[
https://issues.apache.org/jira/browse/MATH-1205?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14330465#comment-14330465
]
Thomas Neidhart commented on MATH-1205:
---------------------------------------
I am working on a patch, but there are at least two other odd things that I
have seen wrt storeless statistics:
a) default implementation of equals / hashCode that compares solely on
getResult() and getN(): I fail to see a use-case for this and it means that two
different types of storeless statistics are considered to be equal if their
current state is equal:
{code}
Mean m = new Mean();
Variance v = new Variance();
System.out.println(m.equals(v)); // outputs true
{code}
I find this very strange, especially as a default.
b) evaluate(double[], ...): the default implementation in
AbstractStorelessUnivariateStatistic does the following:
* clear the internal result
* call increment for all input values in the array
* return the result
thus the method alters the internal state, which is different for non-storeless
statistics.
Some storeless statistics override this method, e.g. Mean to do something else:
calculate the result using an intermediate object and then return the result
without altering state. This is quite confusing imho.
> AbstractStorelessUnivariateStatistic should not extend
> AbstractUnivariateStatistic
> ----------------------------------------------------------------------------------
>
> Key: MATH-1205
> URL: https://issues.apache.org/jira/browse/MATH-1205
> Project: Commons Math
> Issue Type: Bug
> Affects Versions: 3.4.1
> Reporter: Thomas Neidhart
>
> For a storeless statistic it is wrong to extend AbstractUnivariateStatistic
> as various fields and methods are inherited that do not make sense in case of
> a storeless statistic.
> This means a user can accidentially use a storeless statistic in a wrong way:
> {code}
> Mean mean = new Mean();
>
> mean.increment(1);
> mean.increment(2);
>
> mean.setData(new double[] { 1, 2, 3});
>
> System.out.println(mean.getResult());
> System.out.println(mean.evaluate());
> {code}
> will output
> {noformat}
> 1.5
> 2.0
> {noformat}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
