plot_histogram fails to operate correctly when numerous starting and ending values are zero

[sanketsharma]

Environment

• Qiskit version: 1.0.1
• Python version: 3.12.2
• Operating system: Red Hat Enterprise Linux release 8.7 (Ootpa)

What is happening?

In the code below, I am trying to plot the probabilities of a particle being at positions given by the dictionary indices. plot_histogram isn't working the way it should in case of the first dictionary d while it works correctly in case of the second dictionary d1. These values are generated by a quantum circuit. In the first case the plot is empty while in the second case, the plot shows a histogram. Also, I get the following warning on the first call to plot_histogram.

/nfs/home/ssharm19/.conda/envs/research/lib/python3.12/site-packages/qiskit/visualization/counts_visualization.py:389: UserWarning: Attempting to set identical low and high ylims makes transformation singular; automatically expanding.
  ax.set_ylim([min_ylim, min([1.1 * sum(all_vals), max(1.1 * val for val in all_vals)])])

How can we reproduce the issue?

from qiskit.visualization import plot_histogram
import matplotlib.pyplot as plt
import numpy as np

d = {0: 0, 1: 0, 2: 0, 3: 0, 4: 0, 5: 0, 6: 0, 7: 0, 8: 0.00324, 9: 0.03881, 10: 0.17425, 11: 0.28411, 12: 0.28588, 13: 0.17294, 14: 0.03769, 15: 0.00308, 16: 0, 17: 0, 18: 0, 19: 0, 20: 0, 21: 0, 22: 0, 23: 0}
assert sum(d.values()) == 1, "Probabilities aren't normalized"
d1 = {k -8: v for k,v in d.items() if v !=0 }
assert sum(d1.values()) == 1, "Probabilities aren't normalized"
plot_histogram(d)
plot_histogram(d1)
plt.show()

What should happen?

I should be able to see histograms in both cases.

Any suggestions?

No response

[jakelishman]

As an immediate workaround, if you convert all the values in your initial dictionary to floats, you should get the expected output (e.g. d = {k: float(v) for k, v in d.items()}).

This happens because, when given a dict, plot_histogram has to guess whether it's been given a dictionary of counts (so should plot in "counts" mode) or a dictionary of probabilities (so should plot in "distribution" mode). Its heuristic to do this only checks the type of a single value, so if it happens to check a value that's an int (such as your zeros), it'll think it's in "counts" mode, and coerces all the values to integers. We can probably fix that by making it examine all the values, and switch to "distribution" mode if any of them are floats.

[ds-mathias]

Hi, I am part of the IBM OpenSource Coalition. This issue looks interesting to me and I would like implement it

[Adithya2685-git]

Hi, is this issue still open? Can I be assigned to this?

[ddri]

I've submitted a fix for this in PR #15497 as it popped up in my workflow this week and saw it was still open.