[LANG] 2.5 Random (mis)behaviour

[sebb <se...@gmail.com>]

I'm hopefully near to finding a version of nextLong(long) that
distributes the numbers better.
There's no such functionality in Random (even in Java 6) so that might
even be a useful addition for Lang3.

But there are some other aspects of RandomUtils and JVMRandom classes
that seem plain wrong in the current 2.4 release:

RandomUtils.nextInt() returns a number in the range [0,Integer.MAX_VALUE)

however

RandomUtils.nextInt(Random) returns a number in the range
[Integer.MIN_VALUE,Integer.MAX_VALUE]

(unless Random is JVMRandom of course)

Similarly for nextLong().

This is not documented for 2.4 and seems like an accident of the
original implementation, rather than a conscious design decision. I've
added some Javadoc to trunk to document the current behaviour, but I
am now wondering whether it would be OK to fix the behaviour so it
agrees with the 2.4 Javadoc?

Also, given that JVMRandom now relies on a static copy of Random, it
would be trivial to implement some of the missing functionality, such
as:

    public static void nextBytes(byte[]);
    public static synchronized double nextGaussian();

with corresponding changes to the instance methods in JVMRandom.

Should this be done, or is it better to leave the methods unimplemented?

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
For additional commands, e-mail: dev-help@commons.apache.org

Re: [LANG] 2.5 Random (mis)behaviour

[Niall Pemberton <ni...@gmail.com>]

On Fri, Feb 5, 2010 at 3:15 PM, sebb <se...@gmail.com> wrote:
> I'm hopefully near to finding a version of nextLong(long) that
> distributes the numbers better.
> There's no such functionality in Random (even in Java 6) so that might
> even be a useful addition for Lang3.
>
> But there are some other aspects of RandomUtils and JVMRandom classes
> that seem plain wrong in the current 2.4 release:
>
> RandomUtils.nextInt() returns a number in the range [0,Integer.MAX_VALUE)
>
> however
>
> RandomUtils.nextInt(Random) returns a number in the range
> [Integer.MIN_VALUE,Integer.MAX_VALUE]
>
> (unless Random is JVMRandom of course)
>
> Similarly for nextLong().
>
> This is not documented for 2.4 and seems like an accident of the
> original implementation, rather than a conscious design decision. I've
> added some Javadoc to trunk to document the current behaviour, but I
> am now wondering whether it would be OK to fix the behaviour so it
> agrees with the 2.4 Javadoc?

I would say fix the javadoc rather than change the behaviour

> Also, given that JVMRandom now relies on a static copy of Random, it
> would be trivial to implement some of the missing functionality, such
> as:
>
>    public static void nextBytes(byte[]);
>    public static synchronized double nextGaussian();
>
> with corresponding changes to the instance methods in JVMRandom.
>
> Should this be done, or is it better to leave the methods unimplemented?

I think we should only be adding new features that are also going into
the trunk. Since RandomUtils & JVMRandom have been deleted from trunk
then I don't think we should do this.

Niall

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
For additional commands, e-mail: dev-help@commons.apache.org