matplotlib-users

[Matplotlib-users] tick label hiding problem

[Michael F. <mp...@be...>]

Hi all,

Sometime in the past couple months, there's been a change in SVN that's 
causing me problems when trying to hide tick labels.  I think the problem is 
related to the get_xticklabels()/get_yticklabels() methods of the Axes class, 
in that they only return a list consisting of a single tick label no matter 
how many there should be.  I'm encountering this problem in SVN rev. 2706.

Here is some example code:

from pylab import *
# set up a plot
x = arange(5)
ax = subplot(111)
ax.plot(x, x)
# hide the first tick label on the x axis
xtl = ax.get_xticklabels()
xtl[0].set_visible(False)
print len(xtl)
ax.set_xlabel('foo')
show()

For me, no x tick labels are drawn, and the length of xtl is one instead of 
the expected 9.  

Can anyone reproduce this?

Mike

Re: [Matplotlib-users] tick label hiding problem

[Michael F. <mp...@be...>]

Hi all,

I'm still having trouble figuring out what's going on with tick  
labels.  I can't figure out why Axes.get_{x,y}ticklabels() is  
returning a list of length unity when I expect to have the list of  
all labels.

If I run the code below (test.py), I expect a list of 9 x tick  
labels.  If I run the code once ('run test' in ipython), then it  
gives me a list of 1 tick label.  If I close the interactive window  
it generates and run it again, it again returns 1 tick label.   
However, if I leave the window open and re-run the test, it gives the  
right answer.

Can anyone else reproduce this?  Is there something I'm missing?  I'm  
using svn rev 2730 of matplotlib, and I think this is new behavior in  
the last couple months.

Re: [Matplotlib-users] tick label hiding problem

[John H. <jdh...@ac...>]

>>>>> "Michael" == Michael Fitzgerald <mp...@be...> writes:

    Michael> Hi all,

    Michael> I'm still having trouble figuring out what's going on
    Michael> with tick labels.  I can't figure out why
    Michael> Axes.get_{x,y}ticklabels() is returning a list of length
    Michael> unity when I expect to have the list of all labels.

    Michael> If I run the code below (test.py), I expect a list of 9 x
    Michael> tick labels.  If I run the code once ('run test' in
    Michael> ipython), then it gives me a list of 1 tick label.  If I
    Michael> close the interactive window it generates and run it
    Michael> again, it again returns 1 tick label.  However, if I
    Michael> leave the window open and re-run the test, it gives the
    Michael> right answer.

    Michael> Can anyone else reproduce this?  Is there something I'm
    Michael> missing?  I'm using svn rev 2730 of matplotlib, and I
    Michael> think this is new behavior in the last couple months.

It's a feature, not a bug :-)

Here is what is happening.  matplotlib allocates a different number of
ticks depending on the zoom level, and the axis seeds itself with a
single tick called the "protoTick".  If you zoom to the right and new
ticks need to be created, the properties from the protoTick are copied
to the newly created ticks. 99% of the time, this is what you want.
If you have large bold tick labels, and zoom to the right, you want
the new ticks that appear to be large and bold.  So this explains why
you see a single tick before the window is drawn (the protoTick), a
list of ticks after the window is drawn, and why the properties of the
first tick (the "visible" property in this case) are transferred to
the other ticks.

So: how do you solve your problem, of making the first tick invisible?
What I do when I need to solve this problem, which comes up a lot with
multiple axes where ticks can overlap, is the following

from matplotlib.ticker import ScalarFormatter

class MyFormatter(ScalarFormatter):
    def __call__(self, x, pos=None):
        if pos==0: return ''
        else: return ScalarFormatter(self, x, pos)

ax.xaxis.set_major_formatter(MyFormatter())

I often want to do this for the last tick as well, but there is
currently no convenient way to do this, as the tick locator produces
the tick location list and the tick formatter doesn't know how many of
them there are.  I've been considering modifying the tick locator code
to pass -1 for the pos for the last tick to solve just this use case.
If others need this or think it desirable, just give me a nudge and
I'll do it.

