when matches=None, set categoryarray=None for that axis

[celia-lm]

Current behaviour When a plot is created with facetted rows and xaxes matches is set to None, in the bottom plot the ticklabels for the values that are in other rows are still visible.

import plotly.express as px
import plotly
df = plotly.data.election()
fig = px.scatter(df, x= 'winner', y='total', facet_row="winner")
fig.for_each_xaxis(lambda x: x.update(showticklabels=True, matches=None))
fig.show()

• When a facetted figure is created, the items regarding xaxes of the resulting fig.layout look like this:

    'xaxis': {'anchor': 'y',
              'categoryarray': [Joly, Coderre, Bergeron],
              'categoryorder': 'array',
              'domain': [0.0, 0.98],
              'title': {'text': 'winner'}},
    'xaxis2': {'anchor': 'y2', 'domain': [0.0, 0.98], 'matches': 'x', 'showticklabels': False},
    'xaxis3': {'anchor': 'y3', 'domain': [0.0, 0.98], 'matches': 'x', 'showticklabels': False},

• xaxis2 and xaxis3 refer to the two top rows, while xaxis refers to the bottom one. The main difference between the two groups is that xaxis has some additional items/properties: 'categoryarray': [Joly, Coderre, Bergeron], 'categoryorder': 'array', 'title': {'text': 'winner'}
• When we apply fig.for_each_xaxis(lambda x: x.update(showticklabels=True, matches=None)), the same part of fig.layout gets updated to:

    'xaxis': {'anchor': 'y',
              'categoryarray': [Joly, Coderre, Bergeron],
              'categoryorder': 'array',
              'domain': [0.0, 0.98],
              'showticklabels': True,
              'title': {'text': 'winner'}},
    'xaxis2': {'anchor': 'y2', 'domain': [0.0, 0.98], 'showticklabels': True},
    'xaxis3': {'anchor': 'y3', 'domain': [0.0, 0.98], 'showticklabels': True},

• The categoryarray and categoryorder elements are still there (with the values for the three facets, not only the bottom one), so we need to remove them, which we can do with:

fig.for_each_xaxis(lambda x: x.update(categoryorder=None, categoryarray=None))

or

fig.update_xaxes(categoryarray=None, selector={'anchor':'y'})

• The updated part of the dict is:

    'xaxis': {'anchor': 'y', 'domain': [0.0, 0.98], 'showticklabels': True, 'title': {'text': 'winner'}},
    'xaxis2': {'anchor': 'y2', 'domain': [0.0, 0.98], 'showticklabels': True},
    'xaxis3': {'anchor': 'y3', 'domain': [0.0, 0.98], 'showticklabels': True},

Desired behaviour

• When a plot is created with facetted rows and xaxes matches is set to None, categoryorder should be set to None for those axes.
• If the categoryorder has been specified on plot creation, that order is respected even after setting categoryorder to None. It would only be changed if we set a new value for that argument:

import plotly.express as px
df = px.data.tips()
df['weekend'] = df['day'].apply(lambda x: "weekend" if x in ["Sat", "Sun"] else "weekday")
fig = px.bar(df, x="day", y="total_bill", color="smoker", barmode="group", facet_row="weekend",
             category_orders={
                 "weekend":["weekday", "weekend"],
                 "day": ["Thur", "Fri", "Sat", "Sun"],
                 "smoker": ["Yes", "No"],
                 "sex": ["Male", "Female"]
                 })
fig.for_each_xaxis(lambda x: x.update(showticklabels=True, matches=None))
fig.update_xaxes(categoryarray=None, selector={'anchor':'y'})
fig.show()

fig.update_xaxes(categoryarray=["Sun", "Sat"], selector={'anchor':'y'})

[celia-lm]

Related, similar logic: https://github.com/plotly/plotly.py/issues/4452

The current workaround for that issue is to set bingroup=None