Thread: [Matplotlib-users] autoscale and "equal aspect" mode

I'm using this aspect mode:

       axes.set_aspect( 'equal', adjustable='datalim' )

With several xy line plots on the axes, autoscaling doesn't seem to 
occur properly. It cuts off part of the data. Is this a known issue? I 
don't have a simple script to replicate it---I'll have to pull that out 
of a larger program---but just wanted to check if it is known. I don't 
see anything in the bug tracker.

Mike

>>>>> "Michael" == Michael P Mossey <mo...@jp...> writes:

    Michael> I'm using this aspect mode: axes.set_aspect( 'equal',
    Michael> adjustable='datalim' )

    Michael> With several xy line plots on the axes, autoscaling
    Michael> doesn't seem to occur properly. It cuts off part of the
    Michael> data. Is this a known issue? I don't have a simple script
    Michael> to replicate it---I'll have to pull that out of a larger
    Michael> program---but just wanted to check if it is known. I
    Michael> don't see anything in the bug tracker.

matplotlib version number?  a lot of work as been done quite recently
on aspect handling.  You should test against the latest svn and then
post a complete example describing what is wrong if you are
unsatisfied.

JDH

Mike,

I thought all the aspect handling was finally working correctly.  If you 
generate a simple example, I will take a look.  The simpler the better, 
of course.  I presume you are working with a recent svn version.

Eric

Michael P. Mossey wrote:
> I'm using this aspect mode:
> 
>       axes.set_aspect( 'equal', adjustable='datalim' )
> 
> With several xy line plots on the axes, autoscaling doesn't seem to 
> occur properly. It cuts off part of the data. Is this a known issue? I 
> don't have a simple script to replicate it---I'll have to pull that out 
> of a larger program---but just wanted to check if it is known. I don't 
> see anything in the bug tracker.
> 
> Mike
> 
> 
> 
> -------------------------------------------------------
> Using Tomcat but need to do more? Need to support web services, security?
> Get stuff done quickly with pre-integrated technology to make your job 
> easier
> Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
> http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
> _______________________________________________
> Matplotlib-users mailing list
> Mat...@li...
> https://lists.sourceforge.net/lists/listinfo/matplotlib-users

Andrew,

Sure enough.  I did not pay a lot of attention to the zoom part of 
pan/zoom mode with fixed aspect ratio because it doesn't make much sense 
to me--it is trying to change the aspect ratio, while the aspect-ratio 
code is desperately trying to keep the aspect ratio fixed. The solution 
is probably to redefine what the right-button-event does in the fixed 
aspect ratio case. It should either be disabled, or arranged so that 
motion to the right and/or up zooms in and to the left and/or down zooms 
out.  What do you think it should do?

Left-button pan is definitely broken for horizontal motion.

I think what we have is not a single simple bug, but a whole set of 
bugs.  It seems to be difficult to fix one thing related to aspect-ratio 
handling without breaking something else for interactive drawing.  That 
probably means the overall design is bad.  When I can, I will take 
another look and see if I can patch it up, but it won't be right away. 
If you can find a solution, that would be great.

Eric



Andrew Straw wrote:
> I'm having trouble with axis('equal') myself -- I'm happy to take a look
> in the source, but maybe it's a simple and easy bug. The issue I'm
> having is readily apparent with examples/axis_equal_demo.py -- using the
> pan/zoom mode and holding the right-mouse button down to zoom, I notice
> several issues:
> 
> 1) moving left-and-right seems to affect the position, not the zoom level
> 2) moving up seems to zoom in (as expected)
> 3) moving down zooms out as expected, but past a certain point, only the
> vertical axis gets re-scaled and breaking the equal-aspect.
> 
> Eric Firing wrote:
> 
> 
>>Mike,
>>
>>I thought all the aspect handling was finally working correctly.  If
>>you generate a simple example, I will take a look.  The simpler the
>>better, of course.  I presume you are working with a recent svn version.
>>
>>Eric
>>
>>Michael P. Mossey wrote:
>>
>>
>>>I'm using this aspect mode:
>>>
>>>      axes.set_aspect( 'equal', adjustable='datalim' )
>>>
>>>With several xy line plots on the axes, autoscaling doesn't seem to
>>>occur properly. It cuts off part of the data. Is this a known issue?
>>>I don't have a simple script to replicate it---I'll have to pull that
>>>out of a larger program---but just wanted to check if it is known. I
>>>don't see anything in the bug tracker.
>>>
>>>Mike
>>>
> 
> 

Andrew,

I have committed another revision of the aspect-ratio handling.  It seems to work pretty well, and I think pan/zoom work in a reasonable fashion now.  Many times in the past, however, I have thought that aspect-ratio handling was working adequately, and every time someone has found a major bug.  This time is probably no exception.

There can be some anomalies when the window is resized quickly--that is, grabbing a window corner and slinging it around with axis('equal') can cause some incorrect choices of the view limits.  I haven't figured out why.  Although I certainly would like to understand and fix it, it is not a top priority at the moment because I don't think it is a major problem in practice. 

Eric

----- Original Message -----
From: Andrew Straw <str...@as...>
Date: Saturday, May 27, 2006 12:07 pm
Subject: Re: [Matplotlib-users] autoscale and "equal aspect" mode
To: Eric Firing <ef...@ha...>, mat...@li...

> I'm having trouble with axis('equal') myself -- I'm happy to take a 
> lookin the source, but maybe it's a simple and easy bug. The issue I'm
> having is readily apparent with examples/axis_equal_demo.py -- 
> using the
> pan/zoom mode and holding the right-mouse button down to zoom, I 
> noticeseveral issues:
> 
> 1) moving left-and-right seems to affect the position, not the zoom 
> level2) moving up seems to zoom in (as expected)
> 3) moving down zooms out as expected, but past a certain point, 
> only the
> vertical axis gets re-scaled and breaking the equal-aspect.
> 
> Eric Firing wrote:
> 
> > Mike,
> >
> > I thought all the aspect handling was finally working correctly.  If
> > you generate a simple example, I will take a look.  The simpler the
> > better, of course.  I presume you are working with a recent svn 
> version.>
> > Eric
> >
> > Michael P. Mossey wrote:
> >
> >> I'm using this aspect mode:
> >>
> >>       axes.set_aspect( 'equal', adjustable='datalim' )
> >>
> >> With several xy line plots on the axes, autoscaling doesn't seem to
> >> occur properly. It cuts off part of the data. Is this a known 
> issue?>> I don't have a simple script to replicate it---I'll have 
> to pull that
> >> out of a larger program---but just wanted to check if it is 
> known. I
> >> don't see anything in the bug tracker.
> >>
> >> Mike
> >>
> 
> 

Eric,

This works great and fixes all the issues I saw. Thanks much. I didn't
see the anomaly you mentioned.

PS My ISP (dreamhost.com)'s email server is SpamCop's blackhole, which
means SourceForge will reject it from the mailing list unless they've
resolved the issue already.

Eric Firing wrote:

>Andrew,
>
>I have committed another revision of the aspect-ratio handling.  It seems to work pretty well, and I think pan/zoom work in a reasonable fashion now.  Many times in the past, however, I have thought that aspect-ratio handling was working adequately, and every time someone has found a major bug.  This time is probably no exception.
>
>There can be some anomalies when the window is resized quickly--that is, grabbing a window corner and slinging it around with axis('equal') can cause some incorrect choices of the view limits.  I haven't figured out why.  Although I certainly would like to understand and fix it, it is not a top priority at the moment because I don't think it is a major problem in practice. 
>  
>