using input from file to define colour of bar in plot (matplotlib)

Python's dicts can only be indexed with a single key. They don't allow "vectorized" indexing.

Let's use a simplified example that gives the same error:

import numpy as np

lookup = {'a':1, 'b':2, 'c':3}
values = np.array(['a', 'b', 'c', 'c', 'a', 'b', 'a'])
data = lookup[values]

Which yields:

TypeError                                 Traceback (most recent call last)
<ipython-input-71-7d8663a08b8d> in <module>()
      1 lookup = {'a':1, 'b':2, 'c':3}
      2 values = np.array(['a', 'b', 'c', 'c', 'a', 'b', 'a'])
----> 3 data = lookup[values]

TypeError: unhashable type: 'numpy.ndarray'

The exact error is because we're trying to use a mutable type as a key to the dict. There are sequences (e.g. tuples) that you could use as a key, but they still wouldn't work in the way that you want them to.

Therefore, instead of using a sequence of keys such as data = lookup[values], you'll need to use a list comprehension:

data = [lookup[item] for item in values]

Putting this back into your original example:

import numpy as np
import matplotlib.pyplot as plt

results_dtype=np.dtype([('name', 'S100'), ('colour', 'S10'),
('data_this', 'float64'), ('data_that', 'float64'), ...])


data = np.genfromtxt('C:/data.csv', delimiter = ',', dtype=results_dtype, 
                     filling_values=np.nan, skip_header=1)

colours = {'w':'#FFFFFF',
           'y':'#ffff00',
           'b':'#cc8033',
           'p':'#CC79A7'}

color = [colours[item] for item in data['colour']]

fig = plt.figure()
plt.bar(np.arange(len(data)), data['data_this'], bottom=data['data_that'],  
        align='center', color=color)
plt.xticks(np.arange(len(data)), data['name'], rotation='vertical')
plt.show()