Re: [Groff] parsing a corner specification

[Joel E. Denny]

On Sun, 26 Aug 2007, Keith Marshall wrote:

> Even if this were true, it makes no sense in geometric terms, to refer 
> to the `left corner' of an object.  What, exactly, should that be 
> interpreted to mean?
> 
> In your example, object `A' is a circle; it doesn't have *any* corners!  

Thanks for your input.  I apologize for my confusing terminology.  I 
realize of course that, geometrically speaking, circles have no corners.  
I have been using the term "corner" in the way that all of the pic 
grammars I've looked at use it.  In that sense, all objects have multiple 
corners.

> additionally, as Ted has already stated, the correct
> syntax is `A.left', or `at left of A', (strictly, this should probably 
> be `at .left of A'), but not `A left'.  Your example can be written, 
> much more intelligibly, as
> 
>   .PS
>   A: circle "A"
>   B: circle "B" at A.left
>   .PE
> 
> or as
> 
>   .PS
>   A: circle "A"
>   B: circle "B" at left of A
>   .PE
> 
> Why do we need to accept the more confusing form of expression?

I've been asking two questions, which I had hoped had the same answer.

In my last email, I was focused mostly on what the grammar author 
originally intended.  This is interesting to me because I wish to 
determine if I've found a real-world example of how canonical LR(1) would 
help in the development of a grammar.  I believe I've found significant 
evidence in gpic's source comments that canonical LR(1) would help achieve 
what the author intended.

My other question is what the current developers, community, and any pic 
specifications say is the desired pic behavior.  It seems they are opposed 
to this alternative use of "left" and "right".