RFR 8023155: Ensure functional consistency across Random, ThreadLocalRandom, SplittableRandom

[Mike Duigou]

On Aug 19 2013, at 13:12 , Guy Steele wrote:

> 
> On Aug 19, 2013, at 3:17 PM, Mike Duigou <mike.duigou at oracle.com> wrote:
> 
>> - I find disallowing the zero bounds and empty ranges slightly annoying. It requires me to externally special case for situations such as:
>> 
>> Random ran = new Random();
>> String[] users = {"Fred"}; 
>> 
>> someUser = users[mine.nextInt(users.length - 1)];
>> 
>> This is a frequently used idiom. Yes, forcing the random number generator to return zero is silly but for symmetry it is convenient. An empty range isn't an obvious error (though the "String[] users = {};" case is obviously an error).

> If you want a free choice among ALL elements in the (non-empty) array, then
> 
>   someUser = users[mine.nextInt(users.length)];
> 
> is what you want; and if the array is empty, then you'll get an exception from the nextInt method rather than the array indexing step, but that amounts to the same thing: an inability to pick an element from an empty array.

Doh, yes. I knew it was important to document that bounds was exclusive--I managed to confuse myself....

Mike