Re: The future of Spark . Spark 2014 : a wreckage

["Peter C. Chapin" <PChapin@vtc.vsc.edu>]

On 07/08/2013 01:56 PM, vincent.diemunsch@gmail.com wrote:

>     procedure Swap
>        with Global  => (In_Out => (X, Y)),
>             Depends => (X => Y,   -- to be read as "X depends on Y"
>                         Y => X);  -- to be read as "Y depends on X"
>
> How are we supposed to guess that "X => Y" means X depends on Y, if the arrow goes from X to Y ?

Well the phrase "X depends on Y" has X on the left and Y on the right 
just as X => Y does. So it reads fairly naturally to me. I understand 
there is value in being consistent with existing notations although I 
get the impression there is already conflicting practice in that area.

More significantly, though, I think you may be assigning too much 
semantic weight to the '=>' token in this case. In the aspect syntax it 
is used more as a separator than anything else. Consider

   Pre => X < 0

Nobody would say that there is information flowing from "Pre" to the 
expression 'X < 0' or visa-versa. The SPARK 2014 syntax extends this 
notion of "arrow as separator" to the subcomponents of the aspects. Thus

   Depends => (X => Y)

The "inner" arrow just separates X from Y syntactically. It's not really 
related to information flow... although its easy to see why someone 
would assume that it was considering the semantics of the Depends aspect 
overall.

Peter