1 Apr 21:27 2010

### Bugfix release 1.3.1

Dear pgfplots users,

hereby I announce the release of the bugfix release 1.3.1 . It contains
the following changes:

- improved parametric plots with gnuplot
- improved normalsize, small and footnotesize scale styles and added tiny
- a lot of bugfixes

It is available from CTAN, sourceforge, and MikTeX within the next days.

Best regards

Christian

------------------------------------------------------------------------------
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev

2 Apr 21:34 2010

### Re: Plotting ellipse in the axis coordinate system

Hello Frank, hello Christian,

I recently had the same problem and Christians solution seems to work fine, as
long as you don't have to rotate the ellipsis. In this case you get some strange
results. Change the \Xmax/\Ymax or width/height values in my given "minimal" to
see what I am talking about.

I used a workaround where I just set a node in the axis and then draw the ellipsis
on that point after the axis ended.

My issue was to "fit" the data points into the ellipsis which didn't work with the
fit library, so I it with trial and error. Of course for that Christians solution
also works, but if somebody needs a correct rotated ellipsis his solution has to
be improved.

Best regards
Stefan Pinnow

\documentclass{article}
\usepackage{pgfplots}

% A helper command because I don't want to fool around with \pgfextractx
% and \pgfextray:
\makeatletter
% it defines "#1' to contain the x coord value and #2 to contain the y
% coord value
\newcommand{\extractpgfxy}[2]{%
\edef#1{\the\pgf <at> x}%
\edef#2{\the\pgf <at> y}%
}%


3 Apr 14:15 2010

### Re: Plotting ellipse in the axis coordinate system

Hello Stefan,

thank you for the example!

I checked your example and got
! Package tikz Error: Cannot parse this radius.'
when I use pgf 2.00. However, I works correctly with pgf CVS (see
attached pdf).

Maybe the 'ellipse []' syntax is newer than pgf 2.00?  If so, we could
try to find work-arounds which work for pgf 2.00 and your application
example... an idea could be to use

and

then, the '{}' should separate the macro from the 'and' (which is the
problem why \RADIUSX and' doesn't work). But the rotate=90 needs to be
applied after the moveto (axis cs:0,0).

Do you have ideas? Otherwise, I'd suggest to wait until the next stable
PGF version or to use the CVS version. It seems to work correctly.

Best regards

Christian

Stefan Pinnow schrieb:


3 Apr 16:25 2010

### Re: Plot through singularity

On Mon, Mar 29, 2010 at 4:03 AM, Christian Feuersaenger
<feuersaenger@...> wrote:
> Hello Liam,
>
> try using
>        restrict y to domain=-4:4,
> it will replace any coordinate outside of this range by either +inf or
> -inf and will interrupt the plots at these points. An even better result
> can be achieved by using a greater range like '-10:10', combined with
> samples=100 while you keep the ymin/ymax.
>
> Does this yield the desired output?
>

Kind of.  If I use the "restrict to domain", it avoids the connecting line
through the singularity, but there are also a couple gaps near
y=-4 and y=+4 points which are not plotted.  Increasing the resolution and
range as you suggest helps.  The workaround I used prior to your
response was to plot two lines, one from x=0 to the x value where y=-4,
then one from the x value where y=+4 to x=180.  Fortunately, these
x values are easy to calculate analytically in this case.  Unfortunately,
I am not a master of color cycling, so when I was trying to do a plot
with multiple curves it didn't come out right.  I switched to multiple
plots with a single curve (which is better for other reasons anyway)
and it looks fine.

It is good to know about the restrict option for the future.  Thank you
for your response (and of course, for pgfplots in its entirety, it is
a very nice package).



10 Apr 23:06 2010

### bug in pgfplots with gnuplot4.4?

Dear pgfplot-experts,
I use debian unstable and there's a update of gnuplot recently. Now the attached
example does not work anymore. Here're the error messages:

pgfplots 1.3.0 + gnuplot4.2.6 => works

pgfplots 1.3.0 + gnuplot4.4   => does not work
! Dimension too large.

\pgfplotsqpointxy #1#2->\pgf <at> x =#1\pgf <at> xx
\pgf <at> y =#2\pgf <at> yy
\pgfplots <at> coord <at> stream <at> finalize <at> currentpt ...t <at> y }

\pgfplots <at> coord <at> stream <at> coord ...ord <at> stream <at> coord <at>
\ifx \pgfplots <at> current <at> poi...

\pgfplots <at> foreach <at> plot <at> coord <at> NORMALIZED <at> NEXT ...d
\pgfplots <at> foreach <at> plot <at> coo...
<argument> ...977.00996000000000,1Y9.8875919e1],,;
1072.2672000000000,1Y9.888...
...
l.30 \stopaxis

pgfplots 1.3.1 + gnuplot4.2.6 => does not work:
Fatal error: catcode push/pop mismatch. Fix this!

