matplotlib 3D surface - gaps / holes in surface

Bug report

Bug summary

A grid is interpolated with scipy's griddata() and contains values and NaNs. However, when the grid is plotted with mplot3d's plot_surface() there are gaps in the surface (see upper plot). The lower the view height, the more of such gaps are rendered. When the grid is plotted in 2D with imshow() there are no gaps (see lower plot).

Code for reproduction

import numpy as np
import matplotlib.pyplot as plt
from mpl_toolkits.mplot3d import Axes3D

Z = np.loadtxt('data.txt')
X = np.linspace(0,3,301)
Y = np.linspace(0,3,301)
X = np.array([X,]*(301)).transpose()
Y = np.array([Y,]*(301))
fig = plt.figure(figsize=(15,15), dpi=100)
ax = fig.add_subplot(2, 1, 1, projection='3d')
ax.view_init(40,300)
surf = ax.plot_surface(X, Y, Z, cmap='jet', rstride=1, cstride=1, antialiased=False, shade=False, alpha=1.0, linewidth=0, vmin=0.25, vmax=0.35);

ax.invert_yaxis()
ax.dist = 11

ax = fig.add_subplot(2, 1, 2)
plt.imshow(Z, cmap='jet', vmin=0.25, vmax=0.35, origin='lower');

plt.show()

Actual outcome

Close-up:

Expected outcome

No holes in the 3D surface plot.

Matplotlib version

Matplotlib 1.3.0 to 1.5.3 (WinPython 3.3.2.3 to 3.5.2.3), but it was also reproduced up to Matplotlib 2.0: http://stackoverflow.com/questions/41268550/matplotlib-3d-surface-gaps-holes-in-surface

why would you expect no holes for the surface if the grid has NaNs?

…

On Tue, Mar 7, 2017 at 8:19 AM, fmevers ***@***.***> wrote: Bug report *Bug summary* A grid is interpolated with scipy's griddata() and contains values and NaNs. However, when the grid is plotted with mplot3d's plot_surface() there are gaps in the surface (see upper plot). The lower the view height, the more of such gaps are rendered. When the grid is plotted in 2D with imshow() there are no gaps (see lower plot). *Code for reproduction* import numpy as npimport matplotlib.pyplot as pltfrom mpl_toolkits.mplot3d import Axes3D Z = np.loadtxt('data.txt') X = np.linspace(0,3,301) Y = np.linspace(0,3,301) X = np.array([X,]*(301)).transpose() Y = np.array([Y,]*(301)) fig = plt.figure(figsize=(15,15), dpi=100) ax = fig.add_subplot(2, 1, 1, projection='3d') plt.hold(True) ax.view_init(40,300) surf = ax.plot_surface(X, Y, Z, cmap='jet', rstride=1, cstride=1, antialiased=False, shade=False, alpha=1.0, linewidth=0, vmin=0.25, vmax=0.35); ax.invert_yaxis() ax.dist = 11 ax = fig.add_subplot(2, 1, 2) plt.hold(True) plt.imshow(Z, cmap='jet', vmin=0.25, vmax=0.35, origin='lower'); plt.show() *Actual outcome* <https://camo.githubusercontent.com/41f23a4df762cff426c5c9848c83decddf50190e/68747470733a2f2f692e737461636b2e696d6775722e636f6d2f72536a30752e706e67> Close-up: <https://camo.githubusercontent.com/b72cb7d99580d00453e93b5a57c434f2e2b35f1e/68747470733a2f2f692e737461636b2e696d6775722e636f6d2f7473646b682e706e67> *Expected outcome* No holes in the 3D surface plot. *Matplotlib version* Matplotlib 1.3.0 to 1.5.3 (WinPython 3.3.2.3 to 3.5.2.3), but it was also reproduced up to Matplotlib 2.0: http://stackoverflow.com/ questions/41268550/matplotlib-3d-surface-gaps-holes-in-surface — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#8222>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AARy-FFO4u20mblt_XgmyeAlDkatstM_ks5rjVlUgaJpZM4MVeBX> .

Ah, I see, there are holes appearing where there shouldn't be any holes. Well, this is a known issue with plot_surface(). It has never supported NaNs or masked arrays, unfortunately. Yes, this is a departure from just about every single other matplotlib function, but the nature of plot_surface() makes it difficult to properly handle them. If you wish to try your hand at adding support for it, I encourage you to give it a try!

