argument checking, was: About submatrix and its meaning

Robert Dodier wrote:
> On 11/23/09, Richard Fateman <fateman at cs.berkeley.edu> wrote:
>
>   
>> foo(a,b,c):=  (
>>    check_args(foo,[a,listp],[b,integerp],[[a,b,c], a+b>c]),
>>     
>
> This is needlessly clumsy and incomprehensible.
>   

Well, you could add more, and indent, like

check_args(foo,
  assert(listp(a)),
  assert(integerp(b)),
  assert(a+b>c))

and you could have optional clauses to indicate error messages, or what 
to do with failure, etc.

A remark:

in 1969 or shortly thereafter, I proposed extra syntax for matchdeclare, 
tellsimp, tellsimpafter, which was
rejected (by Joel Moses, who thought...  needs extra syntax).
So instead of something like..

f(a:true*x+b:true) ->  g(a)+b

we got

matchdeclare([a,b],true)
tellsimp(f(a*x+b, g(a)+b).

compare this to Mathematica, which uses lots of extra syntax for patterns.

Anyway, it depends on what you are used to.  I think the lisp 
perspective tends toward "no new syntax".

RJF





> check_args merges into the body of the function where it is
> easily overlooked, declarations for all variables are mashed
> together, and there is no syntactic clue that the variables are
> different from the predicates.
>
> Yes, it's reminiscent of the syntax for other declarations,
> but we needn't propagate it any further.
>
> I could be convinced that assert(P(x)) is OK, though.
>
>   
>> The addition of more "syntax" is certainly not required.
>>     
>
> Few things are necessary. Where does that leave us?
>
> Robert Dodier
>