update_xaxes(categoryorder="total descending") does not work correctly on px.histogram

Hello,
When using update_xaxes(categoryorder="total descending") on a histogram, I expect the bars to be reordered by size.
The following example, tested in a Notebook, will not work according to my expectation:

import pandas as pd
import plotly.express as px

data =[['A','left'],['B','left'],['C','right'],['D','up'],['E','right'],['F','down'],
       ['A','right'],['B','righ'],['B','right'],['B','up'],['A','left'],['F','up'],
       ['B','left'],['F','left'],['B','right'],['A','up'],['A','right'],['F','down'],
      ]
df = pd.DataFrame(data, columns=['letter','direction'])
px.histogram(df, x='letter', color='direction').update_xaxes(categoryorder="total descending")

Note that the behavior is unclear to me. In certain condition, the order might be correct. For instance, if you change the last row to ['E','down'], then the order of the plot is correct.

In any case, I would expect the same result than the following:

import pandas as pd
import plotly.express as px

data =[['A','left'],['B','left'],['C','right'],['D','up'],['E','right'],['F','down'],
       ['A','right'],['B','righ'],['B','right'],['B','up'],['A','left'],['F','up'],
       ['B','left'],['F','left'],['B','right'],['A','up'],['A','right'],['E','down'],
      ]

df = pd.DataFrame(data, columns=['letter','direction'])
df['count'] = 1
df = df.groupby(['letter','direction'])['count'].sum().reset_index()
px.bar(df, x='letter', y='count', color='direction').update_xaxes(categoryorder="total descending")

[nicolaskruchten]

Unfortunately I cannot reproduce this... your first example orders correctly for me:

And your second does also... Could you try upgrading to the latest version of plotly which is 4.6.0 to see if this is a bug we've already fixed?

So, I was indeed on 4.5.2 but I upgraded and still the same:

Strange... What else could it be?
I'm on Win10 + Anaconda3.

I downgraded pandas to 0.25, because who knows, but to no avail.

[wcj365]

I have the same issue.
Plotly: 4.5.2
Pandas: 1.0.1
Ubuntu on AWS
The sorting works if I remove the color option.

[audelorAd]

Hello,
I have the same issue too.
Sorting isn't working with histogram but with bar chart it's ok.
Tried with latest version pandas : 1.1.2 & plotly : 4.10.0

[jframe-fid]

I see this same issue with histograms, and switching to bar fixes it. I would love for it to be fixed so I can move back to the simplicity of histograms.

plotly: 4.11.0
pandas: 1.0.5

[nicolaskruchten]

Are folks who are encountering this bug running JupyterLab? If so, please note that the version of the jupyterlab-plotly labextension must match the version of the plotly Python module to run correctly... See https://plotly.com/python/troubleshooting/#jupyterlab-problems for more details.

If this is happening in a "classic" Jupyter Notebook or is happening in JupyterLab with an up-to-date extension then there is definitely something to fix here.

Here is a Javascript-only version of the figure produced by this Python call: https://codepen.io/nicolaskruchten/pen/ExyWMZj ... if someone who is seeing this issue can please click on that link and tell me if things are rendering in or out of order (in the same browser/on the same machine as they're seeing the problem), that might help narrow down the possibilities!

Note: the expected order is BAFEDC

Hello @nicolaskruchten, I just tested on:

• Win10 + Jupyter Notebook + Plotly 4.12
• Manjaro + Jupyter Notebook + Plotly 4.11

and the problem was there on both. However, your codepen looks fine.

[audelorAd]

@nicolaskruchten I'm testing in a Dash App and with Terminal with same result.

I'm working with data in this attach file ( test_dataframe.txt ) and running the following exmeple :

col = ['lib','type','year','sup']

# Get data with attach file
df = pd.DataFrame(data, columns=col)

fig1 = px.histogram(df, x='year', y='sup', color='type').update_xaxes(type="category", categoryorder="total descending")

fig1.show()

fig2 = px.histogram(df, x='lib', y='sup', color='type').update_xaxes(type="category", categoryorder="total descending")

fig2.show()

fig1 is always displaying in the right order and fig2 always displaying in the wrong order

I really can't guess a logic that would generate that order...!
@audelorAd: you are studying Breton names, aren't you?!

[jframe-fid]

When I create a color by histogram, it never comes out correctly, and like @Kerybas I can't figure out how it's ordering things. Even if it acts up with more than one color bin, if I turn off all colors but one, it orders properly again. If I turn even one color back on though, it goes back to the really odd ordering.

[audelorAd]

I really can't guess a logic that would generate that order...!
@audelorAd: you are studying Breton names, aren't you?!

Yes I am ;)

[jenna-jordan]

I am also having this issue, but with px.bar rather than px.histogram
Like @wcj365 said, the problem goes away if you don't pass a "color" argument. However, that kinda defeats the point of a stacked bar chart...
I am rendering this with streamlit, not in a jupyter notebook.
code:

df = df.sort_values(by=["company_count", "Category"])
    fig = px.bar(
        df,
        y="Keyword",
        x="company_count",
        color="Category",
        barmode="stack",
        template="plotly_white",
        orientation="h",
        hover_name="Keyword",
        text="company_count",
        color_discrete_sequence=px.colors.qualitative.Vivid,
    )
    fig.update_yaxes(
        type="category",
        categoryorder="total ascending")

Plot:

[nicolaskruchten]

OK, we've got a fix in the underlying JS engine, which will become available with version 4.13 (edit: 4.14 now) of Plotly for Python. Hopefully that'll resolve all these issues!

[jframe-fid]

I've updated to plotly 4.13 today, and I'm still seeing the odd ordering behavior with px.histogram. Has anyone else seen this fixed updating to 4.13?