…

On Tue, Mar 7, 2017 at 9:46 AM, Benjamin Root ***@***.***> wrote: why would you expect no holes for the surface if the grid has NaNs? On Tue, Mar 7, 2017 at 8:19 AM, fmevers ***@***.***> wrote: > Bug report > > *Bug summary* > > A grid is interpolated with scipy's griddata() and contains values and > NaNs. However, when the grid is plotted with mplot3d's plot_surface() > there are gaps in the surface (see upper plot). The lower the view height, > the more of such gaps are rendered. When the grid is plotted in 2D with > imshow() there are no gaps (see lower plot). > > *Code for reproduction* > > import numpy as npimport matplotlib.pyplot as pltfrom mpl_toolkits.mplot3d import Axes3D > > Z = np.loadtxt('data.txt') > X = np.linspace(0,3,301) > Y = np.linspace(0,3,301) > X = np.array([X,]*(301)).transpose() > Y = np.array([Y,]*(301)) > fig = plt.figure(figsize=(15,15), dpi=100) > ax = fig.add_subplot(2, 1, 1, projection='3d') > plt.hold(True) > ax.view_init(40,300) > surf = ax.plot_surface(X, Y, Z, cmap='jet', rstride=1, cstride=1, antialiased=False, shade=False, alpha=1.0, linewidth=0, vmin=0.25, vmax=0.35); > > ax.invert_yaxis() > ax.dist = 11 > > ax = fig.add_subplot(2, 1, 2) > plt.hold(True) > plt.imshow(Z, cmap='jet', vmin=0.25, vmax=0.35, origin='lower'); > > plt.show() > > *Actual outcome* > > > <https://camo.githubusercontent.com/41f23a4df762cff426c5c9848c83decddf50190e/68747470733a2f2f692e737461636b2e696d6775722e636f6d2f72536a30752e706e67> > Close-up: > > > <https://camo.githubusercontent.com/b72cb7d99580d00453e93b5a57c434f2e2b35f1e/68747470733a2f2f692e737461636b2e696d6775722e636f6d2f7473646b682e706e67> > > *Expected outcome* > > No holes in the 3D surface plot. > > *Matplotlib version* > > Matplotlib 1.3.0 to 1.5.3 (WinPython 3.3.2.3 to 3.5.2.3), but it was also > reproduced up to Matplotlib 2.0: http://stackoverflow.com/quest > ions/41268550/matplotlib-3d-surface-gaps-holes-in-surface > > — > You are receiving this because you are subscribed to this thread. > Reply to this email directly, view it on GitHub > <#8222>, or mute the > thread > <https://github.com/notifications/unsubscribe-auth/AARy-FFO4u20mblt_XgmyeAlDkatstM_ks5rjVlUgaJpZM4MVeBX> > . >

Did I get you right, that if I create an array with values instead of NaNs, plot_surface() would be able to handle the array correctly? This would provide me a workaround. Thank you for your encouragement! But, to be honest, adding support would exceed my coding skills by light-years.

Can you attach the data that causes the problem?

This is the data set (data.txt): https://gist.github.com/anonymous/bf886433298528bfa2f0f80448ddaa6b

I still have that problem with matplotlib 3.0.3. Would be happy to know about a workaround or solution.

Same issue!

This is indeed related to the NaN values.

If one does Z[np.isnan(Z)] = np.nanmean(Z) the artifacts are gone.

Note: Updated the example code since plt.hold(True) is no more (and is unneeded) in matplotlib 3.1.

Navigation Menu

Repository navigation

matplotlib 3D surface - gaps / holes in surface #8222

Bug report

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

matplotlib 3D surface - gaps / holes in surface #8222

Description

Bug report

Activity

WeatherGod commented on Mar 7, 2017

WeatherGod commented on Mar 7, 2017

fmevers commented on Mar 7, 2017

eric-wieser commented on Jul 12, 2018

fmevers commented on Jul 20, 2018

kbruegge commented on Mar 7, 2019

franchesoni commented on Oct 16, 2019

timhoffm commented on Oct 24, 2019

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions