annotate

Christopher_Brown · July 14, 2011, 8:59pm

Hi,

The following code displays a figure with both a red and a blue arrow (generated by annotate). But the pdf that is saved only shows the red one. It seems that the line ax1.set_xscale('log') is causing the trouble, because if it is commented out, the blue arrow reappears in the pdf. Any hints?

import numpy as np
from matplotlib import pyplot as pp

fig = pp.figure()
ax2 = fig.add_subplot(111)
ax1 = ax2.twinx()
ax1.set_xlim(xmin=100,xmax=10000)
ax1.set_ylim(ymin=-60, ymax=5)
ax1.set_xscale('log')
ax2.set_xlim(xmin=100,xmax=10000)
ax2.set_ylim(ymin=-5,ymax=35)
ax2.set_xscale('log')

ax1.annotate("", (10000,-30), (2000,-30),
arrowprops=dict(arrowstyle="-|>",fc="b", ec="b"))
ax2.annotate("", (100, 5), (2000, 5),
arrowprops=dict(arrowstyle="-|>",fc="r", ec="r"))
pp.savefig('test_arrow.pdf')
pp.show()

···

--
Christopher Brown, Ph.D.
Associate Research Professor
Department of Speech and Hearing Science
Arizona State University
http://pal.asu.edu

Michael_Droettboom1 · July 18, 2011, 8:07pm

Thanks for the report. I have created a pull request with a solution for this bug here:

github.com/matplotlib/matplotlib

Log annotate bug

matplotlib:master ← mdboom:log_annotate_bug

opened 08:05PM - 18 Jul 11 UTC

mdboom

+35 -27

This was reported on matplotlib-users by Christopher Brown. The test to determi…ne whether an annotation is on the axes, and thus should be displayed, is a little too sensitive to floating point rounding error. When the axes scale is log, for example, and the annotation is placed right at the edge of the axes, it often is considered outside of the axes and disappears (and this is dpi-dependent). The fix is to use Agg's "contour" functionality to expand the size of the axes by a single pixel and hit-test within that. This contour/radius functionality for hit testing may come in handy in other places in the future as well, and thus is now exposed at the Python level. This example also illustrated a related bug in the SVG backend when mathtext and regular text using the same font are mixed (some characters appear upside down).

If you are able to build from git, test these changes and give me feedback in your specific environment, that would be great.

Cheers,
Mike

···

On 07/14/2011 04:59 PM, Christopher Brown wrote:

Hi,

The following code displays a figure with both a red and a blue arrow
(generated by annotate). But the pdf that is saved only shows the red
one. It seems that the line ax1.set_xscale('log') is causing the
trouble, because if it is commented out, the blue arrow reappears in the
pdf. Any hints?

import numpy as np
from matplotlib import pyplot as pp

fig = pp.figure()
ax2 = fig.add_subplot(111)
ax1 = ax2.twinx()
ax1.set_xlim(xmin=100,xmax=10000)
ax1.set_ylim(ymin=-60, ymax=5)
ax1.set_xscale('log')
ax2.set_xlim(xmin=100,xmax=10000)
ax2.set_ylim(ymin=-5,ymax=35)
ax2.set_xscale('log')

ax1.annotate("", (10000,-30), (2000,-30),
arrowprops=dict(arrowstyle="-|>",fc="b", ec="b"))
ax2.annotate("", (100, 5), (2000, 5),
arrowprops=dict(arrowstyle="-|>",fc="r", ec="r"))
pp.savefig('test_arrow.pdf')
pp.show()

--
Michael Droettboom
Science Software Branch
Space Telescope Science Institute
Baltimore, Maryland, USA