\wait=



11 Apr 09:12 2010

### Re: bug in pgfplots with gnuplot4.4?

Hello Hubertus,

thank you for the bug report!

I investigated the problem and found that it wasn't actually caused by
the gnuplot update. Instead, it was caused by a recent change in PGF
CVS: a change introduced to improve the number printer in LaTeX captions
wasn't portable.

I have just committed a patch to PGF CVS; now your example compiles fine.

Sorry for the trouble and thank you for the report.

Best regards

Christian

Hubertus Bromberger schrieb:
> Dear pgfplot-experts,
> I use debian unstable and there's a update of gnuplot recently. Now the attached
> example does not work anymore. Here're the error messages:
>
> pgfplots 1.3.0 + gnuplot4.2.6 => works
>
> pgfplots 1.3.0 + gnuplot4.4   => does not work
>   ! Dimension too large.
>   <recently read> \pgf <at> xx
>
>   \pgfplotsqpointxy #1#2->\pgf <at> x =#1\pgf <at> xx
>                                             \pgf <at> y =#2\pgf <at> yy


12 Apr 20:07 2010

### Re: bug in pgfplots with gnuplot4.4?

Hello Hubertus,

thank you for providing the gnuplot output files.

I investigated and found that the .table file produced by gnuplot 4.4
does not respect the logscale.

I installed gnuplot 4.4 on my windows and verified that it, indeed,
doesn't apply the logs. gnuplot 4.2.* does (also tested on that windows
machine).

This behavior has nothing to do with pgfplots, and it can be reproduced
using

set table "test.saturableAbsTrans.table"; set format "%.5f"
set format "%.7e";; set samples 100; set logscale x 2.71828182845905;
plot [x=1:10000] 99.*log(1+97./99.*(exp(x/60.)-1))/(x/60.);

in gnuplot 4.4. The expected output contains log scale for x, as in
-------------------------------
#Curve 0 of 1, 100 points
#x y type
0.0000000e+000 9.7016243e+001  i
9.3033741e-002 9.7017818e+001  i
1.8606748e-001 9.7019544e+001  i
2.7910122e-001 9.7021436e+001  i
3.7213496e-001 9.7023510e+001  i
-------------------------------

but gnuplot 4.4 writes


13 Apr 23:50 2010

### Re: bug in pgfplots with gnuplot4.4?

Dear Christian,
> In case no one else has a further idea, you may need to ask the
> gnuplot team why the old .gnuplot scripts do not run anymore.
> Perhaps it is a bug and they are happy for the report?
I posted the problem on the gnuplot-bug list and it seems, it was bug. However,
in the old version...
there was a bug up to version 4.2.6. this has been repaired in version 4.4,
see the ChangeLog file at 2009-10-10.
So, what's now? Sorry I'm lost.

Thanks

Hubertus


15 Apr 22:08 2010

### [SPAM] Re: bug in pgfplots with gnuplot4.4?

Hello Hubertus,

just for you to note: I've written a patch for pgfplots which checks for
the version number of gnuplot explicitly by issueing a system call.

That works now (although I'll need to test it on windows someday).

In case you need gnuplot 4.4 (or you don't want to fiddle with
compilation of older versions), you can pull the git stable branch of
pgfplots which contains the fix. If that's too much trouble and you
still need it, you can also ask me kindly to pack a zip archive

Best regards

Christian

Hubertus Bromberger schrieb:
> Helle Christian,
>> funny enough, I suppose there is no way to configure gnuplot to
>> *sample* logarithmically without setting logscale. But what is a
>> logarithmic plot if samples are drawn linearly!?
> I don't get it neither... But I also have to admit, that after handing in my
> thesis last week, my brain is still verrrryyy slow ;))
>
>> Or do you know a method to configure logarithmic sampling in gnuplot
>> (like matlabs logspace function) without actually setting logscale?
> Sorry, no. I have no clue about gnuplot. I usually work with R and
> python/matplotlib for such things. I asked the mailing list again, but since I
> have no idea about gnuplot, maybe it would be better if you overtake the
> discussion from here on:


17 Apr 09:35 2010

### [Pgfplots features] extra bracing multiline tick labels

Hello all,

I noticed that one has to extra brace multiline tick labels in version 1.3.1
now, otherwise one will get the error

% -----
Runaway argument?
test,\tolist \pgfplots <at> xticklabels \relax \else \expandafter \pgfplotslistnew <at>
impl <at> comma \ETC.
! File ended while scanning use of \pgfplotslistnew <at> impl.
<inserted text>
\par
<*> C:/Temp/minimal/minimal.tex

?
% -----

In version 1.3.0 that wasn't necessary. Is that a "bug" or a feature?

Best regards
Stefan

% ----- minimal example -----
\documentclass{article}
\usepackage{pgfplots}
\begin{document}
\begin{tikzpicture}
\begin{axis}[
xmin=0,xmax=1,
x tick label style={