Histogram range appears to be calculated incorrectly when using "bins='auto'" with 2D data

Bug report

Bug summary

Using 'auto' to compute the number of bins with 2D data only uses the first column to compute the range. This causes all other columns to be incorrectly displayed.

Calculating histograms using bins='auto' with numpy gives the correct behavior.
Code for reproduction

from pylab import *

# four values to examine
x = randn(1000, 4)*0.1 + arange(4)/7

figure()
ax1 = subplot(211)
plt.hist(x, histtype='step')
legend(['1', '2', '3', '4'])
ax2 = subplot(212)
plt.hist(x, bins='auto', histtype='step')
legend(['1', '2', '3', '4'])

# Match axes so it's clear where the problem is
ax2.set_xlim(ax1.get_xlim())
plt.show()

# Check that numpy doesn't have this bug
_, bins_default = np.histogram(x)
_, bins_auto = np.histogram(x, bins='auto')
assert(np.min(bins_default) == np.min(bins_auto))
assert(np.max(bins_default) == np.max(bins_auto))

Actual outcome
The range of the histogram corresponds to the range of the last column of data for all columns, see second subplot.

Expected outcome
The range of the histogram should correspond to the min/max of all data, as it is in the first subplot of the example code.

Matplotlib version

Operating System: Ubuntu
Matplotlib Version: 2.0.0
Python Version: 3.5.2

It looks to me that the problem may come from this section in axes.hist:

for i in xrange(nx):
        # this will automatically overwrite bins,
        # so that each histogram uses the same bins
        m, bins = np.histogram(x[i], bins, weights=w[i], **hist_kwargs)

If the bins parameter that is initially passed to hist is “auto¹”, then bins becomes a sequence after the first iteration, and this iteration will be kept during the following runs. My first idea would be to just add some supplementary logic (in the loop?) to take care of the case of bins being a string.

¹ : maybe I am wrong, but one could pass any string that is valid for numpy.histogram, don't we? Is it deliberate to not mention it in the docstring of ax.hist? Otherwise, it looks like a pretty kind nice but unfortunately hidden feature ^^.

Edit: English...

passing any string was intentional since we knew the plan was for numpy to add more kinds of auto-ranges. If it isn't documented, it should be something along the lines of "passes this value directly to numpy.hist()".

…

On Mon, Jun 19, 2017 at 11:15 AM, Adrien F. Vincent < ***@***.***> wrote: It looks to me that the problem may come from this section <https://github.com/matplotlib/matplotlib/blob/master/lib/matplotlib/axes/_axes.py#L6102> in axes.hist: for i in xrange(nx): # this will automatically overwrite bins, # so that each histogram uses the same bins m, bins = np.histogram(x[i], bins, weights=w[i], **hist_kwargs) If the bins parameter that is initially passed to hist is “auto¹”, then bins becomes a sequence after the first iteration, and this iteration will be kept during the following runs. My first idea would be to just add some supplementary logic (in the loop?) to take care of the case of bins being a string. ¹ : maybe I am wrong, but one could pass any string that is valid for numpy.histogram, don't we? Is it deliberate to not mention it in ax.docstring? Otherwise, it looks like a pretty kind nice but unfortunately hidden feature ^^. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#8778 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AARy-GhOb8cYoyVtegtxfo1tfGy9R1xaks5sFpClgaJpZM4N-WCA> .

Closing as a duplicate of #8636

Indeed (and the on-going PR #8638 may fix this).

Navigation Menu

Repository navigation

Histogram range appears to be calculated incorrectly when using "bins='auto'" with 2D data #8778

Bug report

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Histogram range appears to be calculated incorrectly when using "bins='auto'" with 2D data #8778

Description

Bug report

Activity

afvincent commented on Jun 19, 2017

WeatherGod commented on Jun 19, 2017

tacaswell commented on Jun 19, 2017

afvincent commented on Jun 20, 2017

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions