Bokeh not displaying plot for pandas

Question

I can't get Bokeh to display my plot. This is my Python code.

import pandas as pd
from bokeh.plotting import figure, ColumnDataSource
from bokeh.io import output_file, show


if __name__ == '__main__':
    file = 'Overview Data.csv'
    overview_df = pd.read_csv(file)
    overview_ds = ColumnDataSource(overview_df)
    output_file('Wins across Seasons.html')
    print(overview_ds.data)
    p = figure(plot_width=400, plot_height=400)

    # add a circle renderer with a size, color, and alpha
    p.circle('Season', 'Wins', source = overview_ds, size=20, color="navy", alpha=0.5)

    # show the results
    show(p)

I checked my Chrome browser Inspect Element and the console shows the following.

Wins across Seasons.html:17 [bokeh] could not set initial ranges e.set_initial_range @ Wins across Seasons.html:17

This only seems to happen when I am reading from a file. Hard-coding x and y coordinates work.

I have checked other posts but none of the fixes worked. All my packages are up to date.

This is the file I am reading

Season,Matches Played,Wins,Losses,Goals,Goals Conceded,Clean Sheets
2011-12,38,28,5,89,33,20
2010-11,38,23,4,78,37,15
2009-10,38,27,7,86,28,19
2008-09,38,28,4,68,24,24
2007-08,38,27,5,80,22,21
2006-07,38,28,5,83,27,16

This is the output of the print statement.

{'Season': array(['2011-12', '2010-11', '2009-10', '2008-09', '2007-08', '2006-07'],
      dtype=object), 'Matches Played': array([38, 38, 38, 38, 38, 38], dtype=int64), 'Wins': array([28, 23, 27, 28, 27, 28], dtype=int64), 'Losses': array([5, 4, 7, 4, 5, 5], dtype=int64), 'Goals': array([89, 78, 86, 68, 80, 83], dtype=int64), 'Goals Conceded': array([33, 37, 28, 24, 22, 27], dtype=int64), 'Clean Sheets': array([20, 15, 19, 24, 21, 16], dtype=int64), 'index': array([0, 1, 2, 3, 4, 5], dtype=int64)}

Thanks guys for the help. Seems to be working. :D

arka roy
– arka roy

2018-03-01 13:09:35 +00:00
Commented Mar 1, 2018 at 13:09 — arka roy
– arka roy, Commented Mar 1, 2018 at 13:09

bigreddot · Accepted Answer · 2018-02-19 22:23:46Z

4

Bokeh does not know what to do with those string dates unless you tell it. There are two basic possibilities:

Keep them as strings, and treat them as categorical factors. You can do that by telling Bokeh what the factors are when you create the plot:
```
p = figure(plot_width=400, plot_height=400, 
           x_range=list(overview_df.Season.unique()))
```
That results in this figure:

If you want a different order of categories you can re-order x_range however you like.
Convert them to real datetime values and use a datetime axis. You can do this by telling Pandas to parse column 0 as a date field:
```
overview_df = pd.read_csv(file, parse_dates=[0])
```
and telling Bokeh to use a datetime axis:
```
p = figure(plot_width=400, plot_height=400, x_axis_type="datetime")
```
That results in this figure:

answered Feb 19, 2018 at 22:23

bigreddot

34.8k5 gold badges73 silver badges128 bronze badges

Sign up to request clarification or add additional context in comments.

5 Comments

Abhinav Upadhyay Over a year ago

Was facing the same problem right now, thanks. I am not sure if it's too difficult but a warning from bokeh would have saved me some time. I had set the x_axis_type='datetime' but my x axis values were strings, had to do pd.to_datetime() and it worked.

bigreddot Over a year ago

It's not really possible to warn until things hit the browser, and the type is ultimately wrong in the place it is needed. Before then, there is no way to know what the ultimate type will be, i.e. you might have a datetime axis and strings, and a CustomJSTransform that turns the strings to dates in the browser. Warning about that would be incorrect, but it's not really possible to distinguish those cases. Perhaps we could make a better JS console message, but it would still just be a JS console message that most people won't even know to look for.

Abhinav Upadhyay Over a year ago

I understand. I just learned to use it yesterday and already feel productive with it. Are you a developer behind it? I want to say thank you for creating and maintaining it.

bigreddot Over a year ago

Yes, I am the project lead, thank you very much for the kind words

Joris Over a year ago

great solution. Note that you can get the same error when you pass a list of date objects instead of datetime objects. --> convert them to datetime and it works.

habet · Accepted Answer · 2018-02-19 22:31:13Z

1

you can convert the 'Season'-column to datetime to get an output.

overview_df = pd.read_csv(file)
overview_df.Season = pd.to_datetime(overview_df.Season)
overview_ds = ColumnDataSource(overview_df)

answered Feb 19, 2018 at 22:31

habet

14610 bronze badges

Collectives™ on Stack Overflow

Bokeh not displaying plot for pandas

2 Answers 2

5 Comments

Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

5 Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related