commons-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Phil Steitz <>
Subject Re: [math] math 1.1 RCs and FCS
Date Fri, 14 Oct 2005 12:59:06 GMT
On 10/14/05, Mauro Talevi <> wrote:
> Phil,
> Phil Steitz wrote:
> >
> >
> > There are two open bugs holding up the release:
> > (trivial)
> > (requires some work)
> >
> > I will try to get these sorted this weekend.  Patches, feedback
> > welcome!  Any [math] committers out there, pls review mails and bug
> > report on Complex classes and weigh in so we can resolve and get the
> > release out.
> Thanks for the update. Bugzilla server is down ATM.
> But I've been following the issue 37019.
> It's still not 100% clear what the "correct" scenario is:
> should or should the NaNs have an impact on the result?
> Two comments you made seem slightly inconsistent.
> One:
> "The general policy in [math] is to use NaN consistently with floating
> point arithemtic specs (IEEE 754 for real, C99x Annex G for complex
> arithmetic) so for statistics such as Sum, Mean, etc. that involve
> arithmetic computations, NaN values force the return value to be NaN."
> Two:
> "This is definitely a bug, as the javadoc indicates clearly that NaN
> values should have no impact on the result:
> "The result is NaN iff all values are NaN (i.e. NaN values have no
> impact on the value of the statistic)."
> "
> Could you clarify please? Is the javadoc wrong or the impl?

What is incorrect is the impl, assuming the test results presented in
the bug report are correct (I have not turned these into Junit test
cases yet to verify).  The "iff" in the javadoc means "if and only if"
- so, e.g. the min of {NaN, 0} should be 0, not NaN.
> > We created a release branch once the contents of the 1.1 release were
> > fixed.  That branch is MATH_1_1.  The svn trunk version includes some
> > numerical analysis enhancements that will not be included in 1.1.
> > That's why the trunk version says 1.2
> That's fine, but I guess what I was also asking is whether you have
> a policy of releasing jars labeled 1.1-rc-* (I haven't been able to find
> them) or only the final release.

What you are looking for should be in the tars/zips here

Note that these are not "public" apache releases, but release
candidates for internal evaluation.

Thanks for the feedback!


> Cheers
> ---------------------------------------------------------------------
> To unsubscribe, e-mail:
> For additional commands, e-mail:

To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message