Barpolar heatmap doesn't show all values in plotly 5.13.0

[Rfromearth]

Hello,

I'm trying to create a polar heatmap with the days of the week as r-values and the hour of the day as theta-values.
Minimal example:

import plotly.graph_objects as go
import numpy as np
import pandas as pd
import calendar
import locale

locale.setlocale(locale.LC_TIME, locale.normalize("en_US.utf8"))

def hr2angle(hr):
    return (hr * 15) % 360

def hr_str(hr):
    return str(hr) + " Uhr"

df1 = pd.DataFrame({'weekday': list(calendar.day_name)})    # 'Monday', 'Tuesday',...,'Sunday' -> radial axis
df2 = pd.DataFrame({'hour': list(range(24))})               # 0, 1,..., 23 -> angular axis
df2['theta'] = df2['hour'] * 360 / 24 + 360 / 24 / 2        # hour to angle + 1/2 slice extra rotation

df3 = pd.merge(df1, df2, how='cross')                       # dataframe with 7x24=168 rows
df3['frequency'] = np.random.randint(0, 20, df3.shape[0])   # Add random frequencies

# print(df3)
#     weekday  hour  theta  frequency
# 0    Monday     0    7.5         17
# 1    Monday     1   22.5         15
# 2    Monday     2   37.5         11
# 3    Monday     3   52.5          4
# 4    Monday     4   67.5          2
# ..      ...   ...    ...        ...
# 163  Sunday    19  292.5          4
# 164  Sunday    20  307.5          4
# 165  Sunday    21  322.5         19
# 166  Sunday    22  337.5         17
# 167  Sunday    23  352.5          2
#
# [168 rows x 4 columns]

fig = go.Figure(go.Barpolar(r=df3['weekday'],
                            theta=df3['theta'],
                            marker_color=df3['frequency'],
                            marker_colorscale='ylgn'
                           )
)

fig.update_layout(polar_angularaxis_rotation=90,
                  polar_angularaxis_direction='clockwise',
                  polar_angularaxis_tickmode='array',
                  polar_angularaxis_tickvals=[hr2angle(hr) for hr in range(24)],
                  polar_angularaxis_ticktext=[hr_str(hr) for hr in range(24)],
                  polar_bargap=0
)

fig.show()

Despite of wrong placed r-axis ticks, the plot shows an empty innermost circle instead of the segment for the 'Monday'-values.
When I add e.g. 'Extraday' by replacing the line

df1 = pd.DataFrame({'weekday': list(calendar.day_name)})

with

daylist = ['Extraday']
daylist.extend(list(calendar.day_name))
df1 = pd.DataFrame({'weekday': daylist})                    # 'Extraday', 'Monday'...,'Sunday' (8 rows)

I can see the values for Monday.

[AaronStiff]

Hi @Rfromearth.

This behaviour is by design when r is given a list of string values. Because there are no numerical values for Barpolar to use, it infers numerical values starting from 0, e.g. Monday=0, Tuesday=1 etc, hence why adding an "extra" day at the beginning of the list allows Monday to have a radial coordinate of 1

[AaronStiff]

Perhaps this would be the way to do what you're intending?

import plotly.graph_objects as go
import numpy as np
import pandas as pd
import calendar
import locale

locale.setlocale(locale.LC_TIME, locale.normalize("en_US.utf8"))

fig = go.Figure()

frequencies = np.random.randint(20,size=(7,24))
daynames = list(calendar.day_name)

for i in range(7):
    fig.add_trace(go.Barpolar(
        r=[1 for x in range(24)],
        name=daynames[i],
        marker_color=frequencies[i],
        marker_colorscale='ylgn'
    ))

fig.update_layout(polar_angularaxis_rotation=97.5,
                  polar_angularaxis_direction='clockwise',
                  polar_angularaxis_tickmode='array',
                  polar_angularaxis_tickvals=[i + 7.5 for i in range(0, 360, 15)],
                  polar_angularaxis_ticktext=[str(i)+' Uhr' for i in range(24)],
                  polar_bargap=0,
                  polar_radialaxis_tickmode='array',
                  polar_radialaxis_tickvals=[i for i in range(1,8)],
                  polar_radialaxis_ticktext=daynames
)

fig.show()

[Rfromearth]

Thank you @AaronStiff for diving into my little problem. Your solution opened my eyes for other options to create graph objects in plotly. Interesting to see that every graphing library comes with it's own philosophy.

Btw. adding

hovertemplate=daynames[i] + ': Sum of daily frequencies {}'.format(frequencies.sum(axis=1)[i]),

to the fig.add_trace-part in your solution would even correct the hover-info. But it does not solve my problem because of the following reasons:

1. The calculated colors of the heatmap don't rely on the common reference values (the 'frequencies) I want to visualize. Every trace has it's own color values which are decoupled from the other traces.
   Because of this it's also impossible to show a colorbar for the frequencies.

2. I see no chance to calculate the hourly sum of the frequencies inside fig.add_trace to show the values in the hover-info which is important for me.

Nevertheless thank you very much.

[AaronStiff]

You're very welcome @Rfromearth!

I believe the first of your problems can be solved by adding

marker_coloraxis='coloraxis'

to fig.add_trace, and then moving the colorscale specification into update_coloraxes with

fig.update_coloraxes(colorscale='ylgn')

somewhere before fig.show(). Optionally, showscale=False will hide the colorbar.

This forces each trace to share the same color axis and thus the coloring should be consistent across traces.

I'm not sure I understand the second issue, but are you wanting to display the sum of all the hours up to the current hour for each hour?

[Rfromearth]

Hello @AaronStiff,

you're absolutly right. I wasn't aware of this part of the documentation.

[..] color axes connect color scales, color ranges and color bars to a trace's data. By default, any colorable attribute in a trace is attached to its own local color axis, but color axes may also be shared across attributes and traces by setting e.g. marker.coloraxis in go.Scatter traces or coloraxis in go.Heatmap traces. Local color axis attributes are configured within traces e.g. marker.showscale whereas shared color axis attributes are configured within the Layout e.g. layout.coloraxis.showscale.

Meanwhile I solved also my second demand, which was to calculate the values in the last row and show them in the hover-info:

I will prepare a working example in the next days and put it here. Out there I found a lot of questions regarding the same topic but also a lot of insufficient answers.

Thank you very very much Aaron.

Best regards from Germany
R@iner

[AaronStiff]

No problem! Glad you were able to find a solution.