Hello,
I am not sure if this is a Bokeh question or a Pandas question. I am trying to plot x-axis as datetime. It is showing up correctly. However when I switch timezome (from UTC to an American one) for the dataframe column, Bokeh still recognizes the original timezone in the plot. Is there anything I can do for it plot the local time zone instead of UTC?
Time in UTC and conversion to America/Los_Angeles:
Plot still showing up in UTC:
I do see an open issue on Bokeh repo trying to fix this? Bokeh Issue 1135
Wrote a wrapper to remove timezone information and achieved what I needed. Here is what I did:
df_nts_IR['time'] = pd.to_datetime(df_nts_IR['time'])
df_nts_IR['time'] = df_nts_IR['time'].dt.tz_convert('America/Los_Angeles')
df_nts_IR['time_local'] = np.NaN
Could not remove timezone info on the whole column at once so had to iterate over each row of the column:
for index_df in range(0,len(df_nts_IR)):
df_nts_IR['time_local'][index_df] = df_nts_IR['time'][index_df].replace(tzinfo=None)
I use time_local as the x_axis and it works:
@swamilikes2code That old issue is ancient, and probably should be closed just because it’s hard to know what is relevant or not anymore. Nominally I would say that Bokeh strips all tz info and does not send any tz info to BokehJS. The latter statement about not sending tz info is certainly the case, so I’d still say tz-formatting of ticks is best handled by a FuncTickFormatter. But some experiments have me wondering about the first part about stripping tznfo wrt to pandas specifically. Can you construct a small, complete, toy example for me to investigate? It may be that a new issue is in order.
Sounds good, Bryan. Will share one by tomorrow.
Hello @Bryan,
Here is an example. Looking at is it day? in NYC on Dec 20 2020, one of the shortest day of the year.
First row, I am passing the local time. Second row, I am passing UTC transformed to EST time. Code is below.
import pandas as pd
from bokeh.plotting import figure
from bokeh.models import ColumnDataSource, BoxAnnotation, Span, DatetimeTickFormatter
from bokeh.layouts import column, layout
from bokeh.io import show, output_file
from datetime import datetime
import time
import pytz
# import CSV file as a dataframe
df_is_it_day = pd.read_csv(".//NYC_shortest_day_is_it_day.csv")
df_is_it_day
# Convert time zone data
df_is_it_day['time_UTC'] = pd.to_datetime(df_is_it_day['time_UTC'])
df_is_it_day['time_local_EST_NYC'] = pd.to_datetime(df_is_it_day['time_local_EST_NYC'])
df_is_it_day['time_UTC'] = df_is_it_day['time_UTC'].dt.tz_convert('America/New_York')
# Setup Plot
output_file(".//vis_html//is_it_day_NYC.html")
TOOLS = "pan,wheel_zoom,box_zoom,reset,save,box_select,lasso_select"
data_source = ColumnDataSource(df_is_it_day)
# Is it Day in local time
plot_title = "Is it day?: in NYC local time."
local_time_figure = figure(tools=TOOLS, width=900, height=300, x_axis_type='datetime',
title=plot_title)
local_time_figure.xaxis.axis_label = 'Time'
local_time_figure.yaxis.axis_label = 'Is it day?(-)'
local_time_figure.title.text_font_size = '16pt'
local_time_figure.axis.axis_label_text_font_size = '12pt'
local_time_figure.axis.major_label_text_font_size = '12pt'
local_time_figure.axis.major_label_text_font_style = 'bold'
local_time_figure.axis.axis_label_text_color = 'black'
local_time_figure.ygrid.minor_grid_line_color = 'gainsboro'
local_time_figure.circle('time_local_EST_NYC', 'is_it_day', source=data_source,
color='mediumblue', size=10, fill_alpha=0.8)
# Setup properties of figure handle
local_time_figure.xaxis.formatter=DatetimeTickFormatter(days="%m/%d %H:%M", months="%m/%d %H:%M",
hours="%m/%d %H:%M", minutes="%m/%d %H:%M")
# Is it Day in UTC
plot_title = "Is it day?: in NYC local time."
