Skip to content

[Bug]: barplot does not show anything when x or bottom start and end with NaN #23687

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
Gullumluvl opened this issue Aug 20, 2022 · 4 comments · Fixed by #23751
Closed

[Bug]: barplot does not show anything when x or bottom start and end with NaN #23687

Gullumluvl opened this issue Aug 20, 2022 · 4 comments · Fixed by #23751

Comments

@Gullumluvl
Copy link

Bug summary

I am not sure if this is a bug or an optimization, but in most cases it is fine to plot bars of arrays containing NaN (only the NaN values are not displayed, the rest is ok).

Here I found that when the x array or the bottom array have NaNs at the extremities, no bar is displayed.

(barh has the same behavior).

Code for reproduction

import numpy as np, matplotlib as mpl, matplotlib.pyplot as plt

barx = np.arange(3, dtype=float)
barheights = np.array([0.5, 1.5, 2.0])
barstarts=np.array([0.77]*3)

barx[[0,2]] = np.NaN
# alternatively:
#barstarts[[0,2]] = np.NaN

fig, ax = plt.subplots()

ax.bar(barx, barheights, bottom=barstarts)
# Not ok, nothing shown!

# Same behavior with ax.barh:
ax.barh(barx, barheights, left=barstarts)

# However, the behavior is correct when NaN only affect the 'heights' argument:

fig, ax = plt.subplots()
ax.bar(np.arange(3), np.array([np.NaN, 1, np.NaN]), bottom=np.array([0.2, 0.3, 0.4]))
# OK, one bar is displayed.

Actual outcome

zero bar displayed

Expected outcome

Non-NaN elements should be shown

Additional information

No response

Operating system

Ubuntu 20.04.3

Matplotlib Version

3.5.3

Matplotlib Backend

TkAgg, pdf

Python version

3.8.10

Jupyter version

ipython

Installation

pip

@tacaswell tacaswell added this to the v3.7.0 milestone Aug 26, 2022
@tacaswell
Copy link
Member

I can reproduce this, but do not understand it yet.

@tacaswell
Copy link
Member

In 7c4b000 / #12903 we added some logic to make sure that width would work with interesting units, notable time deltas and and datetime data.

To do this we effective add the width to the first element of the x data, run it through the conversion machinery, and then subtract off the already converted value of the first data point to convert the "width" in (possibly unitful) user data into our internal version of it. However, if the first point in nan, then the width is poisoned to also be nan which means that when we make the Rectangles the x positions are nan which then (correctly) does not render anything.

@jklymak
Copy link
Member

jklymak commented Aug 26, 2022

I guess we should be using the new safe_first_element excluding NaNs?

@tacaswell
Copy link
Member

The new one only drops None, but I just finished adapting it to work with nan as well. PR shortly....

tacaswell added a commit to tacaswell/matplotlib that referenced this issue Aug 26, 2022
Due to the way we handle units on the bar width having an invalid value in the
first position of the x bar (of y of barh) would effectively poison all of the
widths making all of the bars invisible.

This also renames the cbook function _safe_first_non_none function ->
_safe_first_finite and adjusts the behavior to also drop nans

closes matplotlib#23687
@QuLogic QuLogic modified the milestones: v3.7.0, v3.6.1 Sep 19, 2022
melissawm pushed a commit to melissawm/matplotlib that referenced this issue Dec 19, 2022
Due to the way we handle units on the bar width having an invalid value in the
first position of the x bar (of y of barh) would effectively poison all of the
widths making all of the bars invisible.

This also renames the cbook function _safe_first_non_none function ->
_safe_first_finite and adjusts the behavior to also drop nans

closes matplotlib#23687
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging a pull request may close this issue.

4 participants