Re: kawa java.util.Random

[Per Bothner <per at bothner dot com>]

On 10/30/2015 11:40 AM, Debabrata Pani wrote:
> Sorry, the previous mail was sent even before I could complete it.
>
> To continue..
>
> While using java random numbers, I am getting the following warning
>
> #|kawa:43|# (define xx (java.util.Random (java.lang.System:currentTimeMillis)))
> #|kawa:44|# xx
> java.util.Random@5bfbf16f
> #|kawa:45|# (xx:nextInt 10)
> /dev/stdin:45:2: warning - no known slot 'nextInt' in java.lang.Object
> 8
> #|kawa:46|# xx
>
> This is observable in both kawa 2.1 and kawa 2.0
>
> Is this a known/expected behavior ?

The reason is the Kawa compiler can't know that you won't later do
something like:
  (set! xx "foo")
so it can't safely infer the type.  (It could in the case where nextInt is
called directly at the top-level, but in the general case if the nextInt
call is inside a function, which might be called at a later time.)

Do any one of:

(define-constant xx (java.util.Random (java.lang.System:currentTimeMillis)))
(! xx (java.util.Random (java.lang.System:currentTimeMillis)))
(define xx ::java.util.Random (java.util.Random (java.lang.System:currentTimeMillis)))

Using '!' is more-or-less equivalent to define-constant.

I'm considering changing Kawa so that the type of xx in the plain-define case
would be 'dynamic' rather than 'java.lang.Object'.  That will suppress the warning.
Alternatively, this could be part of the "interactive mode" improvements
I'm also considering.
--
	--Per Bothner
per@bothner.com   http://per.bothner.com/