Path: csiph.com!eternal-september.org!feeder.eternal-september.org!news.eternal-september.org!.POSTED!not-for-mail
From: Ethan A Merritt <sfeam@users.sourceforge.net>
Newsgroups: comp.graphics.apps.gnuplot
Subject: Re: Ignore missing data in the desired way
Date: Fri, 20 Jan 2017 10:19:57 -0800
Organization: gnuplot development
Lines: 50
Message-ID: <o5tk9d$9f4$1@dont-email.me>
References: <4eb939ed-8e0f-45aa-a163-d1714b70536a@googlegroups.com> <o5sqtt$ese$1@solani.org>
Mime-Version: 1.0
Content-Type: text/plain; charset="ISO-8859-1"
Content-Transfer-Encoding: 7Bit
Injection-Date: Fri, 20 Jan 2017 18:18:21 -0000 (UTC)
Injection-Info: mx02.eternal-september.org; posting-host="0c1b7ba6bd8c11bd5d96215cf2c30ee2"; logging-data="9700"; mail-complaints-to="abuse@eternal-september.org";	posting-account="U2FsdGVkX1/Z2S+yaBFh4j8nMOYO08wt"
User-Agent: KNode/4.14.5
Cancel-Lock: sha1:URkznHUI6VYdvlm6Ut5Rl8+K+Z8=
Xref: csiph.com comp.graphics.apps.gnuplot:3546

Karl Ratzsch wrote:

> Am 20.01.2017 um 10:35 schrieb u.gotzes@gmail.com:
>> 
>> I can't get the script below to do what I want, namely make gnuplot draw
>> continuous lines between the respective datasets "Yes", "No" and "Maybe".
>> Can anybody help?
>> 
> 
> (simplified example)
> set dataf missing '?'
> $dat << EOD
> No Maybe
> 0 2
> 1 ?
> 2 1
> 3 2
> 4 ?
> 5 9
> 6 ?
> ? 1
> 8 2
> 9 5
> EOD
> 
> plot "$dat" using "No":"Maybe" w lp # missing dp are invalid
> pause -1
> plot "$dat" using 1:2 w lp    # works
> 
> 
> 
> There are few connecting lines and no bezier curve, because an
> invalid datapoint breaks those connections.
> 
> Aaand you know about that and "set datafile missing '?'", meaning
> the connections should not be broken.
> 
> I find that it works if I use column numbers instead of header
> strings in the "using" specifier. Looks like a bug to me.

Version 5 was supposed to remove the difference between 
   using N    and   using ($N)
but it seems that this is a case where they are still treated
differently, as they were in earlier gnuplot versions.
So on the one hand it's always been this way, but on the other hand
it was intended to change. Let's call it a bug.

Note that using ($N) and using (column(N)) and using (column("head"))
all share the same input processing.  So this bug is not specific
to column headers.