Sun 18 Sep 2011 04:48:14 PM UTC, comment #10:
There is already a report of missing TeX/LaTeX support below
https://savannah.gnu.org/bugs/index.php?32822
I'd like to see the number of bug reports pruned down to something more manageable. Since this is essentially a duplicate of 32822, I'm closing it.
|
Sat 17 Sep 2011 12:51:39 AM UTC, comment #9:
The basic problem is that the OpenGL backend support does not support TeX/LaTeX markup.
I recommend closing this and starting a new report.
|
Tue 06 Sep 2011 06:35:44 PM UTC, comment #8:
I've pushed a slightly modified changeset.
http://hg.savannah.gnu.org/hgweb/octave/rev/c7512d0d52e8
|
Tue 06 Sep 2011 06:10:53 PM UTC, comment #7:
Actually, where I saw the code was in scripts/plot/
colorbar.m at line 557, so this is pretty irrelevant.
I think it is pretty obvious where to code goes
in graphics.cc, but it will need a bit of C++
cleverness...
Michael
|
Tue 06 Sep 2011 06:02:28 PM UTC, comment #6:
Ben,
This definitely helps. But, it does not
do the form 10^xx for the tick labels.
Particularly not for xx < 1.
I thought I saw code (which appeared not to be used)
in graphics.cc which setup the exponent form. I wll
look again. In any case, code of that form is needed.
This also means that the minimum range can be less than
10.
Good start at least, and this patch could, I think,
be pushed as is.
Michael
|
Tue 06 Sep 2011 03:29:27 PM UTC, comment #5:
This fixes most of the problem for me.
However, for plots like loglog([1 1.1]) the result would look better if the axis limits were tighter (meaning I think the tick_sep is correct, but the axis limits are too loose).
|
Tue 06 Sep 2011 01:12:46 AM UTC, comment #4:
Here looks like a place to start:
Line: 5236 in graphics.cc:
void
axes::properties::calc_ticks_and_lims (array_property& lims,
array_property& ticks,
array_property& mticks,
bool limmode_is_auto, bool is_logscale)
{
// FIXME -- add log ticks and lims
|
Tue 06 Sep 2011 01:07:49 AM UTC, comment #3:
The code in graphics.cc appears to be intended
to generate correct log tick labels, but it seems
not to be working.
I will make a try at understanding where it goes
wrong, but attention from someone who understands
the code would likely get this fixed sooner.
Michael
|
Tue 06 Sep 2011 12:58:13 AM UTC, comment #2:
Ben,
My quick reading of the gnuplot path code suggested that
if gnuplot is told that the axis is log-scale, it does the
"right" thing with tick-labels. The code also tells gnuplot
to reduce the
number of minor ticks for log axis and this seems to work.
Michael
|
Tue 06 Sep 2011 12:45:23 AM UTC, comment #1:
The tick mark/label positions aren't calculated for gnuplot either. I haven't checked, but it appears that either gnuplot or the gnuplot backend fixes this.
In graphics.cc the is a FIXME which mentions log ticks.
|
Mon 05 Sep 2011 03:06:18 AM UTC, original submission:
When using log axes with fltk the tick labels are
displayed as for linear axes. Thus, for semilogy(1:10)
the x axis tick labels are:
1.58489 2.51189 3.98107 6.30957 10
instead of
10^0.2 10^0.4 10^0.6 10^0.8 10^1.0
or, for ML compatibility:
10^0 10^1
Also, ML shows tick marks as follows for semilogx(1:2):
10^0 10^0.1 10^0.2 10^0.3
and, in the ML plot the ^xx is actually a superscript.
Not sure if OpenGL knows how to do this. Not too important
compared to using 10^xx for the log tick labels and reducing
the number of minor ticks.
I looked at the code in gl-render.cc, which seems to be
where this fix should go, but it was not obvious what the
correct code should be.
So, can someone with OpenGL knowledge look at this?
Michael
|