Re: manual ticks of an axis

Le mardi 25 novembre 2014 à 07:36 -0700, John Denker a écrit :
On 11/25/2014 05:15 AM, Jean Brefort wrote:

 There is already an enhancement request:

I'm unsure about random ticks. I'd like a significant use case for them
before trying to implement them.

Could we please call them /manual/ ticks or /arbitrary/
ticks rather than "random" ticks?

No problem, but I intentionally used "random" because the bug reporter
seems to be able to add ticks not following any rule, which seems weird
at first sight, at least.

On the other side, things as T^n (n real such as -1, 1/2 or whatever is
meaningful, and I plan to implement that), actually we might have
different ticks on different lines of the same axis like T and 1/T, or
the same with different units (cm and in or so)

The rule is that we should be able to build any chart in gnumeric (this
just needs a lot of time), at least if it seems meaningful.

If we are to go on with this discussion, we should use the bug report
since mails are prone to vanish.

Sorry for my approximate english which might not perfectly reflect what
I actually mean.


Also note that it is the ticks and labels that are important;
the "axis" itself becomes a nuisance and a source of confusion
as soon as you try to do anything nontrivial, such as a
spacetime diagram or ternary phase diagram;  for details see


The discussion of use cases starts with the observation that a
gnumeric XY plot offers two "types" of axes: linear and logarithmic.
These must be considered two special cases in a vast universe of
possibilities.  It would be nice to handle the general case.

Case 1:  Arrhenius law:  rate proportional to exp(1/T).
  To represent this as a straight line, it is necessary
  to plot log(rate) versus 1/T.
  It would be nice to have ticks labeled in terms of T, not 1/T:

             position        label
               1/4            T=4
               1/2            T=2
               2/3            T=1.5
                1             T=1

Case 2:  Ohm's law.  Current as a function of inverse resistance
  at constant voltage.  Similar to case 1, without the exponential.

Case 3:  The resistance of a carbon-comp resistor at very
  low temperatures is exponential in the square root of 1/T.
  Such things are used as thermometers.  For plotting the
  calibration curve, it's really nice to have a straight
  line plot, plotted as a function of 1/√T, with ticks
  labeled in units of T.

Case 4:  If you want to get serious, consider tilted ticks
  and iso-contours as in any ternary plot:

Case 5:  Tilted ticks and iso-contours due to rotation in the
   XY plane, e.g. magnetic north versus geographical north.

Case 6:  Tilted ticks and iso-contours due to rotation in the 
   XT plane in spacetime, using /hyperbolic/ trigonometry.

Case 7:  If you want to get really serious, consider the
   task of labeling all *four* sets of iso-contours in
   a chart such as this psychrometric chart:

Case 8:  Similarly, the *four* sets of iso-contours in
   a thermodynamic "indicator diagram" such as


I could go on, but I reckon you get the idea.  There are
a lot of serious real-world applications for tick-marks
and iso-contours that are not simply linear in X or 
logarithmic in X.

gnumeric-list mailing list
gnumeric-list gnome org

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