Re: [Groff] proposal: add variable traps to groff.

[Werner LEMBERG]

> In addition several number registers are changed by special purpose
> requests.  For example the read only \n[.p] is set using .pl.

>From an internal point of view, this is not true.  The `pl' request
sets a variable in the top diversion structure (to specify the page
length), and as soon as the user accesses the `.p' register, this
variable is directly read.

> This is I guess because the groff implementation itself has to
> maintain a .vpt at the bottom of the page, so groff must notice when
> \n[.p] changes.

groff tests after each line whether the current position is greater
than the actual page length (or another vertical trap).  It uses the
internal variable, not the `.p' register.

> So the request .pl is the groff engine wrapper macro for .nr .p.

Well, you could interpret the groff engine that way, but it probably
hides the real mechanism too much.

> I am using tcl quite a lot nowadays and I found its variable traps
> useful in handling multiple access to a single variable.  Maybe this
> is why I thought about having register traps in groff.  This might
> be a dumb idea, I do not know.  I am just a macro user, so I might
> completely miss the point here.

At the cost of speed you already can trace such registers by
redefining the corresponding requests:

  .rn pl pl-orig
  .de pl
  .  pl-orig \\$1
  .  nr .pl-debug \\n[.p]
  ..

What you can't trace are escapes like \n[...] or \*[...], so it
actually makes sense to have your `.wr' and `.ws' requests.  I've
saved your proposal in my `groff' directory, but I can't promise any
realization in the near future.


    Werner