#2837 In 6.0.3, some of the problems with log scale axes remain

[Ilya Zakharevich]

First, a lot of thanks for processing my bug report [#2717].¹⁾

¹⁾ Especially compared a few of my other bug reports about ticks, (wrongly) marked as “not bugs” — alas! — (And many more which I sit on due to such mishandling.)

However, a few problems still remain. For example,

reset
set logscale xy
plot [7:530] x

exposes two:
* The part between 7 and 100 has neither ticks, nor labels.
* The x- and y-axes are treated differently (in practically all the senses).

Finally:

plot exp(x**3)

Shows the the code to decimate minor marks is not vigilant enough.

Thanks again, [tested on Windows]
Ilya

[Ethan Merritt]

• The part between 7 and 100 has neither ticks, nor labels.

True. But this is a distinct improvement over what you got prior to gnuplot 6.0.2 (only two labeled marks, at 10 and 100). IMHO the current result is acceptable, but I agree it would be nice to allow the auxiliary command likeset xtics add 10,10,100 to fill in a specific region of the axis.

Right now the add option only works for an explcit list of labels, not for a range+interval specification. And even an explicit add list currently may not work correctly for logscale. I will look into it.

• The x- and y-axes are treated differently (in practically all the senses).

Not really. You have set an explicit range on x, so that exactly defines the end points. Y defaults to autoscaling, and in autoscaling the range is extended to the next tick mark at each end. You can turn this off by either set ytics noextend or set autoscale y noextend, after which it will match the x axis range for this plot.

• plot exp(x**3)
  Shows the the code to decimate minor marks is not vigilant enough.

Yeah. That one is clearly a bug.

[Ilya Zakharevich]

[Yes, it is absolutely an improvement — but I already commented on this ☺😊 — and thanks anyway!]

However: Why should one be forced to use

the auxiliary command like set xtics add 10,10,100 to fill in a specific region of the axis.

Let me recall one of the “axioms” of [#2717]: the ticks ”should” be placed

were users expect them.

The gnuplot⸣s agent for placing-the-ticks/marks-intelligently already found the best strategy to visualize (essentially) the range 100..1000 at the particular scale. — So all that remains is: just repeat these ticks/marks in the (remaining present-on-the-plot) ranges 10..100 and 1..10. — All that is needed is the to translate the already known results for 100..1000 parallelly!

[Ethan Merritt]

• plot exp(x**3)
  Shows the the code to decimate minor marks is not vigilant enough.
  Yeah. That one is clearly a bug.

Fixed now.