On Thu, 11 Apr 2013 16:45:42 +0200, Arne Ploese wrote:
Am Donnerstag, den 11.04.2013, 16:08 +0200 schrieb Gilles:
Hello.
>
> in class org.apache.commons.math3.linear.RealVector there is the
> embedded class Entry which is protected, but it is exposed by
method
> public Iterator<Entry> iterator() ...
Good catch. Thanks.
>
> My solution:
> 1. make class Entry public
> 2. make class RealVector implement Iterable<RealVector<Entry>>
>
> so one can do this:
>
> for (RealVector.Entry e : v) {
> if (Double.isNaN(e.getValue()) ||
> Double.isInfinite(e.getValue())) {
> throw new RuntimeException("roots: inputs must not contain Inf
or
> NaN");
> }
> }
IMO, exposing an "Entry" is a mistake: From a user point-of-view, a
"RealVector" should be a collection of "[Dd]ouble".
IIUC, the "Entry" was meant to allow storage or iteration
optimizations
in subclasses.
I think that the current "iterator()" method must be deprecated and
removed; in its place, there should be a
protected Iterator<Entry> entryIterator()
and a new
public Iterator<Double> iterator()
(that would use the above method), such that "RealVector" would
implement "Iterable<Double>", as is appropriate for this concept.
Eventually, your use-case would become:
-----
for (Double e : v) {
if (e.isNaN() ||
e.isInfinite()) {
throw new RuntimeException("inputs must not contain Inf or
NaN");
}
}
-----
For me it would be handy to have a RealVector isNaNOrInfinite to
avoid
one iteration - I mean handy not necessary.
There is a JIRA issue about refactoring both matrix and vector APIs.
We are probably not going to add new methods until the whole issue is
tackled.
Gilles
> JIRA + diff???
A bug report mentioning the inconsistency is welcome.
A diff is not necessary until we agree on the way to go.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@commons.apache.org
For additional commands, e-mail: dev-h...@commons.apache.org