Details about the tick locators and formatter can be found at
http://matplotlib.sourceforge.net/matplotlib.ticker.html and there are
a number of examples

  ~/mpl/examples> grep -l Formatter *.py
  custom_ticker1.py
  dashtick.py
  date_demo1.py
  date_demo2.py
  date_demo_rrule.py
  finance_demo.py
  major_minor_demo1.py
  major_minor_demo2.py
  newscalarformatter_demo.py
  shared_axis_demo.py

Hope this helps,
JDH

Re: [Matplotlib-users] tick label hiding problem

[Michael F. <mp...@be...>]

On Aug 31, 2006, at 4:36 AM, John Hunter wrote:
> So: how do you solve your problem, of making the first tick invisible?
> What I do when I need to solve this problem, which comes up a lot with
> multiple axes where ticks can overlap, is the following
>
> from matplotlib.ticker import ScalarFormatter
>
> class MyFormatter(ScalarFormatter):
>     def __call__(self, x, pos=None):
>         if pos==0: return ''
>         else: return ScalarFormatter(self, x, pos)
>
> ax.xaxis.set_major_formatter(MyFormatter())
>
> I often want to do this for the last tick as well, but there is
> currently no convenient way to do this, as the tick locator produces
> the tick location list and the tick formatter doesn't know how many of
> them there are.  I've been considering modifying the tick locator code
> to pass -1 for the pos for the last tick to solve just this use case.
> If others need this or think it desirable, just give me a nudge and
> I'll do it.

Hi John,

Hmm.... I have 3 use cases where I want to isolate individual tick  
labels and make them invisible:

1) the first tick (0; for stacked plots)
2) the last tick (-1; for stacked plots)
3) every other tick for the last few ticks (e.g. -2, -4, -6; for a  
log-scale axis with large tick labels, which can overlap on one end)

Is there a way to force some sort of pseudo-draw event, such that the  
sizing is done and the ticks are created from the protoTick, but the  
draw isn't actually performed by the backend?  That way one could set  
the visibility property rather than wrap the Formatter.

Thanks,
Mike

Re: [Matplotlib-users] tick label hiding problem

[John H. <jdh...@ac...>]

>>>>> "Michael" == Michael Fitzgerald <mp...@be...> writes:

    Michael> Is there a way to force some sort of pseudo-draw event,
    Michael> such that the sizing is done and the ticks are created
    Michael> from the protoTick, but the draw isn't actually performed
    Michael> by the backend?  That way one could set the visibility
    Michael> property rather than wrap the Formatter.

Try fig.canvas.draw() and see if that helps.

If not, we'll come up with something else...

JDH

Re: [Matplotlib-users] tick label hiding problem

[Michael F. <mp...@be...>]

On Aug 31, 2006, at 2:56 PM, John Hunter wrote:

>>>>>> "Michael" == Michael Fitzgerald <mp...@be...> writes:
>
>     Michael> Is there a way to force some sort of pseudo-draw event,
>     Michael> such that the sizing is done and the ticks are created
>     Michael> from the protoTick, but the draw isn't actually performed
>     Michael> by the backend?  That way one could set the visibility
>     Michael> property rather than wrap the Formatter.
>
> Try fig.canvas.draw() and see if that helps.
>
> If not, we'll come up with something else...

I placed that command before the ax.get_xticklabels(), and no dice.

Thanks for looking into this,
Mike

Re: [Matplotlib-users] tick label hiding problem

[John H. <jdh...@ac...>]

>>>>> "Michael" == Michael Fitzgerald <mp...@be...> writes:


    Michael> I placed that command before the ax.get_xticklabels(),
    Michael> and no dice.

    Michael> Thanks for looking into this, Mike

I took a look at the formatter code and it turns out it *does* know
the list of locations it has to format.  I was wrong about this in my
previous post.  The formatter has a locs attribute you can inspect to
see how many ticks there are.  So you should be able to do something
like

class MyFormatter(ScalarFormatter):
    def __call__(self, x, pos=None):
        N = len(self.locs)
        if pos==0: return ''        # turn off first
        elif pos==(N-1): return ''  # turn off last
        else: return ScalarFormatter(self, x, pos)

and you can do other things similarly, eg to turn off every other tick

if (pos%2)==0: return ''

If you want to get very clever with turning on and off certain ticks,
you can also create a custom locator derived from the Locator class
you are using.

JDH