无法在Bokeh .rect字形中按颜色分组ColumnDataSource数据

Question

I'd like to group my plot by color based on a column that holds strings in a .csv file. I was under the impression that "color = 'ColumnName'" would work.

p.rect(x = range(len(source.data['Project Description'])),
   y = adj_h,
   color= 'Client',
   source = source,
   width = .4,
   height = 'Budget'     
  )

This just has my rectangle glyphs appear with black outlines and no fillcolor.

Answer 1

Colors in Bokeh are plotted using a color mapping scheme. First you need to set the colormap which allows Bokeh to know what color to use for which object. You can either manually set the colormap by using a dictionary colormap = {'Client_1':'red', Client_2:'blue'} or if you can use the built in CategoricalColorMapper function to create a colormap colormap = CategoricalColorMapper(palette=palette, factors=list_of_factors) .

Then you can specify the colors in the plotting function. Either you can use a list containing all the colors for all data points colors = [colormap[x] for x in source['Client']] and then pass the colors to the plot color=colors . Or you can let bokeh do the transformation for you with directly in the plot function color={'field': 'Client', 'transform': colormap}

---

In your code, I'd try something like

colormap = CategoricalColorMapper(palette=palette, factors=list(set(data['Client'])))

p.rect(x = range(len(source.data['Project Description'])),
   y = adj_h,
   color = {'field': 'Client', 'transform': colormap}',
   source = source,
   width = .4,
   height = 'Budget'     
  )

Where the palette is chosen depending on the number of categories and data is your original dataset.

---

References:

iris.py - Example on manually setting colors to each item in the dataset and using the list of colors in the plot.

texas.py - Example using the palette and color mapper and then letting the plotting function set the colors with 'transform' .