Re: [Groff] Re: begin page blues

[Miklos Somogyi]

On 14/03/2006, at 3:46 AM, Werner LEMBERG wrote:

> > .ev 1                   \" 1a) why is this necessary?
> > .di x                   \" 1b)
> > .  trf psline.ps	\" 1c) dumps psline.ps with \\$1 replaced onto macro 
> > x?
> > .di                     \" 1d)
> > .ev                     \" 1e)
>
> > Why are 1a) and 1e) necessary?
>
> Switching to another environment protects the current partial output
> line.  Consider this:
>
>   foo
>   .di x
>   I'm in a diversion.
>   .br
>   .di
>   bar
>   .di x
You mean `.x' I presume.
>      => bar foo I'm in a diversion.
>
> And now this:
>
>   foo
>   .ev 1
>   .di x
>   I'm in a diversion.
>   .br
>   .di
>   .ev
>   bar
>   .di x
>
>      => foo bar I'm in a diversion.

Werner, this should be part of the `.di' documentation. With an 
explanation of what `.br' is
doing here (or possibly before `I' m in a diversion').
`.di' documentation does not even mention how to `download' the macro's 
content.
Yes, there are useful things in `.box' but one might not notice that it 
should be read
in conjunction with `.di'.


> > .de line                \" 4a)
> > .  de line-aux END      \" 5a)
> >      ps: def \\*x       \" 5b) appends `x' to prologue
> > .  END
> > .
> > .  nop \Y[line-aux]             \" 4b) ??
> > .  nop \X'ps: exec psline'      \" 4c) ??
> > ..
>
> > What exactly happens at 4b) and 4c) ?
>
> The `line' macro defines another macro, `line-aux'.  During execution
> of `line', normal register, macro, and string expansion takes place.
> This means that `line-aux' no longer contains expandable material
> (except if you you protect it with an additional level of backslashes,
> something like `\\\\*x').  The \Y escape takes the contents of
> `line-aux' and inserts it into the output data stream as a special
> node which is eventually emitted as `x X ...' in the intermediate
> output.  \X does the same; it embeds its argument into a special node
> in the same way as \Y (but no newlines are allowed in the argument).
>
> The `x X' commands in the intermediate output must start with `ps:
> xxx'; otherwise grops don't recognize them.
>
> > .\"--- external file psline.ps
> > .\"
> > .\" /psline {
> > .\"   \$1 0 rlineto
> > .\"   stroke
> > .\" } bind def
> > .\"
> > .\"--- questions
> > .\"
> > .\"   - Where (and in which pass) does the $1 replacement occur?
>
> An external file read in with .trf is copied verbatim -- backslashes
> are not interpreted as escape characters.  The .di request doesn't
> interpret anything by itself; it just diverts the data stream to a
> diversion.  This means that \$1 is expanded one level higher, during
> the call of the `line' macro.

Oh jeez, there's a lot to learn!


> > .\" - Why can't `..' and `.  END' have side comments like anything
> >       else?
>
> This is possible.  How do you come to a different conclusion?

Yeah, you are right. It's not `\" comment'  that causes problems, but 
the TABS before it ...

>   .de foo
>   bar
>   ..       \" a comment
>   .
>   foo
>   .foo
>   foo
>
>       => foo bar foo
>
> > .\"        Why can't they handle trailing tabs?
>
> This is a known limitation.  To cite section 7.3 of the original AT&T
> troff reference:
>
>   The argument separator is the space character (not tab) [...]
>
> Any line starting with a dot is treated as either a request or as a
> macro.  You can even do something perverse like this:
>
>   .de .
>   bar
>   ..
>   .
>   foo
>   ..
>   foo
>
>       => foo bar foo
>
> With other words, `..' is basically handled as an (undefined) macro
> which finishes the `.de' request.
>
> If you input `..<tab>', groff doesn't interpret this as `..' because a
> tab doesn't separate a macro name from its arguments.
>
> In general, groff doesn't handling trailing whitespace.  This is the
> very reason why you always should say
>
>   .ds foo blablabla\"
>
> to avoid an accidentally inserted whitespace at the end of the string
> definition.

I'll remember that!

>     Werner

Thank you for these (lots of) clarifications. I seem to fall into every 
pothole between
SGI troff and groff.

Miklos