Mailing-List: contact dev-help@commons.apache.org; run by ezmlm
Precedence: bulk
Reply-To: "Commons Developers List" <dev@commons.apache.org>
Received-SPF: pass (nike.apache.org: domain of yangshuan45@gmail.com
 designates 209.85.215.49 as permitted sender)
MIME-Version: 1.0
In-Reply-To: <526043D0.6040806@gmail.com>
References: 
 <CA+LiXWYj+GNz5rYR7wgbvY8ffE3C4CrB+5zYHfciSV+uqoozpw@mail.gmail.com>
	<526043D0.6040806@gmail.com>
Date: Fri, 18 Oct 2013 09:33:42 +0800
Message-ID: 
 <CA+LiXWa-oNXdC91+QS6d+6cnsSSqv3oTpZaX8D-y=7++ohR7Vg@mail.gmail.com>
Subject: Re: [math] Method min, max, percentile in StatUtils support int[],
 long[], etc
From: Karl Yang <yangshuan45@gmail.com>
To: Commons Developers List <dev@commons.apache.org>
Content-Type: multipart/alternative; boundary=089e01227ca4b5c65404e8f9ecaa

--089e01227ca4b5c65404e8f9ecaa
Content-Type: text/plain; charset=GB2312
Content-Transfer-Encoding: quoted-printable

Define statistics such as percentiles, mean etc that expect integer arrays
as arguments is what I wish, but the return type is decided by the method,
for method min(int[]) and max(int[]), the return type must be integer; for
method mean, the return type may be double.
If not I need to convert int[] to double[] every time, this would cost a
lot memory. Due to the data is very large, JVM would call GC to free the
memory many times in a thread, it's a big cost. The only way I can fix this
performance problem is to implement my method based on
commons-math's algorithm that can accept integer arrays.


2013/10/18 Phil Steitz <phil.steitz@gmail.com>

> On 10/16/13 6:36 PM, =D1=EE=CB=A8 wrote:
> > Hi,
> >
> > I always need to compute min value, max value, percentile values of
> > millions data, these data may are int, long or double. But class
> StatUtils only
> > supports double[], I need to convert int[], long[] to double[]. Shall w=
e
> > add a feature that StatUtils supports any other numeric type?
> >
> It would be trivial to just add overloaded versions of the methods
> that do the conversion for you. Is that what you mean? They would
> still return doubles as results. Is that what you mean? Or do you
> mean defining statistics such as percentiles, mean, etc. that expect
> integer arrays as arguments and return results coerced to integer
> values? If the latter, what exactly is the use case - i.e., what is
> the practical need for this?
>
> Phil
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
> For additional commands, e-mail: dev-help@commons.apache.org
>
>

--089e01227ca4b5c65404e8f9ecaa--