References: <58C1DA5F.8010109@clear.net.nz>
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:45.0) Gecko/20100101
            Thunderbird/45.7.0
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Message-ID:  <1b5ec418-2f2d-d39d-af53-a8921d3e8176@gmail.com>
Date:         Fri, 10 Mar 2017 19:49:58 -0500
Reply-To: Mailing list for the LaTeX3 project
              <LATEX-L@LISTSERV.UNI-HEIDELBERG.DE>
Sender: Mailing list for the LaTeX3 project <LATEX-L@LISTSERV.UNI-HEIDELBERG.DE>
From: Bruno Le Floch <blflatex@GMAIL.COM>
Subject: Re: Two more (simple) functions for l3fp
To: LATEX-L@LISTSERV.UNI-HEIDELBERG.DE
In-Reply-To:  <58C1DA5F.8010109@clear.net.nz>
Precedence: list
Envelope-To: <rainer.schoepf@GMX.NET>
Content-Transfer-Encoding: 8bit
Status: R

On 03/09/2017 05:42 PM, Andrew Parsloe wrote:
> Two functions which I think could be beneficially added to l3fp without
> straining its role as a kernel module (in the way that, say, the gamma
> function would) are the signature function (sgn) and a unit step
> function (Heaviside? -- the difficulty is what value to assign at 0).
> Given the numerical values assigned to true and false in l3fp, it is
> easy to mimic these with simple inequalities (e.g. (#1 > 0) - (#1 < 0)
> for sgn) but perhaps the functions should be there in the first place?
> 
> I found myself wanting these when considering a table of fp values where
> it made sense to blank cells containing negative values which "got in
> the way". The code creating the table contained the option of blanking
> zero-value cells, so I multiplied the tabulated formula by 1 - (#1 < 0).
> 
> Andrew
> 
> ---
> This email has been checked for viruses by Avast antivirus software.
> https://www.avast.com/antivirus

Will copied your message to https://github.com/latex3/latex3/issues/346
and replied there that

> sign() or sgn() definitely makes sense to me, but I’m not sure about
> step() / heaviside() / whatever. Can you give another example for the
> use case there?

I'm adding sign().  I don't think heaviside is needed: (#1) > 0 ? (#1) :
0 seems to do what you want pretty nicely.

What do you think?

Concerning Gamma, the main reason I haven't implemented it yet is that
it requires determining some magic constants that were only given to me
for a lower precision than what I need (10 digits instead of 16).

Bruno