WORDLIST, about all we can
reasonably ask is that definitions for the same name are returned in
*time sequenced* order, not address order. "TRAVERSE- WORDLIST may visit
words in any order, with one exception: words with the same name are
called in the order newest-to-oldest (possibly with other words in
between)."

> 
> The macro could also depend on more context (maybe some words access
> some variables during compilation), and that additional context still
> has to be set up manually, but at least the common and, for dictionary
> searches, cumbersome cases would be covered.
> 
> What would be needed to implement [CONTEXT]-EVALUATE, that systems
> don't have at the moment is the FIND limit.  The rest can be defined
> in standard Forth.
> 
> The question is if there is enough interest in these kinds of macros
> to convince system implementors (including me) to implement at least
> the FIND limit.

I have not implemented FORGET or MARKER because of these difficulties;
but then, I don't claim 100% ANS compliance either, so my interest is a
moot point.

> 
> - anton

Back to comp.lang.forth | Previous | Next — Previous in thread | Next in thread | Find similar | Unroll thread

Thread

Macros based [CONTEXT]-EVALUATE anton@mips.complang.tuwien.ac.at (Anton Ertl) - 2015-06-03 07:12 +0000
  Re: Macros based [CONTEXT]-EVALUATE "Alex McDonald" <blog@rivadpm.com> - 2015-06-03 14:50 +0100
    Re: Macros based [CONTEXT]-EVALUATE anton@mips.complang.tuwien.ac.at (Anton Ertl) - 2015-06-03 14:01 +0000
      Re: Macros based [CONTEXT]-EVALUATE "Alex McDonald" <blog@rivadpm.com> - 2015-06-03 21:21 +0100
  Re: Macros based [CONTEXT]-EVALUATE humptydumpty <ouatubi@gmail.com> - 2015-06-03 09:02 -0700
    Re: Macros based [CONTEXT]-EVALUATE hughaguilar96@gmail.com - 2015-06-03 14:07 -0700

csiph-web