I was writing a proof of a*0=0
and I stumbled on some strangeness. Why does the sif a >= 0
on line 7 need to be a >=
, and does not compile when its just sif > 0
?
prfn mul_ax0_0 {a:int} () : MUL(a,0,0) =
let
prfun auxnat {a:nat} .<a>. () : MUL(a,0,0) =
sif a == 0 then MULbas()
else MULind(auxnat{a-1}())
in
sif a >= 0 then auxnat{a}() // line 7
else MULneg(auxnat{~a}())
end
implement main0 () = ()
Intuitively, the a=0
should be handled fine by either path, yet only the first path works.
Why?
MULneg is declared as follows:
| {m:pos}{n:int}{p:int}
MULneg (~(m), n, ~(p)) of MUL_prop (m, n, p)
Note that 'm' is required to be positive. In your case, 'm' is '~a'. If you use '>' instead of '>=', then it cannot be inferred that '~a > 0' holds when the test 'a > 0' fails.