7

I'm trying to replicate the HeatMap shown in this question using Bokeh instead of matplotlib. I can't get it quite right though. The existing examples have not helped me to understand what I'm doing wrong. My humble attempt

from bokeh.io import output_notebook; output_notebook()
from bokeh.charts import HeatMap, show
from bokeh.palettes import RdYlGn6
import pandas as pd
import numpy as np

nba = pd.read_csv(urlopen("http://datasets.flowingdata.com/ppg2008.csv"), index_col=0)

# Normalize the data columns and sort.
nba = (nba - nba.mean()) / (nba.max() - nba.min())
nba.sort('PTS', inplace=True)

score = []
for x in nba.apply(tuple):
  score.extend(x)

data = {
  'players': list(nba.index) * len(nba.columns),
  'metric':  list(nba.columns) * len(nba.index),
  'score':   score,
}

hm = HeatMap(data, x='metric', y='players',values='score', title='Fruits', stat=None)
show(hm)

Gives

enter image description here

Note that, despite the similar title, this does not answer my question. It's not the same error and I'm using Bokeh 0.12.3

Community
  • 1
  • 1
Niklas R
  • 16,299
  • 28
  • 108
  • 203
  • The data aren't lined up correctly. You've repeated both `players` and `metric` list-wise, one of them needs to be repeated element-wise. – p-robot Dec 22 '16 at 12:20

2 Answers2

13

In case you still want to create a heatmap using Bokeh: the charts module was removed in more recent versions. In other words, this command will not work with newer versions of Bokeh:

from bokeh.charts import HeatMap

Since it gives the error:

ImportError: cannot import name 'charts'

Charts was moved to bkcharts package, which was than discontinued (further reference in in this answer). Holoviews still has some support for Bokeh, but has some different syntax.

A solution for creating Heatmaps in Bokeh is using p.rect() as instructed in this link about unemployment.py, which results in something like this:

p = figure()
hm = p.rect(data, x='metric', y='players',values='score', title='Fruits', stat=None)

Which yields a results that looks like this:

enter image description here

Guido
  • 6,182
  • 1
  • 29
  • 50
  • Note that the `values=` kwarg does not exist anymore since at least version `1.4.0`. In the updated "unemployment" example, a `LinearColorMapper` is used instead. – mbugert Feb 21 '22 at 15:58
3

Change the generation of the data metric to repeat element-wise and it should be correct:

'metric': [item for item in list(nba.columns) for i in range(len(nba.index))],

So the code that works for me is the following: 

from bokeh.charts import HeatMap, show, output_file
import pandas as pd, numpy as np
from urllib2 import urlopen

nba = pd.read_csv(urlopen("http://datasets.flowingdata.com/ppg2008.csv"), index_col=0)

# Normalize the data columns and sort.
nba = (nba - nba.mean()) / (nba.max() - nba.min())
nba.sort_values(by = 'PTS', inplace=True)

score = []
for x in nba.apply(tuple):
  score.extend(x)

data = {
  'players': list(nba.index) * len(nba.columns),
  'metric':  [item for item in list(nba.columns) for i in range(len(nba.index))],
  'score':   score,
}

output_file('test.html')
hm = HeatMap(data, x='metric', y='players',values='score', title='Fruits', stat=None)
show(hm)

enter image description here

p-robot
  • 4,652
  • 2
  • 29
  • 38
  • 1
    Ah, of course! Thanks! :-) Just as a side note, `list(nba.index)` is not necessary, at least here it works wihout. – Niklas R Dec 22 '16 at 13:49