Bar Chart of sleep time duration-CodePudding

I have the foloowing dataset:

user_id: name of user - eg: 'user1'
start: when user started sleeping - eg: '2022-09-24 22:00:00 02:00'
stop: when user stopped sleeping - eg: '2022-09-25 04:30:00 02:00'

Desired output:

I would like to plot a bar/line chart with:

x axis: time of clock (0 to 24 hours)
y axis: user_id

With each bar showing when user started and stopped sleeping.

Current results:

I am able to plot the bar chart with x axis "minutes from midnight". Here is the code:

import pandas as pd 
import numpy as np
import plotly.graph_objs as go
from datetime import timedelta
import datetime

df = pd.DataFrame({'user_id':  ['user1', 'user2', 'user3', 'user4'],
                   'start':  ['2022-09-24 22:00:00 02:00','2022-09-04 23:30:00 02:00', '2022-11-11 02:20:30 02:00',  '2022-11-12 03:20:30 02:00'],
                   'stop':   ['2022-09-25 04:30:00 02:00','2022-09-05 07:15:30 02:00', '2022-11-11 11:20:30 02:00',  '2022-11-12 10:40:30 02:00']})

 # change to datetime
datetime_columns = ['start', 'stop']
df[datetime_columns]= df[datetime_columns].apply(pd.to_datetime, errors='coerce')

def get_time_from_midnight (datetime_series):
    # if above 12 you get next day, if below you get same day
    conditions_midnight = [
    datetime_series.dt.hour > 12,
    datetime_series.dt.hour <= 12]
    choiches_midnight = [(datetime_series - (datetime_series.dt.normalize()   datetime.timedelta(days=1))) / pd.Timedelta(minutes=1),
    (datetime_series - (datetime_series.dt.normalize())) / pd.Timedelta(minutes=1)]

    time_from_midnight = np.select(conditions_midnight, choiches_midnight, default=np.nan) 
    return time_from_midnight


df['start_minutes_from_midnight'] = get_time_from_midnight(df['start'])

df['time_in_bed'] = (df['stop'] -df['start'])/ timedelta(minutes=1)


fig = go.Figure()
fig.add_bar(y=df['user_id'],
            x=df['time_in_bed'],
            base=df['start_minutes_from_midnight'],
            orientation='h')
fig.update_layout(title_text='Current result: sleep time each user', xaxis_title="minutes_from_midnight")

fig.show()

However I am not able to have on the X-axis the "clock time"

EDIT current progress I have found a possible solution with px.timeline, howver it doens't work when:

start before midight (eg 22:00)
end after midnight (eg: 4:30) where in this case it revereses the order (it starts at 4:30 and ends at 22)

The ideal solution would be centred at midnight.

import pandas as pd 
import numpy as np
from datetime import datetime
import plotly.express as px




df = pd.DataFrame({'Task':     ['user1', 'user2', 'user3', 'user4'],
                   'Start':    ['2022-09-24 22:00:00 02:00','2022-09-04 23:30:00 02:00', '2022-11-11 02:20:30 02:00',  '2022-11-12 03:20:30 02:00'],
                   'Finish':   ['2022-09-25 04:30:00 02:00','2022-09-05 07:15:30 02:00', '2022-11-11 11:20:30 02:00',  '2022-11-12 10:40:30 02:00']})

datetime_columns = ['Start', 'Finish']
 #remove Year month day so it plots all within 24hr period
df[datetime_columns]= df[datetime_columns].apply(pd.to_datetime, errors='coerce')
df['Start'] =  df['Start'].dt.strftime('%H:%M:%S')
df['Finish'] =  df['Finish'].dt.strftime('%H:%M:%S')
df[datetime_columns]= df[datetime_columns].apply(pd.to_datetime, errors='coerce')


fig = px.timeline(df, x_start="Start", x_end="Finish", y="Task")
fig.update_yaxes(autorange="reversed") # otherwise tasks are listed from the bottom up
fig.update_xaxes(
    tickformat="%H:%M")
fig.update_layout(height=400, width=800)
fig.show()

CodePudding user response：

I think what you're looking for is a gantt chart. See the following example:

import pandas as pd
import numpy as np
import plotly.graph_objs as go
from datetime import timedelta
from datetime import datetime
import plotly.figure_factory as ff
from dateutil.parser import parse

df = pd.DataFrame({'Task':  ['user1', 'user2', 'user3', 'user4'],
                   'Start':    ['2022-09-24 22:00:00 02:00', '2022-09-24 23:30:00 02:00', '2022-09-25 02:20:30 02:00',  '2022-09-25 03:20:30 02:00'],
                   'Finish':   ['2022-09-25 04:30:00 02:00', '2022-09-25 07:15:30 02:00', '2022-09-25 11:20:30 02:00',  '2022-09-25 10:40:30 02:00']})


datetime_columns = ['Start', 'Finish']
df[datetime_columns] = df[datetime_columns].apply(
    pd.to_datetime, errors='coerce')


fig = ff.create_gantt(df,
                      showgrid_x=True, showgrid_y=False,
                      show_colorbar=True, bar_width=0.025)
fig.update_layout(height=400,
                  width=800,
                  xaxis_range=[parse('2022-09-24 12:00:00 02:00'), parse('2022-09-25 12:00:00 02:00')])
fig.update_xaxes(
    tickformat="%H:%M",
    autorange=False)
fig.show()

Notice that I 1) removed the day and month information from your dataframe, and 2) changed the tickformat information to show only hours and minutes (otherwise it would put today's date as a default).