[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Plotting large values with small limits cause to make axes mad.

From: Philip Nienhuis
Subject: Re: Plotting large values with small limits cause to make axes mad.
Date: Wed, 13 Feb 2019 22:10:15 +0100
User-agent: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:51.0) Gecko/20100101 Firefox/51.0 SeaMonkey/2.48

Nicholas Jankowski wrote:

    That is due to OpenGL working with single precision floats. See
    a.o., bugs
    32980, 33748, and several later duplicates. Indeed an annoying issue.

these bugs have been around since 2011.  My understanding is that OpenGL
still has no intention to move to general double precision handling
since most graphics hardware still handles everything in single
precision (no double benefit for the number and size of pixels our eyes
are dealing with so far).

is there a particular hesitancy in working up a scaling fix for this
problem on the octave side rather than waiting for OpenGL to go double?

AFAIU the Octave devs are waiting for an OpenGL expert to fix this.

FYI Matlab has the same challenge (it can also use OpenGL) but seems to manage it well.

Note that AFAIU fixing this isn't an easy undertaking. It may appear easy from the outside, just some scaling s you say, but deep inside Octave's C++ OpenGL code it probably looks a bit more involved.

I've looked at adapting plot.m by subtracting the minimum axis value from the data before plotting and adapting the tick labels afterwards. As I often want to plot time series of measured field data the X-axis tick labels have to be adapted to a date/time format, again when zooming in or out. Soon it turned out I would also need to patch e.g., datetick.m. Altogether it became too challenging an affair for me. At work I have Matlab anyway so I gave up and didn't bother much. And then there is patch, line, polar, contour, ..., etc etc. all of which will have to be patched as well.

IIRC Rik has uploaded a workaround patch for (again: IIRC) plotyy.m in one of those aforementioned bug reports.


reply via email to

[Prev in Thread] Current Thread [Next in Thread]