ENH: Flip order of bar chart bars for multi-column data about pandas HOT 7 OPEN

Sorry, I got confused.

The expected behavior of matplotlib is for the categories to be displayed from bottom-to-top.
But within the categories, the bars should display in proper order of the legend.

The current barh function gives this plot with the sample code:

Flipping it with plt.gca().invert_yaxis() gives:

What we would ideally want is a barh function that preserves the X-Y-Z ordering from bottom to top while ordering the bars as C-B-A from bottom to top.

I am working on implementing this.

from pandas.

Thanks for the report. Can you provide a full reproducible example.

from pandas.

I was able to reproduce it with the following simple bar plot:

data = pd.DataFrame({
    'A': [3, 4, 2],
    'B': [1, 2, 3],
    'C': [4, 1, 2]
}, index=['X', 'Y', 'Z'])

data.plot.barh()
plt.show()

This is the produced output.

As reported, the legend follows the reverse order of the bars.

If this should be fixed I can begin working on a solution.

from pandas.

Thanks, the solution would just be plt.gca().invert_yaxis() I think.

from pandas.

I agree - this is a bit odd. I'm +1 on a fixing.

from pandas.

I did some further looking and this is expected behavior of matplotlib. The only fix then is to hardcode a y_axis flip. This would cause a lot of issues with backward compatibility and is generally bad practice. I think it should be left as is.

from pandas.

Thanks for reporting back.

I did some further looking and this is expected behavior of matplotlib.

Are there any links you can share in this regard?

from pandas.