importnumpyasnpimportmatplotlibimportmatplotlib.pyplotaspltimportmatplotlib.colorsascolorsfig,ax=plt.subplots(2,1)test_data1=np.zeros((100,100), )# data at the bottom bound of the rangedisp_obj1=ax[0].imshow (test_data1,norm=colors.PowerNorm(gamma=0.5,vmin=0.0,vmax=1.0,),)# ************ to reproduce the bug, use the line with 'both' or 'min' extends *************#fig.colorbar(disp_obj1, ax=ax[0], )fig.colorbar(disp_obj1,ax=ax[0],extend='max')#fig.colorbar(disp_obj1, ax=ax[0], extend='min')#fig.colorbar(disp_obj1, ax=ax[0], extend='both')test_data2=np.ones((100,100), )*0.25# all positive data that's inside the rangedisp_obj2=ax[1].imshow (test_data2,norm=colors.PowerNorm(gamma=0.5,vmin=0.0,vmax=1.0,),)# ************ to reproduce the bug, use the line with 'both' or 'min' extends *************#fig.colorbar(disp_obj2, ax=ax[1], )#fig.colorbar(disp_obj2, ax=ax[1], extend='max')fig.colorbar(disp_obj2,ax=ax[1],extend='min')#fig.colorbar(disp_obj2, ax=ax[1], extend='both')plt.show()

Before:

/Users/jklymak/matplotlib/lib/matplotlib/colors.py:1201: RuntimeWarning: invalid value encountered in power  np.power(resdat, gamma, resdat)/Users/jklymak/matplotlib/lib/matplotlib/colorbar.py:999: RuntimeWarning: invalid value encountered in true_divide  z = np.take(y, i0) + (xn - np.take(b, i0)) * dy / db

After:

PR Checklist

Has Pytest style unit tests
Code is PEP 8 compliant

tacaswell added this to thev3.0 milestone

Jul 9, 2018

tacaswell approved these changes

Jul 9, 2018

View reviewed changes

timhoffm approved these changes

Jul 9, 2018

View reviewed changes

Copy link

Member

timhoffm left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others.Learn more.

Minor style issue.

lib/matplotlib/colorbar.py Outdated

		if isinstance(self.norm, colors.LogNorm):
		# If using a lognorm, ensure extensions don't go negative
		if (isinstance(self.norm, colors.LogNorm) or
		isinstance(self.norm, colors.PowerNorm)):

Copy link

Member

timhoffmJul 9, 2018•
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others.Learn more.

if isinstance(self.norm, (colors.lOGNorm, colors.PowerNorm)):

jklymak force-pushed thefix-extend-low-powernorm branch 3 times, most recently from15209b6 toe43763aCompare

July 9, 2018 20:58

dstansby approved these changes

Jul 9, 2018

View reviewed changes

lib/matplotlib/colorbar.py

		@@ -802,8 +802,9 @@ def _process_values(self, b=None):

		b = self.norm.inverse(self._uniform_y(self.cmap.N + 1))

		if isinstance(self.norm, colors.LogNorm):
		# If using a lognorm, ensure extensions don't go negative
		if isinstance(self.norm, (colors.PowerNorm, colors.LogNorm)):

Copy link

Member

dstansbyJul 9, 2018

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others.Learn more.

I feel like this kind of stuff should be handled by the specificNorm class, instead of being "special cased" here in this code, but since this is howLogNorm is dealt with at the moment I guess this is fine as a fix.

Copy link

MemberAuthor

jklymakJul 9, 2018

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others.Learn more.

Hmmm, interesting idea. This "b" is just the points that are used to make the pcolor mesh that makes the colorbar. So its a bit of a toss up where making that list should get created and how.

Copy link

Member

dstansby commentedJul 9, 2018

Looks like this is failing on the pytz tests,@pganssle any ideas what's going on with the travis build?

Copy link

MemberAuthor

jklymak commentedJul 10, 2018

I think its coming from flake-8. I guess we should rerun flake8 on any new merges.

./lib/mpl_toolkits/mplot3d/axes3d.py:0:1: X100 Superfluous per-file-ignores for E3031     X100 Superfluous per-file-ignores for E303

jklymak added2 commits

July 9, 2018 17:28

FIX: let extends work for PowerNorm

1eccdbe

TST: fix flake8 error

092e4da

jklymak force-pushed thefix-extend-low-powernorm branch frome43763a to092e4daCompare

July 10, 2018 00:29

Copy link

Member

pganssle commentedJul 10, 2018•
edited
Loading

Maybeflake8 should be run on a whitelist instead of a blacklist? I didn't do theflake8 stuff, though, so I may not be much help.

Copy link

Member

QuLogic commentedJul 10, 2018

pytz testsstarted failing on master before theflake8 issue.

Copy link

Member

pganssle commentedJul 10, 2018

I dunno if this is the issue to talk about it on, but it occurs to me that thetox -e pytz command canprobably be folded into the tests withpandas, sincepytz is a dependency ofpandas anyway (... for now). If it's causing problems that's probably the most expedient way to fix it.