UTC_time_figure = figure(tools=TOOLS, width=900, height=300, x_axis_type='datetime',
title=plot_title)
UTC_time_figure.xaxis.axis_label = 'Time'
UTC_time_figure.yaxis.axis_label = 'Is it day?(-)'
UTC_time_figure.title.text_font_size = '16pt'
UTC_time_figure.axis.axis_label_text_font_size = '12pt'
UTC_time_figure.axis.major_label_text_font_size = '12pt'
UTC_time_figure.axis.major_label_text_font_style = 'bold'
UTC_time_figure.axis.axis_label_text_color = 'black'
UTC_time_figure.ygrid.minor_grid_line_color = 'gainsboro'
UTC_time_figure.circle('time_UTC', 'is_it_day', source=data_source,
color='mediumblue', size=10, fill_alpha=0.8)
# Setup properties of figure handle
UTC_time_figure.xaxis.formatter=DatetimeTickFormatter(days="%m/%d %H:%M", months="%m/%d %H:%M",
hours="%m/%d %H:%M", minutes="%m/%d %H:%M")
# Add night time for Friday to Sunday
# Till Sunrise
start_time = (pd.to_datetime('20201220T0000').timestamp())*1000
end_time = (pd.to_datetime('20201220T0716').timestamp())*1000
morning_night_range = BoxAnnotation(left=start_time, right=end_time, fill_color='silver')
local_time_figure.add_layout(morning_night_range)
# Post Sunset
start_time = (pd.to_datetime('20201220T1631').timestamp())*1000
end_time = (pd.to_datetime('20201221T0000').timestamp())*1000
evening_night_range = BoxAnnotation(left=start_time, right=end_time, fill_color='silver')
local_time_figure.add_layout(evening_night_range)
show(column(local_time_figure, UTC_time_figure))
df_is_it_day
Data from CSV here:
time_UTC,time_local_EST_NYC,is_it_day
2020-12-20T05:00:00Z,2020-12-20T00:00:00,0
2020-12-20T06:00:00Z,2020-12-20T01:00:00,0
2020-12-20T07:00:00Z,2020-12-20T02:00:00,0
2020-12-20T08:00:00Z,2020-12-20T03:00:00,0
2020-12-20T09:00:00Z,2020-12-20T04:00:00,0
2020-12-20T10:00:00Z,2020-12-20T05:00:00,0
2020-12-20T11:00:00Z,2020-12-20T06:00:00,0
2020-12-20T12:00:00Z,2020-12-20T07:00:00,0
2020-12-20T13:00:00Z,2020-12-20T08:00:00,1
2020-12-20T14:00:00Z,2020-12-20T09:00:00,1
2020-12-20T15:00:00Z,2020-12-20T10:00:00,1
2020-12-20T16:00:00Z,2020-12-20T11:00:00,1
2020-12-20T17:00:00Z,2020-12-20T12:00:00,1
2020-12-20T18:00:00Z,2020-12-20T13:00:00,1
2020-12-20T19:00:00Z,2020-12-20T14:00:00,1
2020-12-20T20:00:00Z,2020-12-20T15:00:00,1
2020-12-20T21:00:00Z,2020-12-20T16:00:00,1
2020-12-20T22:00:00Z,2020-12-20T17:00:00,0
2020-12-20T23:00:00Z,2020-12-20T18:00:00,0
2020-12-21T00:00:00Z,2020-12-20T19:00:00,0
2020-12-21T01:00:00Z,2020-12-20T20:00:00,0
2020-12-21T02:00:00Z,2020-12-20T21:00:00,0
2020-12-21T03:00:00Z,2020-12-20T22:00:00,0
2020-12-21T04:00:00Z,2020-12-20T23:00:00,0
To update on my previous solution, you can switch the timezone of a whole pandas column using the following:
df_nts_IR['local_time'] = df_nts_IR['time'].dt.tz_convert('America/Los_Angeles')
df_nts_IR['local_time'] = df_nts_IR['local_time'].dt.tz_localize(None)
time has timezone in UTC and local_time has the local time zone of LA in USA.
Previously I was looping through each element, which is inefficient.