Re: @value, macro and other expansion questions

[Patrice Dumas]

On Fri, Aug 22, 2008 at 09:20:11PM -0500, Karl Berry wrote:
>     expansion in @-commands arguments.
> 
> I think I agree with everything, except there is one complication: for
> nodes and anchors, there is no question that it would be very useful to
> allow some @-commands (especially accents).  Eli more or less
> implemented this for makeinfo with --commands-in-node-names.

In fact, I tried to look at what --commands-in-node-names did, and, as
far as I can tell @-commands are more or less already expanded in nodes,
the only thing that --commands-in-node-names does is to remove some 
error messages linked with @value{} not expanded in node names at some
point where some validation is done. Did I missed something?
 
> The problem is TeX.  The complications of getting TeX to expand some
> commands within node names., to just the right level, are myriad.  This
> is the main reason for the present rule of "no @-commands in node names".

Indeed, but this is merely an implementation limitation, not really a
texinfo feature. I agree that the texinfo manual should enforce the
lowest common constructs, though.

> Oleg, if we carefully specified the list of allowed commands in @node
> (starting with, say, @'), do you think there's a chance we could make it
> work in texinfo.tex?  I haven't looked into it in a long time.

If I remember well, accented letters in node names work in TeX (even
@,{} that doesn't work in makeinfo), but I don't know for other commands.

> The other processors could conceivably allow other commands.  And your
> expansion rules could be followed in the case of
> --commands-in-node-names, anyway.

Do you agree on node names being in code_style? This is one thing I am
not sure about.

> As for the things you marked debatable:
> 
>     2 in @-commands taking a filename as argument on the line, namely
>       @setfilename, @include and @verbatiminclude
>       my personal opinion is that the 
> 
>         @value, commands defined by @macro and other @-commands should be 
>         expanded, in code_style. 
> 
> I definitely agree.  It is nonsensical to do anything but code_style in
> filenames.  In fact, I recall recently making a bug fix related to this
> in response to a GCC bug report.

Ok. I'll report what I find.

>     6 I think that everything on a @def* line 
> 
>         should have @values and @-command defined by @macros expanded, and, 
>         in my opinion, be in code_style. There should only be line @-commands.
>         It is not completly clear to me if text within style @-commands that
>         are not in code_style, like @var, should be in code_style. 
> 
> Again I agree.  There is no use in doing anything but code_style in
> @def* lines.

Ok. This is not what is done curerntly in makeinfo. I'll also report
what I find.

>     9 I don't know if @key should be in code_style, but in most cases it
>       appears in @kbd and should 'inherit' the code_style.
> 
> I don't think the question arises, because keyboard key names never
> contain expandable constructs.  Do they?  At least I don't see any on my
> US keyboard.  But in principle, I think code_style would be desirable
> since again, the user is going to writing literally what should appear
> and won't expect Texinfo to be changing it.

It wouldn't be impossible to have something like '' in a keyboard
layout, for example, maybe meaning a double acute accent or something like
that.

--
Pat